Zbornik 20. mednarodne multikonference INFORMACIJSKA DRUŽBA - IS 2017 Zvezek I Proceedings of the 20th International Multiconference INFORMATION SOCIETY - IS 2017 Volume I Delavnica za elektronsko in mobilno zdravje ter pametna mesta Workshop Electronic and Mobile Healt and Smart Cities Uredila / Edited by Matjaž Gams, Aleš Tavčar http://is.ijs.si 9.–13. oktober 2017 / 9–13 October 2017 Ljubljana, Slovenia Zbornik 20. mednarodne multikonference INFORMACIJSKA DRUŽBA – IS 2017 Zvezek I Proceedings of the 20th International Multiconference INFORMATION SOCIETY – IS 2017 Volume I Delavnica za elektronsko in mobilno zdravje ter pametna mesta Workshop Electronic and Mobile Health and Smart Cities Uredila / Edited by Matjaž Gams, Aleš Tavčar http://is.ijs.si 9. - 13. oktober 2017 / 9th – 13th October 2017 Ljubljana, Slovenia Urednika: Matjaž Gams Odsek za inteligentne sisteme Institut »Jožef Stefan«, Ljubljana Aleš Tavčar Odsek za inteligentne sisteme Institut »Jožef Stefan«, Ljubljana Založnik: Institut »Jožef Stefan«, Ljubljana Priprava zbornika: Mitja Lasič, Vesna Lasič, Lana Zemljak Oblikovanje naslovnice: Vesna Lasič Dostop do e-publikacije: http://library.ijs.si/Stacks/Proceedings/InformationSociety Ljubljana, oktober 2017 Kataložni zapis o publikaciji (CIP) pripravili v Narodni in univerzitetni knjižnici v Ljubljani COBISS.SI-ID=39259653 ISBN 978-961-264-120-7 (pdf) PREDGOVOR MULTIKONFERENCI INFORMACIJSKA DRUŽBA 2017 Multikonferenca Informacijska družba (http://is.ijs.si) je z dvajseto zaporedno prireditvijo osrednji srednjeevropski dogodek na področju informacijske družbe, računalništva in informatike. Letošnja prireditev je ponovno na več lokacijah, osrednji dogodki pa so na Institutu »Jožef Stefan«. Informacijska družba, znanje in umetna inteligenca so spet na razpotju tako same zase kot glede vpliva na človeški razvoj. Se bo eksponentna rast elektronike po Moorovem zakonu nadaljevala ali stagnirala? Bo umetna inteligenca nadaljevala svoj neverjetni razvoj in premagovala ljudi na čedalje več področjih in s tem omogočila razcvet civilizacije, ali pa bo eksponentna rast prebivalstva zlasti v Afriki povzročila zadušitev rasti? Čedalje več pokazateljev kaže v oba ekstrema – da prehajamo v naslednje civilizacijsko obdobje, hkrati pa so planetarni konflikti sodobne družbe čedalje težje obvladljivi. Letos smo v multikonferenco povezali dvanajst odličnih neodvisnih konferenc. Predstavljenih bo okoli 200 predstavitev, povzetkov in referatov v okviru samostojnih konferenc in delavnic. Prireditev bodo spremljale okrogle mize in razprave ter posebni dogodki, kot je svečana podelitev nagrad. Izbrani prispevki bodo izšli tudi v posebni številki revije Informatica, ki se ponaša s 40-letno tradicijo odlične znanstvene revije. Odlične obletnice! Multikonferenco Informacijska družba 2017 sestavljajo naslednje samostojne konference:  Slovenska konferenca o umetni inteligenci  Soočanje z demografskimi izzivi  Kognitivna znanost  Sodelovanje, programska oprema in storitve v informacijski družbi  Izkopavanje znanja in podatkovna skladišča  Vzgoja in izobraževanje v informacijski družbi  Četrta študentska računalniška konferenca  Delavnica »EM-zdravje«  Peta mednarodna konferenca kognitonike  Mednarodna konferenca za prenos tehnologij - ITTC  Delavnica »AS-IT-IC«  Robotika Soorganizatorji in podporniki konference so različne raziskovalne institucije in združenja, med njimi tudi ACM Slovenija, SLAIS, DKZ in druga slovenska nacionalna akademija, Inženirska akademija Slovenije (IAS). V imenu organizatorjev konference se zahvaljujemo združenjem in inštitucijam, še posebej pa udeležencem za njihove dragocene prispevke in priložnost, da z nami delijo svoje izkušnje o informacijski družbi. Zahvaljujemo se tudi recenzentom za njihovo pomoč pri recenziranju. V 2017 bomo petič podelili nagrado za življenjske dosežke v čast Donalda Michija in Alana Turinga. Nagrado Michie-Turing za izjemen življenjski prispevek k razvoju in promociji informacijske družbe bo prejel prof. dr. Marjan Krisper. Priznanje za dosežek leta bo pripadlo prof. dr. Andreju Brodniku. Že šestič podeljujemo nagradi »informacijska limona« in »informacijska jagoda« za najbolj (ne)uspešne poteze v zvezi z informacijsko družbo. Limono je dobilo padanje slovenskih sredstev za akademsko znanost, tako da smo sedaj tretji najslabši po tem kriteriju v Evropi, jagodo pa »e-recept«. Čestitke nagrajencem! Bojan Orel, predsednik programskega odbora Matjaž Gams, predsednik organizacijskega odbora i FOREWORD - INFORMATION SOCIETY 2017 In its 20th year, the Information Society Multiconference (http://is.ijs.si) remains one of the leading conferences in Central Europe devoted to information society, computer science and informatics. In 2017 it is organized at various locations, with the main events at the Jožef Stefan Institute. The pace of progress of information society, knowledge and artificial intelligence is speeding up, and it seems we are again at a turning point. Will the progress of electronics continue according to the Moore’s law or will it start stagnating? Will AI continue to outperform humans at more and more activities and in this way enable the predicted unseen human progress, or will the growth of human population in particular in Africa cause global decline? Both extremes seem more and more likely – fantastic human progress and planetary decline caused by humans destroying our environment and each other. The Multiconference is running in parallel sessions with 200 presentations of scientific papers at twelve conferences, round tables, workshops and award ceremonies. Selected papers will be published in the Informatica journal, which has 40 years of tradition of excellent research publication. These are remarkable achievements. The Information Society 2017 Multiconference consists of the following conferences:  Slovenian Conference on Artificial Intelligence  Facing Demographic Challenges  Cognitive Science  Collaboration, Software and Services in Information Society  Data Mining and Data Warehouses  Education in Information Society  4th Student Computer Science Research Conference  Workshop Electronic and Mobile Health  5th International Conference on Cognitonics  International Conference of Transfer of Technologies - ITTC  Workshop »AC-IT-IC«  Robotics The Multiconference is co-organized and supported by several major research institutions and societies, among them ACM Slovenia, i.e. the Slovenian chapter of the ACM, SLAIS, DKZ and the second national engineering academy, the Slovenian Engineering Academy. In the name of the conference organizers we thank all the societies and institutions, and particularly all the participants for their valuable contribution and their interest in this event, and the reviewers for their thorough reviews. For the fifth year, the award for life-long outstanding contributions will be delivered in memory of Donald Michie and Alan Turing. The Michie-Turing award will be given to Prof. Marjan Krisper for his life-long outstanding contribution to the development and promotion of information society in our country. In addition, an award for current achievements will be given to Prof. Andrej Brodnik. The information lemon goes to national funding of the academic science, which degrades Slovenia to the third worst position in Europe. The information strawberry is awarded for the medical e-recipe project. Congratulations! Bojan Orel, Programme Committee Chair Matjaž Gams, Organizing Committee Chair ii KONFERENČNI ODBORI CONFERENCE COMMITTEES International Programme Committee Organizing Committee Vladimir Bajic, South Africa Matjaž Gams, chair Heiner Benking, Germany Mitja Luštrek Se Woo Cheon, South Korea Lana Zemljak Howie Firth, UK Vesna Koricki Olga Fomichova, Russia Mitja Lasič Vladimir Fomichov, Russia Robert Blatnik Vesna Hljuz Dobric, Croatia Aleš Tavčar Alfred Inselberg, Israel Blaž Mahnič Jay Liebowitz, USA Jure Šorn Huan Liu, Singapore Mario Konecki Henz Martin, Germany Marcin Paprzycki, USA Karl Pribram, USA Claude Sammut, Australia Jiri Wiedermann, Czech Republic Xindong Wu, USA Yiming Ye, USA Ning Zhong, USA Wray Buntine, Australia Bezalel Gavish, USA Gal A. Kaminka, Israel Mike Bain, Australia Michela Milano, Italy Derong Liu, Chicago, USA Toby Walsh, Australia Programme Committee Bojan Orel, chair Mitja Luštrek Niko Schlamberger Franc Solina, co-chair Marko Grobelnik Stanko Strmčnik Viljan Mahnič, co-chair Nikola Guid Jurij Šilc Cene Bavec, co-chair Marjan Heričko Jurij Tasič Tomaž Kalin, co-chair Borka Jerman Blažič Džonova Denis Trček Jozsef Györkös, co-chair Gorazd Kandus Andrej Ule Tadej Bajd Urban Kordeš Tanja Urbančič Jaroslav Berce Marjan Krisper Boštjan Vilfan Mojca Bernik Andrej Kuščer Baldomir Zajc Marko Bohanec Jadran Lenarčič Blaž Zupan Ivan Bratko Borut Likar Boris Žemva Andrej Brodnik Janez Malačič Leon Žlajpah Dušan Caf Olga Markič Saša Divjak Dunja Mladenič Tomaž Erjavec Franc Novak Bogdan Filipič Vladislav Rajkovič Andrej Gams Grega Repovš Matjaž Gams Ivan Rozman iii Invited lecture AN UPDATE FROM THE AI & MUSIC FRONT Gerhard Widmer Institute for Computational Perception Johannes Kepler University Linz (JKU), and Austrian Research Institute for Artificial Intelligence (OFAI), Vienna Abstract Much of current research in Artificial Intelligence and Music, and particularly in the field of Music Information Retrieval (MIR), focuses on algorithms that interpret musical signals and recognize musically relevant objects and patterns at various levels -- from notes to beats and rhythm, to melodic and harmonic patterns and higher-level segment structure --, with the goal of supporting novel applications in the digital music world. This presentation will give the audience a glimpse of what musically "intelligent" systems can currently do with music, and what this is good for. However, we will also find that while some of these capabilities are quite impressive, they are still far from (and do not require) a deeper "understanding" of music. An ongoing project will be presented that aims to take AI & music research a bit closer to the "essence" of music, going beyond surface features and focusing on the expressive aspects of music, and how these are communicated in music. This raises a number of new research challenges for the field of AI and Music (discussed in much more detail in [Widmer, 2016]). As a first step, we will look at recent work on computational models of expressive music performance, and will show some examples of the state of the art (including the result of a recent musical 'Turing test'). References Widmer, G. (2016). Getting Closer to the Essence of Music: The Con Espressione Manifesto. ACM Transactions on Intelligent Systems and Technology 8(2), Article 19. iv KAZALO / TABLE OF CONTENTS Delavnica za elektronsko in mobilno zdravje ter pametna mesta / Workshop Electronic and Mobile Health and Smart Cities ........................................................................................................................................ 1 PREDGOVOR / FOREWORD ................................................................................................................................. 3 PROGRAMSKI ODBORI / PROGRAMME COMMITTEES ..................................................................................... 5 Integracija v EMZ EkoSMART / Gams Matjaž, Tavčar Aleš .................................................................................. 7 Razvoj intervencij na platformi eOskrba / Cukjati Iztok, Bon Jure, Kališnik Jurij Matija, Žibert Janez, Pirtošek Zvezdan .............................................................................................................................................. 10 Visual Working Memory and its impairments in Parkinson’s disease / Pileckyte Indre, Bon Jure, Pirtošek Zvezdan ............................................................................................................................................................ 13 Open Source C++ Libraries for Electrophysiological Data Preprocessing and Analysis / Perellón Alfonso Ruben, Bon Jure, Pirtošek Zvezdan ................................................................................................................. 16 Using brain state-dependent transcranial magnetic stimulation for investigating causal role of cortical oscil ations in functional states / Matkovič Andraž, Bon Jure, Pirtošek Zvezdan ............................................ 20 Diagnosticiranje parkinsonove bolezni iz glasu osebe / Zupanc Andrej, Bratko Ivan .......................................... 24 Arhitektura sistema za oddaljeno spremljanje pacientov / Janković Marko, Žitnik Slavko, Bajec Marko ............ 27 Razvoj zapestnice za pomoč starejšim / Kompara Tomaž................................................................................... 29 Aplikacija tehnologije BLE za avtomatično zaznavanje prisotnosti oseb in predmetov / Planina Andrej, Vidmar Luka ..................................................................................................................................................... 32 Technology for training and assessment of precise movements in persons with Parkinson’s disease / Cikajlo Imre, Matjačić Zlatko, Burger Helena, Peterlin Potisk Karmen ............................................................ 34 Smart Dentistry and Smart Tooth Brushing / Kokol Peter, Colnarić Matjaž, Završnik Jernej, Zorman Milan, Verber Domen, Turčin Marko, Žlahtič Bojan, Moraus Stanislav, Jurič Simon, Žlahtič Grega, Slemik Bojan ..................................................................................................................................................... 36 Prototipi aplikacij za prenos mobilnih EKG meritev od uporabnika senzorja do zdravnika / Pavliha Denis, Planinc Nataša, Depol i Matjaž, Smokvina Aleš ............................................................................................... 39 Particle Accelerators as Medical Devices / Mehle Marko, Kurnjek Luka ............................................................. 41 Mobilno spremljanje okoljskih dejavnikov in njihovega vpliva na zdravje / Golija Andrej ..................................... 44 Application for Viral Hepatitis Infection Risk Assessment / Ajanović Alen, Počivavšek Karolina, Podpadec Matic, Ulčar Andrej, Peterlin Ana Marija, Prodan Ana, Rink Saša, Gradišek Anton, Gams Matjaž, Fele Žorž Gašper, Matičič Mojca ......................................................................................................... 46 Spletno svetovanje študentom v stiskah / Šef Tomaž, Tavčar Aleš, Mlakar Miha, Gams Matjaž........................ 49 Upgrade of AH-Model with Machine Learning Algorithms / Molan Gregor, Molan Martin ................................... 53 Wireless Sensor Prototype for Industrial Harsh Environments / Pavlin Marko, Poklukar Špela, Papa Gregor, Novak Franc ........................................................................................................................................ 57 Napovedovanje časovnih vrst za podporo energetski optimizaciji stavbe / Zupančič Domen ............................. 61 Avtomatizacija in digitalizacija vrta / Colarič Luka ................................................................................................ 65 Indeks avtorjev / Author index ................................................................................................................................ 69 v vi Zbornik 20. mednarodne multikonference INFORMACIJSKA DRUŽBA – IS 2017 Zvezek I Proceedings of the 20th International Multiconference INFORMATION SOCIETY – IS 2017 Volume I Delavnica za elektronsko in mobilno zdravje ter pametna mesta Workshop Electronic and Mobile Health and Smart Cities Uredila / Edited by Matjaž Gams, Aleš Tavčar http://is.ijs.si 10. oktober 2017 / 10th October 2017 Ljubljana, Slovenia 1 2 PREDGOVOR V letu 2017 smo pripravili tretjo delavnico na temo »e&m-zdravstva« (elektronsko in mobilno zdravstvo, kratko EMZ), tj. predlog izvedbe infrastrukture in vpeljave uporabe informacijsko in mobilno podprte celostne zdravstvene oskrbe za izboljševanje preventivne, diagnostične in terapevtske obravnave državljanov, ki bi zmanjšala stroške, obenem pa povečala dostopnost zdravstvene oskrbe v obdobju 2016-2020. V letu 2016 je bil sprejet Raziskovalno Razvojni in Inovacijski (RRI) program EkoSMART v domeni pametne specializacije S4 na področju pametnih mest in skupnosti, kjer EMZ predstavlja enega od šestih nosilnih stebrov programa v obliki RRP (Raziskovalno-Razvojnega Projekta). V okviru javnega razpisa »RRI v verigah in mrežah vrednosti« − sklop 1: »Spodbujanje izvajanja raziskovalno-razvojnih programov (TRL 3-6)« je predvidenih 5,9 milijona evrov nepovratnih javnih sredstev za program EkoSMART. Projekt EMZ sestavlja 5 delovnih sklopov oziroma delovnih paketov (DP), ki jih vodijo UKCL, IJS, FERI UM in FRI UL:  Informacijske tehnologije za podporo celostni oskrbi / bolnice / prof. Z. Pirtošek  Podpora na domu za zdrave, starejše in za kronične bolnike / doma / prof. M. Gams  Mobilno spremljanje vitalnih in okolijskih podatkov / mobilno / dr. R. Trobec  Računalniška podpora, podatki, kreiranje novih znanj /algoritmi / prof. P. Kokol  IKT platforma / prof. M. Bajec Delavnica EMZ omogoča celoletno pregledovanje in usklajevanje sklopa EMZ znotraj programa EkoSMART. Podobno kot v letu 2016 se bodo partnerji javno predstavili vsem drugim z že precej usklajenim predlogom. Vse predstavitve bomo nato dokončno uskladili in pripravili specifikacijo dela na programu za naslednje obdobje (sep. 2017 – avg. 2018). Potrebno se je zavedati, da je prvo leto dela že za nami in da je potrebno pregledati in predvsem povezati prispevke v smiselno celoto. Povezovali jih bomo najprej znotraj delovnih sklopov (delovnih paketov), nato znotraj RRP EMZ, nato pa še znotraj celotnega programa EkoSMART. Prvi povezovalni okvir je narejen za vse slovenske občine na ui-obcine.ijs.si, kjer sta tudi repozitorija prototipov in domen. Na delavnici bomo vse prispevke poskušali povezati z omenjenim okvirjem. Pobudo e&m-zdravstvo so vzpodbudile potrebe po horizontalnem in vertikalnem povezovanju, trendi in dileme področja. Predlagana pobuda e&m-zdravje vpeljuje v zdravstveno oskrbo nove koncepte, ki bodo s svojimi multiplikacijskimi in sinergijskimi učinki sprožili hitrejšo in učinkovitejšo prilagoditev obstoječega sistema celostne zdravstvene oskrbe na današnje izzive. Ključna strokovna komponenta je umetna inteligenca, ki bo po napovedih strokovnjakov revolucionirala zdravstvo skupaj z novimi IKT rešitvami. Javno zdravstvo po vsem svetu se otepa izrednih problemov, najboljšo rešitev pa strokovnjaki po svetu vidijo v vpeljavi storitev IKT in umetne inteligence. E&m-zdravstvo (EMZ) vidimo kot eno najbolj perspektivnih smeri v več pobudah od zdravstva do pametnih mest. E&m-storitve nudijo izboljšano kvaliteto življenja državljanom ob zmanjšanih stroških, hkrati pa omogočajo preboj Slovenije v svet na e&m-področju. E&m-zdravstvo se bo predvidoma vsebinsko oblikovalo delno kot samostojna pobuda s svojo platformo, organizacijo in projekti, ki bo povezana tako s pametnimi mesti kot z zdravjem. Ključne komponente za uspešno izvedbo EMZ so inovativni človeški viri, njihovo usklajeno delovanje in vpeljava EMZ v Sloveniji. 3 Amerika generira dvakrat več pomembnih inovacij v zdravstvu kot EU ter vlaga štirikrat več sredstev v nova, z medicino povezana podjetja. Kitajska namenja največ sredstev za znanost, medtem ko je Slovenija tretja najslabša po državnem financiranju znanosti v Evropi. Leta 2025 bo več kot milijarda, ali skoraj osmina svetovnega prebivalstva, starejša od 60 let. Stroški za zdravstveno oskrbo starejše populacije predstavljajo v EU skoraj polovico vseh stroškov za zdravstvo, kar pomeni, da grozi zdravstvenemu in gospodarskemu sistemu in kvaliteti življenja zlom, če ne bomo vpeljali storitev e&m-zdravja. Druga pomembna komponenta je povezovanje in ustvarjanje kritične mase komplementarnih partnerjev, ki edino omogoča uspešen prodor na svetovna tržišča. Slovenija potrebuje sodelovanje in koordiniranje že zaradi svoje relativne majhnosti, kar dokazuje relativno slaba izkušnja z velikim številom malih in razdrobljenih projektov, ki nimajo dovolj podpore za vpeljavo novih rešitev. Tretja ključna komponenta je vpeljava EMZ v slovensko zdravstvo, ki bo na ta način dobilo novo priložnost, da vzpostavi nacionalno platformo in mednarodne standarde, preseže ujetost v nedopustno dolge čakalne dobe za pregled pri specialistih, poveže razdrobljene in nekompatibilne sisteme in že samo s tem opraviči vložena sredstva. Po zadnjem povečanju sredstev za področje zdravstva so se čakalne vrste povečale, kar kaže, da sedanji tradicionalni pristop ne zmore prinesti realnih izboljšav. Matjaž Gams, Aleš Tavčar 4 PROGRAMSKI ODBOR / PROGRAMME COMMITTEE Matjaž Gams (chair) Marko Bajec (co-chair) Roman Trobec (co-chair) Zvezdan Pirtošek (co-chair) Roland Petek Jure Bon Peter Kokol Andrej Kos Marko Hren Aleš Tavčar Stanislav Erzar Janez Uplaznik 5 6 Integracija v EMZ EkoSMART Matjaž Gams, Aleš Tavčar Jozef Stefan Institute Jamova 39, 1000 Ljubljana, Slovenia matjaz.gams@ijs.si POVZETEK za vse občane - da imajo več informacij, storitev in podpore. V prispevku je na kratko predstavljen modul Elektronsko in Umetna inteligenca in IKT napredujeta neverjetno hitro, raziskovalni oddelki razvijajo fantastične nove sisteme, spotoma mobilno zdravje (EMZ) znotraj programa EkoSMART. Sledijo pa nastajajo tudi sistemi, ki jih lahko namestijo in vzdržujejo navodila za sestavljanje prispevkov in partnerjev v integriran inštitucije lokalne samouprave in civilne družbe. projekt. Sistem sestavljajo naslednji bloki: Ključne besede  Občinska televizija – vsakdo lahko razvije svojo občinsko Elektronsko in mobilno zdravstvo, pametna mesta, pametna televizijo s sledenjem navodilom. Potreben je prenosnik in specializacija kamera in nekaj znanja računalništva. Običajno občinska TV prenaša ali sprotno dogajanje v živo, ali pa se veti 1. UVOD predpripravljena datoteka s tekočimi informacijami za tekoči teden. V letu 2016 je bil sprejet Raziskovalno Razvojni in Inovacijski  3D virtualni asistent – ponovno je z nekaj znanja (RRI) program EkoSMART [3] v domeni pametne specializacije računalništva po navodilih možno izdelati sistem 3D S4 na področju pametnih mest in skupnosti, kjer EMZ predstavlja virtualnega asistenta, ki vodi po stavbah, recimo upravni enega od šestih nosilnih stebrov programa v obliki RRP stavbi občine. (Raziskovalno-Razvojnega Projekta). V okviru javnega razpisa »RRI v verigah in mrežah vrednosti« − sklop 1: »Spodbujanje  Turizem – sistem omogoča informiranj o turističnih izvajanja raziskovalno-razvojnih programov (TRL 3-6)« je znamenitostih v naravnem jeziku in načrtovanje turističnih predvidenih 5,9 milijona evrov nepovratnih javnih sredstev za obiskov. Sistem vsebuje preko 3000 znamenitosti in je program EkoSMART. neposredno uporaben.  Asistenti – za vsako slovensko občino je narejen svoj občinski asistent, ki odgovarja na vprašanja v naravnem Projekt EMZ sestavlja 5 delovnih sklopov oziroma delovnih paketov (DP), ki jih vodijo UKCL, IJS, FERI UM in FRI UL: jeziku. Obstaja tudi pokrajinski asistent in slovenski asistent – slednja sta sestavljena iz pripadajočih občinskih asistentov. 1. Informacijske tehnologije za podporo celostni oskrbi / Obstaja pa tudi asistent za starejše občane (zdusko) in za bolnice / prof. dr. Z. Pirtošek EkoSMART. Narejeni sta dve interaktivni mapi Slovenije, Podpora aktivnostim v bolnici. prva z občinami in druga z društvi upokojencev. Ko kliknete 2. Podpora na domu za zdrave, starejše in za kronične na mapo, se vzpostavi povezava z občino oz. društvom. bolnike / doma / prof. dr. M. Gams Podpora predvsem starejšim doma. Partnerji EkoSMART so povabljeni, da se vključujejo v obstoječe 3. Mobilno spremljanje vitalnih in okolijskih podatkov / bloke oz. dodajajo nove. mobilno / prof. dr. R. Trobec Za EMZ pa je najpomembnejši blok Zdravje. Senzorji. 4. Računalniška podpora, podatki, kreiranje novih znanj /algoritmi / prof. dr. P. Kokol 3. EMZ IN ZDRAVJE Navezava na dentalno zdravstvo. 5. IKT platforma / prof. dr. M. Bajec Blok »zdravje« v okviru občin (ui-obcine-ijs.si) nudi informacije Platforma in povezovanje z EkoSMART. o prvi pomoči, zdravstvene nasvete doma in iz tujine, informacije iz NIJZ, iz programa pametne specializacije EkoSMART ter V letu 2016 smo zastavili povezovanje preko bele knjige EMZ podprojekta Elektronsko in mobilno zdravje, repozitorijev domen [1], v letu 2017 gremo na delavnici EMZ korak dlje. in prototipov, sistemov za nadziranje stresa in skrb za starejše. Storitev je dostopna v asistentih – kliknete na svojo občino, levo 2. EMZ IN OBČINE zgoraj, izberete aplikacijo »Zdravje«. V nadaljevanju so našteti podsistemi in kako se vključevati v njih: Razvili smo ogrodje in nekaj sistemov [2] (ui-obcine.ijs.si) občin Prva pomoč: Tu dobite nasvete v primeru nujne pomoči kot korak dlje od pametnih mest in oboje sistematično skušamo (Mobilno Android IOS). Sistem zna do neke mere tudi sam vpeljati v občine, društva upokojencev in druga društva ter v civilno družbo z namenom, da Slovenija ponudi boljšo izkušnjo odgovarjati na enostavna vprašanja tipa »zlomil sem si nogo«. 7 Integracija: dodati nove storitve prve pomoči. Zdravstveni nasveti slovensko: Če imate zdravstvene težave, je pametno obiskati zdravnika, neodvisno od tega pa lahko pogledate, kaj pravijo strokovnjaki na spletu. Integracija: dodati nove storitve slovenskih zdravstvenih nasvetov. Tuji zdravstveni sistemi za pomoč, svetovanje in drugo mnenje: Uptodate health.com webmd.com 10 najboljših aplikacij zdravstvenih svetovalnic NIJZ: Tu je zbrana vrsta koristnih storitev za zdravje. Slika 1: Del izpisa vsebine repozitorija domen. Integracija: povezovati storitve EMZ in NIJZ. PROJEKTI IN LIFE: s projektom IN LIFE želimo omogočiti starejšim z opešanimi kognitivnimi sposobnostmi bolj samostojno življenje. Če vas zanima testiranje sistema, pišite na jani.bizjak (at) ijs.si E-gibalec: aplikacija za mobilne telefone, ki je bila razvita z namenom približati osnovnošolcem športne aktivnosti in jih spodbuditi k več gibanja. ASPO: spletna aplikacija za prepoznavanje in informiranje o spolno prenosljivih okužbah. Zaznavanje stresa: cilj študentskega projekta OSVET je spletni pogovorni svetovalec (chatbot) za zaznavanje stresa pri uporabnikih in nudenje psihosocialne pomoči preko spleta. Uporabnik lahko izpolni tudi anketo, ki izračuna stopnjo stresa. Aplikacija je še v razvoju. EkoSmart, EMZ: namen programa EkoSmart je razviti ekosistem pametnega mesta. V okviru EMZ (Elektronsko in Mobilno Zdravje) zbiramo repozitorije domen in prototipov, tako lahko vidite, kdo v Sloveniji hrani katere podatke in katere prototipe. Tu razvijamo asistenta za EMZ. Integracija: dodajati projekte partnerjev EMZ. Slika 2: Del izpisa vsebine repozitorija prototipov. Obstajata repozitorija domen in prototipov v obliki spletnih aplikacij. Integracija: dodajati nove prototipe in domene v repozitorija. 4. ZAKLJUČEK Narejen je okvir in znotraj tega vrsta storitev, kamor bodo partnerji EMZ in EkoSMART dodajali svoje storitve, sisteme in dosežke. Če ni jasno, kako se doda oz. integrira, potem je najbolje poslati elektronsko pošto avtorjema tega prispevka. 8 5. ZAHVALA [2] Ogrodje pametnih občin Ui-obcine.ijs.si Raziskave/delo je delno sofinancirano s strani Ministrstva za [3] Program EkoSMART izobraževanje, znanost in šport in Evropske unije iz Evropskega http://ekosmart.net/sl/o-projektu/ sklada za regionalni razvoj (ESRR). 6. REFERENCES [1] Bela knjiga EMZ EkoSMART https://dis.ijs.si/?p=2057 9 Razvoj intervencij na platformi eOskrba Iztok Cukjati Jure Bon Jurij Matija Kališnik Klinični oddelek za bolezni živčevja Klinični oddelek za bolezni živčevja Klin. oddelek za kirurgijo srca in ožilja Univerzitetni klinični center Ljubljana Univerzitetni klinični center Ljubljana Univerzitetni klinični center Ljubljana Zaloška 2, 1000 Ljubljana Zaloška 2, 1000 Ljubljana Zaloška 2, 1000 Ljubljana +38640186268 jure.bon@kclj.si jmkalisnik@gmail.com iztok.cukjati@upr.si Janez Žibert Zvezdan Pirtošek Univerza v Ljubljani Klinični oddelek za bolezni živčevja Zdravstvena fakulteta Univerzitetni klinični center Ljubljana Zdravstvena pot 5, 1000 Ljubljana Zaloška 2, 1000 Ljubljana janez.zibert@zf.uni-lj.si zvezdan.pirtosek@kclj.si POVZETEK vpeljava novih informacijskih in komunikacijskih tehnologij Slovenija in celotna Evropa se srečujeta z demografskimi (IKT). spremembami z izrazitim staranjem prebivalstva in nizko Pri vpeljavi IKT v zdravstveno okolje je med drugim ključno: rodnostjo. Skladno s povečevanjem števila in deleža starejših se povečuje število bolnikov z nenalezljivimi kroničnimi boleznimi - Uporaba že obstoječih tehnologij oz. razvoj novih, ko je (NKB), ki predstavljajo največji delež sredstev zdravstvenega to potrebno/smiselno. sistema. Učinkovito obvladovanje NKB bo v prihodnje ključno za - Zagotavljanje semantične interoperabilnosti podatkov vzdržnost zdravstvenih sistemov. Večja vključenost pacientov v med posameznimi deležniki v zdravstvu. proces zdravljenja skupaj z uporabo sodobnih tehnologij - Aktivno vključevanje medicinske stroke in pacientov že predstavlja osrednjo izhodišče za dosego finančno učinkovitejšega v fazah izdelave z IKT podprtih procesov zdravstvene sistema. oskrbe. V projektu EkoSmart bomo na raziskovalni platformi eOskrba - Zagotavljanje finančne učinkovitosti vpeljevanja novih / razvili in klinično validirali nove intervencije, ki bodo namenjene prilagojenih storitev, tudi s podporo odprtokodnih tako osebju in bolnikom v klinični praksi oz. posameznim rešitev. raziskovalnim skupinam za potrebe znanstveno-raziskovalnega V različnih zdravstvenih sistemih v različnih državah, tudi v dela. Tako bomo v sodelovanju Univerzitetnega kliničnega centra Sloveniji, so v preteklih letih pilotno vpeljali in klinično validirali Ljubljana (UKC), Medicinske fakultete (MF) in Fakultete za že večje število IKT podprtih kliničnih poti (intervencij). Tako je računalništvo in informatiko Univerze v Ljubljani (FRI) razvili bila v Sloveniji med prvimi dokazovana tako klinična [3] kot pilotni sistem za spremljanje EKG in ostalih parametrov pri finančna [4] učinkovitost z IKT podprte intervencije za bolnike z kardiovaskularnih bolnikih v postoperativni fazi. Sistem bo depresijo. omogočal tudi medsebojno primerjavo različnih senzorjev in s Korak naprej v razvoju tehnološko naprednih rešitev za tem njihovo klinično validacijo. implementacijo zdravstvenih intervencij predstavlja odprtokodna rešitev eOskrba, kjer so bile na enotni platformi razvite in pilotno Ključne besede preizkušane različne intervencije (eAstma, eDiabetes, eHujšanje) EkoSmart, e-zdravje, interoperabilnost, Parkinsonova bolezen, na različnih nivojih zdravstvene oskrbe (primarna, sekundarna, srčno-žilne bolezni, diabetes, astma, debelost. terciarna); učinkovitost intervencij pa je bila tudi klinično validirana [5]. V okviru projekta EkoSmart se na osnovi platforme eOskrba razvijajo nove intervencije, tudi z vpeljavo zbiranja in analize 1. IZHODIŠČA večjega števila podatkov iz različnih senzorjev. Zelo pomembno Staranje prebivalstva je v t.i. zahodnem svetu prisotno in mesto pri razvoju novih intervencij bomo namenili (z)možnosti neizogibno, poglavitna razloga pa sta nizka rodnost in daljša sistema za avtomatizirano pošiljanje podatkov v nacionalno življenjska doba. Tako naj bi se po napovedih Eurostata delež informacijsko hrbtenico in posledično dostopnost starejših od 65 let povečal do leta 2040 s sedanjih 19% na 27% (anonimiziranih) podatkov vsem (tudi raziskovalno usmerjenim) celotnega prebivalstva [1]. Ob povečevanju števila (in še bolj zainteresiranim uporabnikom. izrazito deleža) starejših prebivalcev se sočasno povečuje tudi število bolnikov z NKB, ki že danes predstavljajo kar 70% vseh 1.1 Tehnologija vzrokov smrtnosti [2]. Ročno merjenje in beleženje posameznih zdravstvenih podatkov, Obvladovanje stroškov naraščujočih kroničnih bolezni kot so ne glede na mesto njihovega nastanka (pacient doma ali astma, diabetes, bolezni srca in ožilja in nevrodegenerativne zdravstveni delavec v ustanovi) je še vedno zelo prisotno. bolezni, je bistveno za dolgoročno vzdržnost zdravstvenega Avtomatizirano odčitavanje in (brezžično) elektronsko beleženje sistema, k čemur lahko pomembno prispeva dobro načrtovana podatkov je postalo mnogo učinkovitejše (hitrejše, cenejše) od 10 prej omenjenega načina. Tako lahko z relativno majhnimi sredstvi 2. EKOSMART IN RAZVOJ NOVIH uporabimo tehnologije, ki so že na voljo, npr. mobilni telefon, ki INTERVENCIJ lahko v procesu zdravstvene oskrbe deluje kot: V okviru projekta EkoSmart razvijamo dve intervenciji, ki bosta - zaslon za spremljanje (zdravstvenih) podatkov, meritev, namenjeni kardiovaskularnim bolnikom v postoperativni video posnetkov; bolnišnični obravnavi in bolnikom s parkinsonovo boleznijo, oz. - vmesnik za brezžični zajem podatkov iz specializiranih zdravstvenim delavcem pri vodenju zdravstvene oskrbe medicinskih naprav; omenjenih skupin bolnikov. - lokalno skladišče podatkov; - komunikacijska naprava s spletom; 2.1 eOskrba – (tudi) raziskovalna platforma - zajem podatkov kot so: lokacija, smer in hitrost gibanja V preteklih letih se je v slovenskem prostoru že validirala (GPS); pospeški in pojemki (pospeškometer), slika in vpeljava novih intervencij z uporabo IKT storitev na področju video (kamera), zvok (mikrofon). oskrbe bolnikov s NKB. Tako je bila med drugim razvita odprtokodna platforma eOskrba, na kateri bomo v okviru projekta Posebno prednost predstavljajo operacijski sistemi (iOS, Android), ki omogočajo praktično neomejen razvoj in uporabo EkoSmart razvili dve novi intervenciji. Poleg tega se bo razvil vmesnik za pošiljanje podatkov v nacionalno bazo podatkov o mobilnih aplikacij. pacientu. Razviti intervenciji bosta vključevali tudi uporabo sodobnih, brezžičnih senzorjev. 1.2 Izmenljivost informacij 2.2 Postoperativno vodenje Uporaba IKT orodij omogoča hitro in zanesljivo pridobivanje kliničnih in ostalih podatkov oz. informacij, pogosto ključnih za kardiovaskularnih bolnikov obravnavo pacientov. Hiter razvoj naprednih senzorjev z Spremljanje kardiovaskularnih bolnikov v postoperativni fazi je brezžičnim in avtomatiziranim prenosom podatkov na mobilne izrazito pomembno zaradi pojava morebitnih zapletov, med naprave nudi številne možnosti razvoja naprednih rešitev za drugim tudi atrijske fibrilacije. Z uporabo EKG naprav, ki so uporabnike: zdravstveno osebje, paciente in ostale deležnike v lahko bodisi prenosne bodisi stacionarne, se lahko doseže procesu zdravljenja. relativno dober nadzor nad bolniki. Težavo predstavljata visoka cena prenosnih EKG merilcev (t.i. holter aprati) in neažurnost Čeprav so posamezne informacije o pacientu s pomočjo omenjene podatkov – prenosne EKG naprave praviloma ne pošiljajo tehnologije za diagnostiko (in spremljanje terapije) že zelo podatkov v realnem času na strežnik, ampak se podatki hranijo uporabne, pa še vedno obstaja izrazito ozko grlo zaradi lokalne lokalno v napravi in jih zdravniki odčitajo ob ponovnem prihodu v (proizvajalec naprave) hrambe posameznih podatkov. Prav tako bolnišnico. tehnološka podjetja praviloma t.i. »surove podatke« obdelujejo in končnemu uporabniku ponujajo le t.i. »obdelane podatke«. Predvidevamo, da bi uporaba cenejše naprave, ki omogoča prenos Neusklajenost med posameznimi ponudniki in različnimi podatkov »skoraj v realnem času« lahko pomembno hitreje informacijskimi sistemi se rešuje z vpeljavo standardov, ki zaznala morebitne zaplete in na ta način preprečila marsikateri omogočajo semantično interoperabilnost - bistveno za enoznačno postoperativni zaplet. interpretacijo podatka/informacije v različnih informacijskih V okviru projekta razvijamo celovito storitev, ki med drugim sistemih [6]. omogoča sočasno spremljanje/beleženje podatkov iz obstoječih V projektu EkoSmart bomo uporabili standard OpenEHR, standardnih EKG naprav in mobilnega EKG senzorja Savvy, z posamezne podatke pa se bo pošiljalo v nacionalno bazo pri NIJZ; namenom primerjave pridobljenih podatkov in izboljšave klinične s čimer bodo podatki na voljo ostalim zaintereiranim oskrbe. uporabnikom, anonimizirani podatki pa bodo na voljo tudi za 2.3 Vodenje bolnikov s Parkinsonovo raziskovalne namene. boleznijo 1.3 Vključevanje uporabnikov Parkinsonova bolezen je ena izmed najpogostejših oblik Nove storitve so namenjene uporabnikom, ki so v zdravstvenem nevroloških bolezni, zaradi staranja prebivalstva pa se število sistemu praviloma medicinsko osebje, pacienti in (v nekaterih bolnikov hitro povečuje. Tako naj bi se število bolnikov od leta primerih) njihovi svojci. Uspešnost vpeljave novih storitev je v 2010 do 2040 podvojilo, sami stroški zdravljenja in izgube zaradi mnogočem odvisna od dejstva, kako bodo nove storitve bolezni so (za ZDA) ocenjeni na 22.800 USD na pacienta [7]. zadovoljevale pričakovanja in zahteve zdravnikov, medicinskih sester, bolnikov, svojcev itd. Za učinkovit razvoj in uporabo z V okviru projekta razvijamo spletno-mobilno aplikacijo za IKT podprtih zdravstvenih storitev je tako ključno sodelovanje spremljanje bolnikov s parkinsonovo boleznijo, predvsem z namenom aktivne uporabe (tudi v raziskovalne namene) beleženih zgoraj omenjenih skupin uporabnikov od prvih faz (t.j. načrtovanje storitev) do končne validac podatkov. Kasneje predvidevamo integracijo s specifičnimi ije. senzorji, ki se uporabljajo za spremljanje zdravstvenega stanja 1.4 Finančni vidik bolnikov in v prihodnosti razvoj celovite storitve, tako za Inovativna dejavnost na področju zdravstvene oskrbe s pomočjo bolnišnično oskrbo, kot tudi spremljanje pacientov na domu. vpeljave IKT rešitev je lahko finančno učinkovita, vendar pa to ni samoumevno. Pravilno načrtovanje razvoja, vpeljave in validacije 3. REFERENCE posameznih novosti je bistveno za izboljšanje zdravstvenega [1] http://ec.europa.eu/eurostat/statistics- stanja (kroničnih) bolnikov, ki ima hkrati tudi finančno ugodne explained/index.php/Population_structure_and_ageing posledice (cenejše zdravljenje). [2] http://www.who.int/mediacentre/factsheets/fs355/en/ Spremljanje finančnega učinka bomo uvedli tudi pri razvoju novih [3] Meglic M, Furlan M, Kuzmanic M, Kozel D, Baraga intervencij. D,Kuhar I, Kosir B, Iljaz R, Novak Sarotar B, Dernovsek 11 MZ, Marusic A, Eysenbach G. Brodnik. A. Feasibility of an Controlled Trial in Slovenia. Zdr Varst. 2017 May eHealth Service to Support Collaborative Depression Care: 26;56(3):150-157. Results of a Pilot Study. J Med Internet Res, 12(5):e63, 2010. [6] Beštek, M., Brodnik, A. Interoperability and mHealth – [4] Naveršnik K, Mrhar A. Routine real-time cost-effectiveness precondition for successful eCare. Mobile Health: A monitoring of a web-based depression intervention: a risk- Technology Road Map, 2015: Chapter 16; Springer. sharing proposal. J Med Internet Res. 2014 Feb 27;16(2):e67. [7] Kowal SL, Dall TM, Chakrabarti R, Storm MV, Jain A. The [5] Iljaž R, Brodnik A, Zrimec T, Cukjati I. E-healthcare for current and projected economic burden of Parkinson's Diabetes Mellitus Type 2 Patients - A Randomised disease in the United States. Mov Disord. 2013. Mar;28(3):311-8 12 Visual Working Memory and its impairments in Parkinson’s disease Indre Pileckyte Jure Bon Zvezdan Pirtošek Clinical Department of Neurology Clinical Department of Neurology Clinical Department of Neurology University Medical Centre Ljubljana University Medical Centre Ljubljana University Medical Centre Ljubljana Zaloška cesta 2, Ljubljana Zaloška cesta 2, Ljubljana Zaloška cesta 2, Ljubljana indre.pileckyte@gmail.com jure.bon@kclj.si zvezdan.pirtosek@kclj.si ABSTRACT Visual Working Memory (vWM) is a set of processes enabling 1.1.2. Behavioral vWM capacity quantification short-term visual information maintenance and manipulation. Methodological developments in the past twenty years have One of the most prominent and researched properties of vWM is encouraged more standardized and reliable ways to quantify vWM its highly limited capacity. Different theoretical models and capacity and probe its sub-processes. Visual WM is recognized as frameworks have always recognized that short-term storage and a bottleneck in information processing system due to its highly control mechanisms are the bottleneck in the human information limited capacity, yet, its central role in human cognition makes it a processing system. However, despite the long history of research, key research topic in understanding cognitive deficits and the precise quantification of vWM capacity had varied greatly impairments in clinical populations. Parkinson’s disease (PD) is from one study to another due to the general lack of standardized one of the neurodegenerative disorders marked by vWM measurement procedures. The turn point in the field was marked impairments. Considering that vWM impairments start at the early by a seminal paper of S. Luck and E. Vogel in 1997 [9]. Through stages of PD, behavioral and electrophysiological vWM capacity a set of influential experiments, the authors introduced a game- measurements could act as important early disease biomarkers. changing discrete-slot framework of vWM capacity, as well as This work was carried out for the purposes of developing the popularized a change-detection paradigm and a behavioral vWM peripheral sensing techniques for EkoSMART project. capacity index K. According to their discrete-slot model, vWM capacity is limited General Terms to, on average, 3-4 memory representations. As a result, in case of Experimentation, Human Factors, Theory a supra-threshold memory set, only a fraction of information (3-4 representations) can be retained in vWM [5, 9]. Moreover, the individual differences in memory capacity, ranging from 1.5 to 6 Keywords items, are stable both across different stimulus modalities and Visual Working memory, Parkinson’s disease time. In addition, S. Luck and E. Vogel popularized a change-detection 1. INTRODUCTION paradigm for quantification of vWM capacity which remains one of the dominant experimental paradigms in the field. Essentially, 1.1.1. Visual working memory definition in a change detection paradigm the observer is presented with a memory set for a brief period of time (typically 100-200 ms) and Visual Working memory (vWM) is a capacity-limited system after a short period of retention (typically around 900 ms) he is responsible for the short-term maintenance and manipulation of shown a probe and has to indicate whether a change has occurred visual information [4]. It is an integrative part of the visual [9]. There are two variants of the change-detection task: A single- information processing system, responsible for such functions as probe version requires the subject to make a judgment about one integration of pre-saccadic and post-saccadic retinal images and item, while a whole-display version requires to make a decision binocular integration [10]. Furthermore, the functional role of regarding any element in the whole memory set. The change visual WM expands beyond perceptual and processing roles. It detection paradigm has been used and established for a variety of has been repeatedly and reliably shown that vWM underlies a visual features, such as color, shape, size or orientation [9], as range of higher cognitive functions, such as fluid intelligence, well as complex visual stimuli [6]. Although the change-detection reasoning, language comprehension and math abilities [11], acting task is very simple and straightforward, it has been shown to as an online workspace for keeping information “in mind” [7]. correlate well with a range of more complex tasks for fluid Growing recognition of the vWM role in healthy and impaired intelligence and high level aptitude tests [11]. cognition has resulted in a substantial increase in studies and Finally, S. Luck and Vogel have popularized a behavioral publications related to the topic: In 2014 there were 18.224 capacity index K for standardized vWM capacity measurement. citations in PubMed and 1.580.000 search results in Google The capacity index K is derived from accuracy parameters, such Scholar related to working memory, whereas in 2017 the number as false alarm rate and hit rate, and adjusted to the memory set grew to 42.474 PubMed citations and 4.750.000 Google Scholar size and the version of change detection task (single-probe or search results [3]. whole-display) [12]. The index K has enabled reliable 13 quantification of vWM capacity across different stimuli and task modifications. Figure 2. Example of a CDA waveform at memory load of 2 (red line) and 4 (blue line), based on real experimental data Figure 1. Example of simplified bilateral change-detection task from healthy participants design 1.1.4. Visual WM deficits in Parkinson’s disease 1.1.3. Electrophysiological vWM capacity marker One of the clinical populations marked by vWM deficits comprise It has been long observed that visual WM tasks are accompanied Parkinson’s disease (PD) patients. PD is a neurodegenerative by a slow negative wave at temporal and occipital scalp sites [4]. disorder characterized by a range of motor symptoms, such as However, the productive use of this observation was impeded by resting tremor, rigidity, and bradykinesia [2]. The motor difficulties in separating visual working memory processes from symptoms, however, are accompanied by cognitive deficits in those of visual perception, general effort or task difficulty. The executive functions and goal-directed behavior [2, 8]. It is known ground-breaking change in the electrophysiological study of that both the motor and cognitive symptoms are caused by vWM came from E. Vogel and M. G. Machizawa [13] with the dopaminergic deficit in basal ganglia, as L-dopa administration introduction of a modified bilateral variant of the change- has been shown to ameliorate performance deficits in cognitive detection paradigm (Figure 1). The modified task differed from tasks, both in terms of accuracy and reaction times [2]. the original one in a directional cue preceding the memory set, resulting in vWM processing being focused on one visual Visual WM deficits in PD were first defined with regard to its hemifield. This method, known as contralateral control method, reduced capacity, however, development of the CDA allowed can be applied to any lateralized brain system, such as visual or more accurate dissection of the vWM pathology in PD. E. Y. Lee motor, in order to isolate specific processes. and colleagues [8] used a bilateral whole-display change- detection task to quantify vWM performance in 21 idiopathic PD Consequently, E. Vogel and M. G. Machizawa observed a slow patient and 28 healthy controls. The behavioral capacity index K negative wave larger at the contralateral (to the stimuli) temporal and CDA provided convincing evidence that Parkinson’s disease and posterior electrode sites. They used a subtraction method to patients have both reduced vWM capacity and impaired filtering construct a difference wave and termed it the Contralateral Delay efficiency. The authors hypothesized that the loss of dopaminergic Activity (CDA). They immediately noticed that the CDA has a input to the basal ganglia could lead to reduced function of globus great load sensitivity, which is reflected by its amplitude changes pallidus in regulating task-relevant information storage in vWM. (Figure 2). Moreover, the CDA has been shown to mirror the They also suggested that bradyphrenia could cause the observed same interpersonal vWM capacity differences as those reflected reduced capacity and filtering efficiency, as patients could not by the behavioral index K [13, 14]. Importantly, while the index deal efficiently with the rapid pace of the task. K is a summary measurement of the overall performance (accuracy), the CDA allows online probing into different vWM Visual WM studies in PD patients are important not just for sub-processes, such as encoding, maintenance and retrieval. obtaining a better insight into the underlying pathology. P. S. Boggio with colleagues [1] have reported a significant As a result, it has been shown that individual differences in vWM improvement in WM task accuracy after active anodal tDCS capacity in healthy adults can be explained by the differences in stimulation over lDLPFC with 2mA, opening a discussion about the resistance to distractors rather than the capacity itself. E. possible therapeutic cognitive rehabilitation in PD. Vogel and colleagues [14] termed this vWM property Filtering Efficiency (FE). Essentially, FE indicates how well a person can 2. CONCLUSIONS resist encoding task-irrelevant distractors and use his highly Improved methodology in behavioral and electrophysiological limited vWM capacity efficiently. It has been since shown that FE quantification of visual working memory allows more accurate deficits are present in many neurodegenerative and psychiatric understanding of vWM impairments in PD patients. Considering disorders characterized by some kind of vWM impairments. that vWM impairments start at the early stages of PD, behavioral and electrophysiological vWM capacity measurements could act as important early disease biomarkers. 14 3. ACKNOWLEDGEMENTS (2003). Gamma oscillations correlate with working memory This work has been supported by EkoSMART consortium. load in humans. Cerebral cortex, 13(12), 1369-1374. [8] Lee, E. Y., Cowan, N., Vogel, E. K., Rolan, T., Valle-Inclan, F., & Hackley, S. A. (2010). Visual working memory deficits in patients with Parkinson's disease are due to both reduced 4. REFERENCES storage capacity and impaired ability to filter out irrelevant [1] Boggio, P. S., Ferrucci, R., Rigonatti, S. P., Covre, P., information. Brain, 133(9), 2677-2689. Nitsche, M., Pascual-Leone, A., & Fregni, F. (2006). Effects of transcranial direct current stimulation on working memory [9] Luck, S. J., & Vogel, E. K. (1997). The capacity of visual in patients with Parkinson's disease. Journal of the working memory for features and conjunctions. Nature, neurological sciences, 249(1), 31-38. 390(6657), 279-281. [2] Costa, A., Peppe, A., Dell’Agnello, G., Carlesimo, G. A., [10] Luck, S. J., & Vogel, E. K. (2013). Visual working memory Murri, L., Bonuccelli, U., & Caltagirone, C. (2003). capacity: from psychophysics and neurobiology to individual Dopaminergic modulation of visual-spatial working memory differences. Trends in cognitive sciences, 17(8), 391-400. in Parkinson’s disease. Dementia and geriatric cognitive [11] Luria, R., Balaban, H., Awh, E., & Vogel, E. K. (2016). The disorders, 15(2), 55-66. contralateral delay activity as a neural measure of visual [3] D'esposito, M., & Postle, B. R. (2015). The cognitive working memory. Neuroscience & Biobehavioral Reviews, neuroscience of working memory. Annual review of 62, 100-108. psychology, 66. [12] Rouder, J. N., Morey, R. D., Morey, C. C., & Cowan, N. [4] Drew, T. W., McCollough, A. W., & Vogel, E. K. (2006). (2011). How to measure working memory capacity in the Event-related potential measures of visual working memory. change detection paradigm. Psychonomic Bulletin & Clinical EEG and Neuroscience, 37(4), 286-291. Review, 18(2), 324-330. [5] Fukuda, K., Awh, E., & Vogel, E. K. (2010). Discrete [13] Vogel, E. K., & Machizawa, M. G. (2004). Neural activity capacity limits in visual working memory. Current opinion in predicts individual differences in visual working memory neurobiology, 20(2), 177-18 capacity. Nature, 428(6984), 748-751. [6] Gao, Z., Li, J., Liang, J., Chen, H., Yin, J., & Shen, M. [14] Vogel, E. K., McCollough, A. W., & Machizawa, M. G. (2009). Storing fine detailed information in visual working (2005). Neural measures reveal individual differences in memory—Evidence from event-related potentials. Journal of controlling access to working memory. Nature, 438(7067), Vision, 9(7), 17-17. 500-503 [7] Howard, M. W., Rizzuto, D. S., Caplan, J. B., Madsen, J. R., Lisman, J., Aschenbrenner-Scheibe, R., ... & Kahana, M. J. 15 Open Source C++ Libraries for Electrophysiological Data Preprocessing and Analysis Ruben Perellón Alfonso Jure Bon Zvezdan Pirtošek Clinical Department of Neurology Clinical Department of Neurology Clinical Department of Neurology University Medical Centre Ljubljana University Medical Centre Ljubljana University Medical Centre Ljubljana Zaloška cesta 2, Ljubljana Zaloška cesta 2, Ljubljana Zaloška cesta 2, Ljubljana ruben.palfonso@gmail.com jure.bon@kclj.si zvezdan.pirtosek@kclj.si ABSTRACT lack of specific solutions for the analysis of EMG data obtained in In this paper, we describe two open source C++ libraries the context of non-invasive electric and magnetic transcranial developed to facilitate and automate preprocessing and analysis of stimulation, which is a well stablished neuroscientific research electromyography (EMG) and electroencephalography (EEG) data method for studying and altering brain function non-invasively (palMEP and palEEG, respectively). Additionally, we introduce a [4]. Additionally, well-established software solutions for the graphical user interface (GUI) software for EMG data analysis. analysis of EEG and MEG data, such as EEGLAB or FieldTrip, These tools fill existing gaps in open source data preprocessing are mostly developed and implemented within the Matlab and analysis solutions in the field of electrophysiology and neuro- (MathWorks, Inc. Chicago) programing environment, which is a stimulation. They are aimed at enabling research and its leading commercial software in scientific computing. While reproducibility and, at the same time, facilitate the transition of Matlab is commonly available in research institutions throughout experimental clinical therapeutic methods from an open loop the world, many research groups and individuals are often unable brain stimulation paradigm towards an adaptive more to afford the licence fees associated with commercial software personalized closed loop paradigm. This work was carried out for packages, and are, therefore, forced to operate outside the the purposes of developing the peripheral sensing techniques for mainstream of preprocessing and analysis methods. EkoSMART project. First, the purpose of developed tools will be The aim of the work in progress that the libraries here presented introduced in the context of neuroscientific research in general constitute, is to fill the two aforementioned gaps. On the one and non-invasive neuro-stimulation in particular. Secondly, the hand, we have developed an open source solution, named inner workings of the libraries will be thoroughly described. palMEP, to perform preprocessing and analysis of EMG data in Finally, current work in progress and future directions in general, and of data that has been gathered while using electric or development will be discussed. magnetic stimulation, in particular. The methods and functionality of palMEP have been developed according to standard practices Keywords in the research community and reflect the idiosyncrasies and EMG, EEG, Neurostimulation, FFT, open source. challenges that researchers in the field of neuro-stimulation typically face when dealing with EMG data gathered concurrently with methods such as Transcranial Magnetic Stimulation (TMS) 1. INTRODUCTION or peripheral electric stimulation. On the other hand, we are Open source solutions for preprocessing and analysis of currently developing a homologue of the EEGLAB and FieldTrip electrophysiological and neuroimaging data are now mainstream toolboxes, named palEEG, which will not be dependent on a in neuroscientific research. Publically available toolboxes, such as commercial software development framework. At the moment, EEGLAB[1], FieldTrip [2], or SPM[3], have gained major palEEG includes an interface to import Matlab based files adhesion from the international research community. The rapid containing workspace variables in general and EEGLAB and pace at which new preprocessing and analysis methods appear as ERPLAB data structures in particular. palEEG also currently well as the dynamic nature of their implementation made it implements an experimental implementation of time-frequency somewhat necessary that toolboxes be created within the open decomposition using Morlet wavelets. source community so that the software would be up to date with the ongoing progress in research methodologies. Moreover, the open source model enables and promotes the interaction and 2. palMEP exchange of information among scientist in the neuroscientific In this section we describe the philosophy and implementation of community. This, in turn, allows the seamless implementation of an open source GUI based software for the preprocessing and cutting edge methods on common universally available platforms, analysis of EMG data. which is then reflected in evermore standardized data 2.1. Intended usage and Scope preprocessing and analysis pipelines and, hence, higher This software is aimed at students, researchers and clinicians in transparency and replicability of research methods and results. electrophysiology and neuroscience, who have little or no While there is an abundance of well-established open source EEG experience in programming and/or in signal processing and and neuroimaging data preprocessing and analysis tools, there is a analysis. This program is also intended for users who do have this 16 knowledge, but would benefit from an easy to use tool for 1) The peak to peak amplitude: manipulation, visualization and automation of motor evoked 2) The natural logarithm of the peak to peak amplitude: potentials (MEPs) analysis and preprocessing tasks. ln( ). To this end, the user interface is designed to be intuitive and 3) The root mean square amplitude [9]: simple, stripped down to the most necessary elements to reach the most commonly sought after goal: to get reliably preprocessed . Where is the MEP measures in a format that is friendly to mainstream statistical squared value of each datum of EMG within the data analysis softwares, such as SPSS© or R[5]. window [10]. The measures of choice included in this program, peak to peak, log and root mean square amplitude, are the most widely used to 4) The latency in milliseconds: . Where describe MEPs elicited by transcranial magnetic and electric is the index of a data point in a time series. For stimulation. When using these stimulation methods, a large inter- unknown sampling rates (delimited text files), the trial variability is to be expected [6], hence, visual inspection of latency is expressed in number of data points: single trials might be necessary in order to reject MEPs which . might have been influenced by unwanted background muscle contraction [7], or which are far too large or small. This software 5) The sampling state, as defined in the original Signal file, provides a clear and intuitive signal viewer for trial by trial visual or as computed from a marker channel in a Spike file. exploration. Additionally, mean and standard deviation are also For delimited text files, state is always 0. calculated, in order to allow further analysis outside this software. 6) The name of the data channel as described in the This software supports the commonly used CED (Cambridge original Signal or Sipke file. For delimited text files Electronic Design Limited, Cambridge, UK) Signal and Spike2 ‘Channel_ID’ is always ‘unknown’. software files, as well as comma, space or tab delimited text files, which further extends the compatibility of this program to any 7) The full path of the processed file. other software which supports custom text file export. This software is distributed, together with the source code, under the GNU GPLv3 licence. The source code is fully documented according to the Doxygen standard. The software also includes a detailed user manual in pdf format. 2.2. Implementation and Functionality palMEP is implemented in C++ using the QT (The Qt Company Ltd) programing environment. This software depends on the CED Figure 2. palMEP results output sample. CFS library and the CEDS64ML interface libraries, for reading Signal and Spike files, respectively. Additionally, it depends on An innovative feature of this software is an algorithm designed to the QCustomPlot library by Emanuel Eichhammer for signal detect the MEP automatically, without any user input. This allows plotting and visualization. All these dependencies are publically automatic processing of an unlimited number of files without available and routinely maintained. previous visual inspection. The algorithm implements two methods, and if they do not coincide, it picks the solution of the palMEP has two main modes of operation: single-trial based one which provided the smaller result (discrepancy is usually due processing, and automatic processing of single or multiple files. to one method detecting the artifact and the other detecting the In single-trial mode, the user can easily highlight the area of MEP). The first method finds the maximum difference between interest where measures will be performed in for each trial using each pair of consecutive data points in the sweep. The second two interactive vertical cursor bars (see Figure 1). method estimates the rate of change by iteratively calculating the first derivative in 5 data point segments, then the segment with the maximum rate of change is taken. Figure 1. palMEP GUI layout. palMEP computes the following measures and metadata of interest, and stores them in a tab delimited text file of the users choosing (see Figure 2 for a sample of the output): 17 1) ‘read_2dmat’, allows reading a single vector or 2d matrix (rown x coln) from a .mat Matlab file. 2) ‘read_3dmat’, allows reading a 3d matrix (rown x coln x slicen) from a .mat Matlab file. 3) ‘read_EEG’, allows reading an EEGLAB EEG data structure contained in either a .set or .mat data file. 4) ‘read_EEG’, allows reading an ERPLAB ERP data structure contained in either a .erp or .mat data file. 3. palEEG All 4 reading functions are numeric type sensitive (integer, single In this section we describe the philosophy and implementation of or double precision floating point numbers), and will convert the an open source library for importing Matlab based variables, as values to double precision, when appropriate. The functions carry well as EEGLAB and ERPLAB [8] data structures, and out sufficient error checking to avoid crashes and will return performing time-frecuency decomposition using Morlet wavelets. success or failure accordingly. 3.1. Intended usage and Scope Funcitons 3 and 4 store the read data into an eegData or erpData structure, respectively. These data structures are designed to keep This library is work in progress, and is intended to become a the structure of the data as close as possible to the original EEG or homologue alternative to the Matlab based EEGLAB, ERPLAB ERP structure in Matlab, while only retaining the essential and FieldTrip toolboxes. At the moment, the library includes a set information for further processing in palEEG. A simplified of useful functions and, in the near future, will include a GUI declaration of the 2 data structures would be as follows: similar to palMEP to enable easy access to EEG preprocessing and analysis methods and data visualization. This software is distributed, together with the source code, under the GNU GPLv3 licence. 3.2. Implementation and Functionality palEEG is implemented in C++ using the QT programing environment. This software depends on the Armadillo linear algebra library [9] for matrix operations and manipulations. We have chosen Armadillo (namespace: ‘arma’), because it is well- The ‘eventStruc’ stores all the relevant marker and event established and maintained, it has good performance and provides information from the original Matlab event structure, retaining the a syntax and functionality similar to Matlab. For reading Matlab subfield names for easier portability. files, palEEG depends on the MAT File I/O Library by Christopher C. Hulbert. We have chosen this library because it constitutes a decent open source alternative to the proprietary library provided by Matlab. For computing discrete Fourier transforms (DFT) palEEG depends on the Fastest Fourier Transform in the West (FFTW) library [10]. This library has been chosen because it is possibly the fastest implementation of the fast Fourier transform (FFT) algorithm, and is also the library used by Matlab to compute FFT. The time-frequency decomposition functions are declared and implemented in ‘dotf.h’ and ‘dotf.c’, respectively. FFTW uses the following formulas for the FFT and its inverse: The main function is declared as follows: It takes the imported data and first computes the event-related potential (ERP) for any chosen experimental condition and EEG channel. It then performs the FFT of the vectorized original data. Next, it creates the Morlet Wavelet [11], according to the ‘cycles’ parameter defined by the user, and computes its FFT. The FFT of the data is then element-wise multiplied by the FFT of the palEEG currently includes 4 functions for loading Matlab files, wavelet. Finally, the inverse FFT of the resulting multiplication is these functions are declared and implemented in ‘readmat.h’ and computed. This results in a vector of complex elements containing ‘readmat.c’, respectively: power and phase information for all the chosen time points in each of the specified frequencies. The resulting vector is then reshaped into 3 different 3d matrices (channel x Frequency x 18 Time), containing the raw complex-number result of the 6. REFERENCES decomposition, the power and the ITPC results, respectively, and [1] Delorme, A., & Makeig, S. (2004). EEGLAB: an open stored into a ‘tfData’ structure, which is defined as follows: source toolbox for analysis of single-trial EEG dynamics including independent component analysis. Journal of neuroscience methods, 134(1), 9-21. [2] Oostenveld, R., Fries, P., Maris, E., & Schoffelen, J. M. (2011). FieldTrip: open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data. Computational intelligence and neuroscience, 2011, 1. [3] Penny, W. D., Friston, K. J., Ashburner, J. T., Kiebel, S. J., & Nichols, T. E. (Eds.). (2011). Statistical parametric The ‘ mapping: the analysis of functional brain images. Academic do_TFeeg’ function supports three different modes of press. operation, that can be specified via the ‘type’ parameter: [4] Wassermann, E., Epstein, C., & Ziemann, U. (Eds.). (2008). 1) ‘Absolute Power’, computes the total power (phase- Oxford handbook of transcranial stimulation. Oxford locked + non-phase-locked power) and the inter-trial University Press. Chicago. phase coherence. [5] Team, R. C. (2014). R: A language and environment for 2) ‘Phase locked’ Power, computes the power of the ERP statistical computing. Vienna, Austria: R Foundation for only. Statistical Computing; 2014. 3) ‘Non-phase-locked Power’, subtracts the ERP from [6] Kiers, L., Cros, D., Chiappa, K. H., & Fang, J. (1993). each individual trial in the original data before running Variability of motor potentials evoked by transcranial time-frequency decomposition. magnetic stimulation. Electroencephalography and Clinical Neurophysiology/Evoked Potentials Section, 89(6), 415-423. The different time-frequency measures follow the definitions by [7] Darling, W. G., Wolf, S. L., & Butler, A. J. (2006). Brian J. Roach and Daniel H. Mathalon [12], while the algorithms Variability of motor potentials evoked by transcranial used to compute time-frequency decomposition closely follow magnetic stimulation depends on muscle activation. Mike X. Cohen [13]. Experimental brain research, 174(2), 376-385. Subsequent iterations of palEEG will include a GUI, additional [8] Lopez-Calderon, J., & Luck, S. J. (2014). ERPLAB: an data preprocessing tools, such as filtering and ICA based artifact open-source toolbox for the analysis of event-related rejection, new data and result visualization methods, as well as potentials. Frontiers in human neuroscience, 8. time-frequency based connectivity measures. [9] Sanderson, C., & Curtin, R. (2016). Armadillo: a template- 4. CONCLUDING REMARKS based C++ library for linear algebra. Journal of Open Source The libraries here described constitute work in progress meant to Software. fill existing gaps in open source data preprocessing and analysis [10] Frigo, M., & Johnson, S. G. (2005). The design and solutions in the field of electrophysiology and neuro-stimulation. implementation of FFTW3. Proceedings of the IEEE, 93(2), These tools are aimed at enabling research and its reproducibility 216-231. and, at the same time, facilitate the transition of experimental [11] Goupillaud, P., Grossmann, A., & Morlet, J. (1984). Cycle- clinical methods from an open loop paradigm towards an adaptive octave and related transforms in seismic signal analysis. more personalized closed loop paradigm, which we hope to Geoexploration, 23(1), 85-102. achieve by developing fully open source optimized solutions for data preprocessing and analysis. [12] Roach, B. J., & Mathalon, D. H. (2008). Event-related EEG time-frequency analysis: an overview of measures and an If you would be interested in either getting a copy or collaborating analysis of early gamma band phase locking in in the development of these libraries, contact the author of this schizophrenia. Schizophrenia bulletin, 34(5), 907-926. paper. [13] Cohen, M. X. (2014). Analyzing neural time series data: 5. ACKNOWLEDGEMENTS theory and practice. MIT Press. This work has been supported by EkoSMART consortium. 19 Using brain state-dependent transcranial magnetic stimulation for investigating causal role of cortical oscillations in functional states Andraž Matkovič Jure Bon Zvezdan Pirtošek 1Clinical Department of Neurology 1Clinical Department of Neurology 1Clinical Department of Neurology University Medical Centre Ljubljana University Medical Centre Ljubljana University Medical Centre Ljubljana Zaloška cesta 2, Ljubljana Zaloška cesta 2, Ljubljana Zaloška cesta 2, Ljubljana 2Department of Psychology jure.bon@kclj.si zvezdan.pirtosek@kclj.si Faculty of Arts Aškerčeva 2, Ljubljana andraz.matkovic@ff.uni-lj.si ABSTRACT thereby inducing action potentials in neurons. It is a common Non-invasive brain stimulation is being used for manipulation of technique for manipulating neuronal oscillations and is used for cortical oscillations in research and clinical context for research and clinical purposes. Transcranial alternate or direct development of possible therapeutic applications in brain current stimulation (TACS / TDCS) is a similar, although much disorders. Effects of brain stimulation show strong inter- and weaker technique for modulating neural activity. intra-individual variation. In general there are several sources of Effects of the TMS treatment show strong inter- and intra- this variability, e.g. neuroanatomical and neurochemical factors. individual variability and are influenced by neuroanatomical, This article describes our work in the scope of EkoSMART neurochemical and neurophysiological factors [10]. These factors consortium on development of peripheral sensing techniques. can be trait-related and can be stable (e.g. cortical thickness, Here we focus on the rapidly varying neurophysiological factors – individual alpha rhythm frequency) or can vary intra-individually cortical oscillations. Current state of cortical oscillations can be (e.g. circadian fluctuations). On the other hand, state-related continually recorded with peripheral sensors like EEG scalp determinants can vary strongly and rapidly within and between electrodes and used online for continuous monitoring and treatment sessions. For example, it has been shown that phase and adjustments of brain stimulation parameters. By adjusting the amplitude of cortical oscillations influence corticospinal timing, intensity and frequency of transcranial stimulation to excitability as measured with motor evoked responses [1, 15]. In specific brain states it is possible to reduce variation in the the case of trait-related factors, variability of TMS intervention treatment effects. However, since brain state-dependent can be reduced by pre-selection of individuals based on a certain stimulation (BSDS) requires online monitoring and analysis of trait or by homogenizing the influencing variable, e.g. applying neurophysiological data, it is technically demanding. BSDS has the treatment at the same time of the day. However, cortical been made possible by recent technological advances and oscillations can also vary on a millisecond scale and to reduce the advances in analytical procedures. While EEG data has been effects of these factors, brain-state dependent stimulation is traditionally analyzed in time- or frequency-domain only, time- needed. frequency analysis is being increasingly used and it offers better insight into neurophysiology of oscillations. BSDS is useful in the 2. FEASIBILITY OF BRAIN-STATE field of clinical neuroscience, where it can be used to personalize stimulation parameters, e.g. adjust deep brain stimulation DEPENDENT STIMULATION depending on the severity in symptoms in Parkinson’s disease. Brain-state dependent stimulation (BSDS) requires online Because it enables manipulation of cortical oscillations when a monitoring and analysis of neurophysiological data (see Figure 1). specific brain state is detected, it allows stronger causal inferences Karabanov et al. [10] distinguish between (1) state-informed about their role in behavior and brain states. Therefore, BSDS can noninvasive transcranial stimulation (NTBS) and (2) adaptive, be also used as a tool for verification or falsification of hypotheses closed-loop1 NTBS. In the former, the timing, frequency or the in cognitive neuroscience. intensity of stimulation is adjusted according to the predefined state (e.g. phase or power of cortical oscillations), whereas in the Keywords latter, stimulation is dynamically adjusted depending on the stimulation-induced state changes. Brain state-dependent stimulation, transcranial magnetic stimulation, time-frequency analysis, electroencephalography, cortical oscillations. 1 In the literature, the term closed-loop stimulation is sometimes being used for both types of stimulation. However, Karabanov 1. INTRODUCTION et al. [10] emphasize, that state-informed stimulation is not the Transcranial magnetic stimulation (TMS) is a non-invasive brain same as “closed-loop” stimulation and that the latter term stimulation method during which magnetic field in a coil induces should only be used for a stimulation which adapts depending an electric current in nearby conductive tissue – the brain – on stimulation effects in real-time. 20 In recent years several studies have shown feasibility of both types spatial resolution and sensitivity to subcortical structures is of stimulation. An example of state-informed NTBS is a study by important, whereas EEG is more appropriate when timing Bergmann et al. [2]. To answer the question how the phases of precision on sub-second level is more relevant. cortical oscillations affect cortical excitability, Bergmann et al. applied TMS during sleep while concurrently measuring electroencephalographic (EEG) signal. Single-pulse TMS was triggered by automatic detection of up- and down-states in slow- oscillations during non-rapid eye movement sleep. It was shown that motor-evoked potentials (MEPs) and TMS-evoked potentials (TEPs) were larger during slow-oscillations up-states than during down-states. Similarly, Gharabaghi et al. [6] and Kraus et al. [11] showed that single-pulse TMS controlled by beta-band event- related desynchronization (ERD) during motor imagery resulted in an increase of corticospinal excitability whereas in the non- BSDS condition this effect was absent. An example of an adaptive closed-loop stimulation is a study by Brittain et al. [4] who applied transcranial alternating current stimulation (TACS) over the motor cortex of patients with Parkinson’s disease. Stimulation was delivered at tremor- frequency and adjusted in a way to produce phase-cancellation and thus achieving tremor-suppression up to 50%. Similarly, in a study by Little et al. [12] it was shown that deep brain stimulation in Parkinson’s disease can be adjusted by providing feedback from local field potentials from the electrodes. This type of adaptive deep brain stimulation was more effective and efficacious than conventional continuous stimulation. While these studies provide proof-of-principle, it remains to be shown that EEG combined with non-invasive transcranial stimulation can also be used in a closed-loop, adaptive fashion [2]. Figure 2. Same data analysed in time-domain (event-related potential, top) and with time-frequency analysis (bottom). 2.1 Challenges in online real-time analysis of brain-states Traditionally, EEG data has been analyzed in time- or frequency- domain. In time domain analyses (see Figure 2, top), data are typically averaged across epochs based on markers, which represent events (e.g. stimulus presentation or participants’ response) by which event-related potentials (ERPs) are observed. This procedure is based on an assumption that signal is constant across trials, whereas all trial-to-trial variability is considered to be noise. One major disadvantage of ERP technique which originates from this assumption is that it requires a lot of trials to achieve acceptable signal-to-noise ratio. While this procedure is still being used, it is now known that this assumption is false since even temporally fluctuating potentials (so-called non-phase- Figure 1. Closed-loop brain state-dependent non-invasive locked or induced activity) are averaged out from the signal and transcranial brain stimulation. Source: Bergmann et al., therefore cannot be reliably detected with ERP technique. 2016 [3]. Originally published under CC BY license (https://creativecommons.org/licenses/by/4.0/). In frequency-domain analyses, EEG data is decomposed from Both neuroimaging (e.g. functional magnetic resonance) and time-domain into frequency-domain using Fourier transform. electrophysiological methods (e.g. EEG) can be coupled with With this procedure it can easily be estimated which frequencies TMS for BSDS. TMS can be combined with fMRI for brain-states constitute EEG signal, e.g. whether there is more activity in theta- with slow fluctuations (e.g. resting state connectivity) and when band (4–7 Hz), beta-band (16–30 Hz), etc. This type of analysis is 21 relatively simple and therefore widely used, e.g. for research on using BSDS, where oscillations in gamma and theta spectra would resting state EEG. However, the fact that brain is highly non- be manipulated using transcranial brain stimulation. Further, stationary renders results of frequency-domain analyses hard to using BSDT, if low theta or gamma power were detected during interpret, especially when we are interested in how oscillations stimuli presentation, these stimuli could then be presented change in response to events. The third way to analyze EEG data multiple times and in this way learning would be more efficient. is by means of time-frequency analysis. Time-frequency analysis offers good time and frequency resolution (although there exists a 4. CONCLUSION trade-off between them) [5]. Its results are closer to actual Computational advances and advances in statistical methods have neurophysiology in comparison to the other two methods and in recent years enabled analysis of trial-to-trial variations in the therefore easier to interpret. Since it does not require a large field of neurophysiology. This has led to new hypotheses about number of trials to achieve acceptable signal-to-noise ratio and the functional role of cortical oscillations. Brain state-dependent since it can be used to disentangle phase-locked and non-phase- stimulation can be used as a tool for testing these hypotheses and locked activity, it is more appropriate for single-trial analyses. It thus enables making strong causal inferences about cortical also enables calculation of various connectivity measures based oscillations. BSDS is a significant step towards optimizing non- on phase, power, etc. invasive transcranial brain stimulation interventions, thus enabling more efficient stimulation adapted to the individual’s Although time-frequency analysis is more computationally brain and/or functional state. To conclude, research on brain state- consuming, it is more suitable for BSDS in comparison to time- or dependent stimulation is still in its infancy and shows frequency-domain analyses. Besides usefulness in online considerable promise in fostering progress in cognitive and monitoring of cortical oscillations in real-time, time-frequency clinical neuroscience. analysis can also be used to generate hypotheses about the causal role of different types of oscillations and for evaluating the effects 5. ACKNOWLEDGEMENTS of transcranial stimulation. Thanks to Tine Kolenik for checking the grammar. Another important issue in BSDS with TMS are strong artefacts This work has been supported by EkoSMART consortium. produced by stimulation lasting several milliseconds [16] and TEPs caused by stimulation. In state-informed open-loop 6. REFERENCES stimulation a refractory period of several seconds can be used to avoid triggering of TMS by artefacts or TEPs, whereas for closed- [1] Berger, B. et al. 2014. EEG oscillatory phase-dependent loop stimulation methods for online artefact reduction yet need to markers of corticospinal excitability in the resting brain. be developed. BioMed Research International. 2014, (2014). DOI:https://doi.org/10.1155/2014/936096. 3. BRAIN STATE-DEPENDENT [2] Bergmann, T.O. et al. 2012. EEG-guided transcranial STIMULATION ENABLES STRONGER magnetic stimulation reveals rapid shifts in motor cortical CAUSAL INFERENCES ABOUT excitability during the human sleep slow oscillation. Journal CORTICAL OSCILLATIONS of Neuroscience. 32, 1 (2012), 243–253. DOI:https://doi.org/10.1523/JNEUROSCI.4792-11.2012. Whereas the usefulness of BSDS in clinical context is evident as illustrated by examples described above, BSDS can also foster [3] Bergmann, T.O. et al. 2016. Combining non-invasive progress in cognitive neuroscience. Currently, dominant approach transcranial brain stimulation with neuroimaging and for investigating the role of cortical oscillations in cognition is to electrophysiology: Current approaches and future randomly present events and then observe changes in event- perspectives. NeuroImage. 140, (2016), 4–19. related potentials or event-related oscillations. In the case of DOI:https://doi.org/10.1016/j.neuroimage.2016.02.012. investigation of effects of TMS on the brain, EEG is correlated with TEPs or MEPs. This approach is useful for generating [4] Brittain, J.S. et al. 2013. Tremor suppression by rhythmic hypotheses about relationship between brain and functional states, transcranial current stimulation. Current Biology. 23, 5 however, it is essentially a correlational approach. Stronger causal (2013), 436–440. inferences are possible if events are triggered when a specific DOI:https://doi.org/10.1016/j.cub.2013.01.068. brain state is detected. Besides BSDS where transcranial stimulation is adapted to the [5] Cohen, M.X. 2014. Analyzing neural time series data: brain state, a stimulus presentation or task can also be adapted to theory and practice. The MIT Press, Cambridge, the brain state, resulting in the so-called brain-state dependent Massachusetts. task (BSDT) [8]. For example, Ngo et al. [13] applied auditory closed-loop stimulation in phase with slow oscillation up-states [6] Gharabaghi, A. et al. 2014. Coupling brain-machine during sleep. This improved memory consolidation and enhanced interfaces with cortical stimulation for brain-state dependent declarative memory retention. stimulation: enhancing motor cortex excitability for neurorehabilitation. Frontiers in Human Neuroscience. 8, All three approaches can be used complementary: first, it can be March (2014), 1–7. shown that a specific oscillatory pattern is linked to behavior. For DOI:https://doi.org/10.3389/fnhum.2014.00122. example, Osipova et al. [14] have shown that stronger gamma and theta activity during visual stimuli presentation predicted [7] Hartmann, T. et al. 2011. Probing of brain states in real-time: subsequent retrieval. A hypothesis that gamma and theta activity Introducing the Console environment. Frontiers in is causally linked to memory encoding could further be tested 22 Psychology. 2, MAR (2011), 1–17. [14] Osipova, D. et al. 2006. Theta and gamma oscillations DOI:https://doi.org/10.3389/fpsyg.2011.00036. predict encoding and retrieval of declarative memory. Journal of Neuroscience. 26, 28 (2006), 7523–7531. [8] Horschig, J.M. et al. 2014. Hypothesis-driven methods to DOI:https://doi.org/10.1523/JNEUROSCI.1948-06.2006. augment human cognition by optimizing cortical oscillations. Frontiers in Systems Neuroscience. 8, June (2014), 1–21. [15] Schulz, H. et al. 2014. Now I am ready - Now I am not: The DOI:https://doi.org/10.3389/fnsys.2014.00119. influence of pre-TMS oscillations and corticomuscular coherence on motor-evoked potentials. Cerebral Cortex. 24, [9] Jensen, O. et al. 2011. Using brain-computer interfaces and 7 (2014), 1708–1719. brain-state dependent stimulation as tools in cognitive DOI:https://doi.org/10.1093/cercor/bht024. neuroscience. Frontiers in Psychology. 2, MAY (2011), 1– 11. DOI:https://doi.org/10.3389/fpsyg.2011.00100. [16] Veniero, D. et al. 2009. TMS-EEG co-registration: On TMS- induced artifact. Clinical Neurophysiology. 120, 7 (2009), [10] Karabanov, A. et al. 2016. Transcranial brain stimulation. 1392–1399. Current Opinion in Neurology. 29, 4 (2016), 397–404. DOI:https://doi.org/10.1016/j.clinph.2009.04.023. DOI:https://doi.org/10.1097/WCO.0000000000000342. [17] Walter, A. et al. 2012. Coupling BCI and cortical stimulation [11] Kraus, D. et al. 2016. Brain state-dependent transcranial for brain-state-dependent stimulation: methods for spectral magnetic closed-loop stimulation controlled by sensorimotor estimation in the presence of stimulation after-effects. desynchronization induces robust increase of corticospinal Frontiers in Neural Circuits. 6, November (2012), 1–17. excitability. Brain Stimulation. 9, 3 (2016), 415–424. DOI:https://doi.org/10.3389/fncir.2012.00087. DOI:https://doi.org/10.1016/j.brs.2016.02.007. [12] Little, S. et al. 2013. Adaptive deep brain stimulation in advanced Parkinson disease. Annals of neurology. 74, 3 (2013), 449–57. DOI:https://doi.org/10.1002/ana.23951. [13] Ngo, H.V. V et al. 2013. Auditory closed-loop stimulation of the sleep slow oscillation enhances memory. Neuron. 78, 3 (2013), 545–553. DOI:https://doi.org/10.1016/j.neuron.2013.03.006. 23 Diagnosticiranje parkinsonove bolezni iz glasu osebe Andrej Zupanc Ivan Bratko Univ. of Ljubljana, Faculty of Computer and Info. Sc. Univ. of Ljubljana, Faculty of Computer and Info. Sc. Večna pot 113 Večna pot 113 1000 Ljubljana 1000 Ljubljana +386 1 4798278 andrej.zupanc@gmail.com bratko@fri.uni-lj.si POVZETEK Podatkovne zbirke, ki smo jih uporabili, so objavili: Zgodnje odkrivanje parkinsonove bolezni je pomembno za (1) Athanasios Tsanas in Max Little [2], ki sta s pomočjo desetih lajšanje simptomov pri bolnikih. Za prakso bi bile koristne zdravstvenih ustanov in podjetjem Intel s pomočjo naprave za nezahtevne, splošno uporabne avtomatizirane metode odkrivanja nadzor na daljavo (angl. telemonitoring) spremljala 42 oseb v znakov bolezni. Ena možnost v tem pogledu je diagnosticiranje z obdobju šestih mesecev, ki so bile v zgodnji fazi PB. Vsaka oseba avtomatsko analizo glasu osebe. V tem prispevku so opisani je posnela približno 200 zvočnih zapisov. V podatkovni zbirki so poskusi z učenjem diagnosticiranja parkinsonove bolezni iz že izračunani atributi 5875 posnetkov. Vsak posnetek je zvočnih posnetkov glasu oseb v treh bazah podatkov, predvsem predstavljen s 26 atributi. Celotna podatkovna zbirka obsega 890 Sage Bionetworks. V poskusih ocenjena diagnostična točnost iz KB. 10 sekundnega zvočnega posnetka je 80,7%. (2) Max Little [3], ki je skupaj z Nacionalnim centrom za glas in Ključne besede govor (angl. National Centre for Voice and Speech) iz Denverja v Strojno učenje, medicinske aplikacije, parkinsonova bolezen, Koloradu naredil raziskavo z 31 osebami, med katerimi jih je 23 imelo PB, drugih osem je predstavljalo kontrolno skupino. Vsaka diagnostika oseba je posnela približno šest zvočnih zapisov. V podatkovni zbirki so že izračunani atributi 195 posnetkov. Vsak posnetek je 1. UVOD predstavljen s 23 atributi. Celotna podatkovna zbirka obsega Parkinsonova bolezen je nevrodegenerativna bolezen, katere približno 40 KB. glavni simptomi so tremor v mirovanju, počasno začenjanje (3) Sage Bionetworks [4], ki je sponzorirala raziskavo, pri kateri gibov, mišična rigidnost in tudi težave z govorom. Zgodnje so sodelujoči naložili na svoj mobilni telefon aplikacijo mPower. odkrivanje parkinsonove bolezni je pomembno za lajšanje S to aplikacijo so lahko opravili nekaj enostavnih testov (anketa, simptomov pri bolnikih. Za prakso bi bile koristne nezahtevne, snemanje glasu, uporaba aplikacije med hojo, tapkanje in igra za splošno uporabne avtomatizirane metode odkrivanja znakov testiranje spomina). Rezultati teh testov so se shranili na spletnem bolezni. Ena možnost v tem pogledu je avtomatska analiza glasu strežniku za poznejšo obdelavo in analizo. Izmed vseh teh osebe, kot kažejo npr. raziskave M. Littla, A. Tsanasa in informacj smo uporabili rezultate anket in posnetke glasu. Anketo sodelavcev [2,3,7,8,9] . V njihovih raziskavah je šlo predvsem za je izpolnilo 6805 različnih oseb, od katerih je svoj glas posnelo daljinsko spremljanje poteka bolezni. 5826 različnih oseb. Sodelujoči so svoj glas lahko posneli večkrat, tako je v podatkovni zbirki 65022 zvočnih posnetkov, ki so dolgi Nekatera izhodišča za pričujoče delo so opisana v [12]. V tem 10 sekund in še nimajo izračunanih atributov. Število oseb, ki so prispevku želimo z uporabo strojnega učenja razviti model za izpolnile anketo in posnele svoj glas, je 4962. Izmed teh je 970 prepoznavanje znakov parkinsonove bolezni v glasu osebe, ki bi ljudi označilo, da imajo diagnosticirano parkinsonovo bolezen, bil splošno uporaben kot mobilna aplikacija za zgodnje drugih 3992 oseb pa ne. Ta podatkovna zbirka obsega 81.3 GB. diagnosticiranje parkinsonove bolezni. Model je naučen iz Pri interpretaciji naših rezultatov je treba predpostaviti, da je tako primerov zvočnih posnetkov glasu »aaaa«, ki so jih bolniki in pridobljena informacija o zdravstvenem stanju sodelujočih oseb zdrave osebe posnele z elektronsko napravo. Te posnetke je dovolj zanesljiva. potrebno analizirati in izračunati atribute, na osnovi katerih se učimo klasificirati posnetke. Rezultat klasifikacije govori o tem, ali kaže zvočni posnetek znake parkinsonove bolezni ali ne. 3. POSKUSI Poskuse smo izvajali v programskem jeziku Python. Pri implementaciji učnih algoritmov, kot so naključni gozdovi, 2. UPORABLJENI PODATKI metoda podpornih vektorjev in drugih, smo si pomagali s Za učenje učnega modela smo uporabili tri podatkovne zbirke. knjižnico Scikit-learn [5]. Za implementacijo nevronskih mrež smo uporabili knjižnici Keras [6] in CNTK Dve podatkovni zbirki sta prosto dostopni na spletnem (https://www.microsoft.com/en-us/cognitive-toolkit/ ). Pri repozitoriju UCI Machine Learning Repository [1]. Tretjo preverjanju pravilnosti modela smo uporabili 10-kratno prečno podatkovno zbirko smo pridobili pri neprofitni organizaciji Sage preverjanje. Bionetworks. Baza je z ustrezno registracijo dostopna preko spletnega portala Synapse (https://synapse.org). Za lažje nadaljnje delo smo vizualizirali nekatere anketne odgovore v bazi Sage Bionetworks. Tako smo dobili boljši 24 pregled nad podatki, ki smo jih uporabili. Pri tem smo opazili, da dejanske zvočne posnetke. Prav tako smo s to knjižnico analizirali so ljudje s parkinsonovo boleznijo večkrat posneli svoj glas kot le prve veljavne posnetke posamezne osebe. ljudje brez. V povprečju so osebe s parkinsonovo boleznijo 40 krat posnele svoj glas, medtem ko so zdrave osebe v povprečju Z uporabo istih zvočnih posnetkov, le da z atributi izračunanimi s posnele svoj glas šest krat. Med analizo anketnih odgovorov smo knjižnico OpenSmile, je imel klasifikator, naučen z uporabo ugotovili tudi, da so nekatere osebe, ki nimajo diagnosticirane metode podpornih vektorjev klasifikacijsko točnost 77.9%. Na teh parkinsonove bolezni, prav tako navedle letnico, kdaj naj bi se jim podatkih se je med vsemi najbolje odrezal klasifikator z uporabo pojavili prvi simptomi ali pa katerega leta naj bi bili ansambelske metode GradientBoostingClassifier, ki je dosegel diagnosticirani s parkinsonovo boleznijo. Posnetkov teh oseb v klasifikacijsko točnost 80,7%. Ta rezultat je treba interpretirati z nadaljnji analizi nismo uporabili, saj jih nismo mogli obravnavati določeno mero previdnosti glede na to, da podatki v bazi Sage kot dovolj zanesljive. Prav tako v nadaljnjem delu nismo Bionetworks niso uravnoteženi glede na starost ob prisotnosti oz. uporabili posnetkov oseb, ki nimajo parkinsonove bolezni, vendar neprisotnosti bolezni. V podatkovni bazi je namreč povprečna imajo kakšno drugo bolezen. S tem se je zmanjšalo število starost oseb z boleznijo bistveno višja kot pri zdravih osebah. primernih oseb za analizo v podatkovni zbirki pridobljeni preko spletnega portala Synapse na 2909. 5. ZAKLJUČKI Da bi lahko združili podatkovne zbirke v eno, smo morali Naša doslej dosežena točnost diagnosticiranja parkinsonove uporabiti atribute izračunane z enakimi algoritmi. Zbirki, ki sta jih bolezni iz glasu (parkinsonova bolezen da/ne) je 80.7%. V objavila Athanasios Tsanas in Max Little, sta že imeli izračunane atribute, vendar je bilo med temi atributi samo deset takšnih, ki so nadaljevanju projekta nameravamo ta rezultat, dobljen s podatki iz bili izračunani z enakimi algoritmi in katere lahko tudi sami Sage Bionetworks, preveriti glede na možen vpliv porazdelitve izračunamo. Te atribute lahko izračunamo s pomočjo knjižnice oseb po starosti odvisno od prisotnosti bolezni. Izvedli bomo tudi poskuse z neposrednim učenjem z globokimi nevronskimi Voice Analysis Toolbox, ki jo je izdal Athanasios Tsanas mrežami iz zvočnega posnetka, brez računanja izpeljanih (https://people.maths.ox.ac.uk/tsanas/software.html), [7] [8] [9]. To knjižnico smo atributov iz posnetka. v prvem delu analize uporabili za izračun atributov zvočnih posnetkov, ki smo jih pridobili preko spletnega portala Synapse. Za izračun atributov smo uporabili samo prvi veljavni posnetek posamezne osebe. Za tako izbiro smo se 6. ZAHVALA odločili, da bi se izognili morebitnemu vplivu privajanja pri Raziskava je bila delno financirana s strani Ministrstva RS za kasnejših posnetkih. izobraževanje, znanost in šport (projekt EMZ) in Evropske unije iz Evropskega sklada za regionalni razvoj (ESRR) ter 4. REZULTATI raziskovalnega programa ARRS Umetna inteligenca in inteligentni sistemi. Avtorja se zahvaljujeta Dejanu Georgijevu, Klasifikacijski model, induciran iz podatkov iz prvih dveh baz z Zvezdanu Pirtošku in Aleksandru Sadikovu za koristno diskusijo. originalno izračunaninimi atributi (Tsanas in Little), je imel klasifikacijsko točnost 99.5% [11]. Zaradi suma, da tak rezultat ne more biti realen, smo iskali razlago, kako bi lahko prišlo do tako 7. REFERENCE visoke točnosti. Ena možnost je, da podatkovna množica ni bila dovolj uravnotežena glede na zastopanost obeh razredov v učni množici (posnetkov zdravih oseb je bilo manj kot 1%). Druga [1] M. Lichman, UCI Machine Learning Repository, 2013. možnost je pristranskost, saj bi lahko učni model našel povezave http://archive.ics.uci.edu/ml. med posnetki istih oseb, ki so bili v podatkovnih zbirkah, ki sta [2] A. Tsanas, M. A. Little, P. E. McSharry in L. O. Ramig, jih objavila Athanasios Tsanas in Max Little. Tako bi lahko Accurate Telemonitoring of Parkinson's Disease Progression by klasifikator določal razred glede na to, ali gre za isto osebo, ne pa, Noninvasive Speech Tests, IEEE Transactions on Biomedical ali gre za znake bolezni. Verjetna razlaga bi lahko bila tudi, da so Engineering, pp. 884-893, 2010. bili podatki v teh dveh zbirkah drugače obdelani. Zaradi tega smo opustili ti dve podatkovni zbirki in se osredotočili samo na [3] M. A. Little, P. E. McSharry, E. J. Hunter in L. O. Ramig, Suitability of Dysphonia Measurements for Telemonitoring of podatke iz portala Synapse. Z uporabo samo enega posnetka Parkinson's Disease, IEEE Transactions on Biomedical osebe in uravnoteženo množico je imel najboljši učni model klasifikacijsko točnost 67.8%. Ta klasifikator je bil naučen z Engineering, zv. 56, pp. 1015-1022, 2009. metodo podpornih vektorjev. Ta rezultat je zanimiv tudi zato, ker [4] These data were contributed by users of the Parkinson kaže, kako zavajajoči so lahko rezultati učenja iz izrazito mPower mobile application as part of the mPower study neuravnoteženih podatkovnih množic. V našem prvem poskusu developed by Sage Bionetworks and described in Synapse [11] je bila točnost preko 99%, čeprav smo pri tem uporabili [doi:10.7303/syn4993293]. metode za avtomatsko uravnotežanje podatkov. [5] F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Klasifikacijske modele, naučene z atributi izračunanimi z uporabo Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, knjižnice Voice Analysis Toolbox, smo želeli primerjati z drugimi P. M in E. Duchesnay, Scikit-learn: Machine Learning in Python, modeli, ki bi se učili na večjem številu izračunanih atributov. V ta Journal of Machine Learning Research, zv. 12, št. 2011, pp. 2825- namen smo uporabili tudi knjižnico OpenSmile [10], s katero smo -2830, 2011. izračunali za vsak posnetek 6375 atributov. S to knjižnico smo [6] F. Chollet, „Keras,“ Github, 2015 lahko analizirali samo podatke pridobljene preko spletnega . portala Synapse, saj smo le pri teh podatkih imeli na razpolago https://github.com/fchollet/keras. 25 [7] A. Tsanas, M. A. Little, P. E. McSharry in L. O. Ramig, New nonlinear markers and insights into speech signal degradation for effective tracking of Parkinson’s disease symptom severity, v International Symposium on Nonlinear Theory and its Applications (NOLTA), Krakov, 2010. [8] A. Tsanas, M. A. Little, P. E. McSharry in L. O. Ramig, Nonlinear speech analysis algorithms mapped to a standard metric achieve clinically useful quantification of average Parkinson’s disease symptom severity, Journal of the Royal Society Interface, zv. 8, pp. 842-855, 2011. [9] A. Tsanas, Accurate telemonitoring of Parkinson’s disease symptom severity using nonlinear speech signal processing and statistical machine learning. PhD Thesis, University of Oxford, 2012. [11] A. Zupanc, Napovedovanje parkinsonove bolezni z analizo govora s pametnim telefonom. Diplomsko delo, Fakulteta za računalništvo in informatiko, Univerza v Ljubljani, 2017. [12] A. Sadikov, I. Bratko, Some possibilities of applying AI to early detection and monitoring of Parkinson’s disease. Information Society Multiconference, Ljubljana 2016. 26 Arhitektura sistema za oddaljeno spremljanje pacientov Marko Janković Slavko Žitnik Marko Bajec Univerza v Ljubljani Univerza v Ljubljani Univerza v Ljubljani Fakulteta za računalništvo in Fakulteta za računalništvo in Fakulteta za računalništvo in informatiko informatiko informatiko Večna pot 113 Večna pot 113 Večna pot 113 1001 Ljubljana, Slovenia 1001 Ljubljana, Slovenia 1001 Ljubljana, Slovenia marko.jankovic@fri.uni-lj.si slavko.zitnik@fri.uni-lj.si marko.bajec@fri.uni-lj.si POVZETEK 2. ARHITEKTURA mDH V prispevku predstavimo arhitekturo sistema za oddaljeno Sistem mDH sestoji iz zaledne aplikacije, zgrajene na osnovi spremljanje pacientov MyDataHub, ki nastaja v sklopu projekta mikrostoritev, certifikatne agencije, kriptirane podatkovne baze in EkoSmart. Sistem sestoji iz zaledne aplikacije, zgrajene na osnovi avtentikacijskega ter avtorizacijskega strežnika. Namen sistema je mikrostoritev, certifikatne agencije, kriptirane podatkovne baze, zagotavljati varno povezljivost med napravami mDH, ki jih avtentikacijskega ter avtorizacijskega strežnika in mDH naprav, ki uporabljajo pacienti za merjenje svojih vitalnih znakov, in ostalimi jih uporabljajo pacienti za zajem vitalnih znakov. Namen sistema sistemi. Slednji zagotavljajo delovanje celostne zdravstvene je zagotavljati varno povezljivost med napravami mDH in ostalimi storitve od začetka telemedicinske obravnave, tehničnega in sistemi. zdravstvenega nadzora, do zaključka ali predaje naprave drugemu uporabniku. Za zunanje sisteme predstavlja mDH enotno točko za vso realnočasovno ali odloženo komunikacijo s pacienti ter za 1. UVOD dostop do njihovih podatkov. Zaradi ustreznega zagotavljanja V Evropi je prisoten trend staranja prebivalstva [1] ter naraščanja varnosti mDH vključuje lastno certifikatno agencijo in nabor štirih bolnikov s kroničnimi boleznimi [2]. Slednje predstavlja veliko programskih vmesnikov, ki so ločeni glede na funkcionalne zahteve družbeno in ekonomsko breme in kliče po inovativnih IKT rešitvah, zunanjih sistemov. Vsakega od programskih vmesnikov in ki bi razbremenile medicinsko osebje ter posledično zmanjšale primerov zunanjih sistemov na kratko predstavimo v sledečih čakalne vrste s pomočjo oddaljenega spremljanja pacientov. Kot razdelkih (2.1, 2.2, 2.3 in 2.4). odgovor na nastalo stanje se je na trgu že pojavilo veliko različnih rešitev (npr. OpenTeleHealth [3]). Med drugim tudi v sklopu 2.1 Pacientov portal in aplikacije projekta EkoSmart nastaja rešitev MyDataHub (mDH), ki bo Vmesnik /patient omogoča dostop do vseh podatkov vitalnih omogočala oddaljeno spremljanje pacienta s poudarkom na zajemu znakov, ki jih je pacient ustvaril s pomočjo naprav mDH. Preko vitalnih znakov z uporabo različnih merilnih naprav. V prispevku portala jih lahko dopolni z dodatnimi kontestualnimi podatki, deli je predstavljena arhitektura sistema mDH in vmesniki. Pri zasnovi z drugimi ali omogoča dostop do njih svojim družinskim članom. arhitekture je bil glavni poudarek na varovanju osebnih podatkov V primerjavi z napravo mDH, so lahko na portalu na voljo in zagotovitvi varnega kanala za prenos podatkov. naprednejše vizualizacije in morebitne avtomatske interpretacije meritev ali napredek. Pacientov portal in aplikacije Arhitektura mDH Spletni strežnik Podatkovna (SSL proxy) Zdravstvena hrbtenica ali baza zdravniški portal (kriptirana PostgreSQL) Patient HTTPS REST API Health HTTPS REST API Zaledni JavaEE strežnik z množico Management HTTPS mikrostoritev REST API Upravljalski sistem Avtentikacijski Hub HTTPS in avtorizacijski WS/REST API mDH naprave strežnik WSS (Keycloak) Certifikatna agencija Slika 1: Arhitektura integriranega sistema, ki omogoča komunikacijo z mobilnimi zdravstvenimi napravami pacientov in integracijo z zunanjimi sistemi z namenom zagotovitve celostnih zdravstvenih storitev. 27 2.2 Zdravstvena hrbtenica ali zdravniški [2] Reinhard Busse, Miriam Blumel, et al., Tackling chronic disease in Europe, Observatory studies Series Nº 20. portal Dostopno na: Povezava z zdravniškimi sistemi (vmesnik /health) je poleg http://www.euro.who.int/__data/assets/pdf_file/0008/96632/ naprav mDH eden ključnih za doseganje višje dodane vrednosti v E93736.pdf okviru elektronskih in mobilnih storitev v zdravstvu. Integracija z zdravniškimi sistemi omogoča zdravnikom, da nadzorujejo svoje [3] OpenTeleHelath. Dostopno na: paciente, so o njihovem poslabšanem stanju avtomatsko obveščeni, http://www.opentelehealth.com/ preverjajo njihove podatke, neposredno komunicirajo z njimi in jim oddaljeno določajo zahteve po merjenju vitalnih znakov. 2.3 Upravljalski sistem Preko vmesnika /manage je omogočeno celovito upravljanje z arhitekturo mDH. Upravljalski sistem definira storitve, ki zdravnikom in pacientom zakrijejo postopke vzpostavitve novih mDH naprav, reklamacij, menjav, beleženja aktivnosti, zagotavljanja sledljivosti in skalabilnost glede na predvidene potrebe. Preko vmesnika je možno mDH integrirati v obstoječi večji sistem in ga tako upravljati programsko. Prav tako je na voljo razvit upravljalski portal mDH, ki podpira osnovno delovanje in nadzor vseh storitev mDH. 2.4 Naprave mDH Naprave mDH skrbijo za zajem podatkov iz senzorjev ali merilnikov vitalnih znakov in povezavo s sistemom mDH. Preko vmesnika /hub sta izpostavljena dva tipa storitev - HTTPS REST in WSS. Websocket povezava med napravo in sistemom je vzpostavljena ves čas, ko je hub povezan v Internet, saj tako omogoča oddaljen nadzor in pošiljanje novih zahtevkov ali neposredno avdio-video komunikacijo z zdravnikom. Preko te povezave hub tudi pošilja nove rezultate meritev, ki jih v primeru nepovezljivosti hrani do 30 dni in so poslane takoj ob vzpostavitvi povezave. Storitve tipa REST pa omogočajo vzpostavitev nove mDH naprave, pošiljanje večjih binarnih datotek (na primer EKG meritve) ali posodobitve programske opreme. Vsaka mDH naprava ima dodeljen lasten odjemalski certifikat, preko katerega lahko dostopa do vseh storitev. V primeru komprimitiranja naprave pa se lahko ta certifikat doda na črno listo in tako onemogoči vzpostavljanje povezave z mDH sistemom. 3. ZAKLJUČEK V prihodnosti lahko pričakujemo porast sistemov in aplikacij, ki bodo omogočale zajem podatkov o našem zdravstvenem stanju ter podatke posredovale v zaledni sistem, kjer bodo na voljo zdravniku. Pri zasnovi arhitekture takega sistema je potrebno posebno pozornost nameniti zasebnosti in varnosti. V prispevku smo predstavili arhitekturo sistema mDH, ki nastaja v okviru projekta EkoSmart. Z mDH lahko pripomoremo k optimizaciji zdravstvene oskrbe in uspešnosti zdravljenja. 4. ZAHVALA Raziskave/delo je delno sofinancirano s strani Ministrstva za izobraževanje, znanost in šport in Evropske unije iz Evropskega sklada za regionalni razvoj (ESRR). 5. REFERENCE [1] Population structure and ageing, Eurostat. Dostopno na: http://ec.europa.eu/eurostat/statistics- explained/index.php/Population_structure_and_ageing 28 Razvoj zapestnice za pomo č starejšim Tomaž Kompara Elgoline d.o.o. Podskrajnik 34 1380 Cerknica Slovenija tomaz.kompara@elgoline.si POVZETEK 2. FUNKCIJE Razvoj zapestnice za pomoč starejšim narekuje uporabo naj- Na podlagi zbranih mnenj strokovnjakov in uporabnikov, novejših tehnologij in dognanj na področju umetne inteli- smo prišli do nabora funkcij, ki jih mora zapestnica omo- gence, da bi lahko zadovoljili želje in potrebe uporabnikov. gočati. Te so: (1) zvočna komunikacija, (2) ročno proženje Razvoj na področju elektronike in umetne inteligence omo- alarma, (3) detekcija padcev, (4) lokalizacija in (5) merjenje goča izdelavo zapestnic, ki bodo omogočale samodejen klic aktivnosti. na pomoč ob nevarnih situacijah. V tem delu so predsta- vljene ključne funkcije, ki jih zapestnica lahko nudi in kako 2.1 Zvočna komunikacija te vplivajo na uporabnikovo izkušnjo. V drugem delu so Zapestnica omogoča zvočno komunikacijo med uporabnikom predstavljene zahteve, ki so sestavljene na podlagi uporab- in klicnim centrom oz. drugo osebo, ki ji lahko v krizni si- niških izkušenj in mnenj strokovnjakov na tem področju. tuaciji nudi pomoč. Izbrana zvočna komunikacija je starej- šim uporabnikom enostavna za uporabo (velika večina upo- Ključne besede rabnikov zna uporabljati vsaj stacionarni telefon) ter omo- goča komunikacijo ne glede na namen vzpostavitve povezave. zdravje, starejši, zapestnica Prav tako je razlog za izbor v enostavnosti implementacije in majhnih zahtevah v hitrosti povezave, ki je potrebna med 1. UVOD napravo in klicano napravo. Za enako izbiro so se odločili tudi vsi sorodni produkti, ki smo jih pregledali. Trend staranja prebivalstva v razvitih državah narekuje is- kanje novih rešitev tako na družbenem, socialnem, poslov- nem ter finančnem, kot tudi na tehnološkem področju. Ra- zvoj novih tehnologij omogoča izboljšanje kakovost življenja starejšim, vendar mora biti način implementacije prilagojen posebej njim. Naprave namenjene starejšim morajo biti prilagojene njiho- vim željam, sposobnostim ter morajo reševati težave, ki jih imajo. Ena izmed glavnih težav starejše populacije je, da je kljub temu, da je velik del te populacije povsem sposo- ben samostojnega bivanja v svojem gospodinjstvu, prisiljen življenja v varovanih domovih, saj se bojijo, da jim v pri- meru ko bodo potrebovali pomoč, ta ne bo na voljo. Pri starostnikih, ki živijo sami v svojem gospodinjstvu je na- mreč verjetnost, da ob kritičnih situacijah ne bodo zmogli izvesti klica na pomoč večja kot pri ostali populaciji. Prav Slika 1: Govorna komunikacija. tako so lahko alarmantne spremembe v njihovem obnašanju, ki jih starostniki sam ne opazijo. Vse te težave so rešljive z 2.2 Ročno proženje alarma uporabo sodobnih tehnologij. Razvita zapestnica mora omogočati klic v sili ob zahtevi uporabnika. Ta funkcionalnost omogoča klic na pomoč ob Pri pregledu sorodnega dela, smo naleteli na nekaj rešitev vsaki nevarni situaciji, kot je npr. slabost, nesreča, pomoč [1, 2, 3, 4, 5], ki že nudijo funkcionalnost klica v sili, neka- drugi osebi ipd. Tak pristop omogoča hiter klic na pomoč in tere imajo tudi možnost lokalizacije in samodejne detekcije omogoča hitrejše posredovanje pristojnih organizacij ali po- padcev, medtem ko nismo našli rešitve, ki bi obenem me- sameznikov. Čas med nesrečo in obvestilom nujne pomoči rila tudi uporabnikovo aktivnost. Pregledane naprave imajo je namreč pri starejših v primerjavi z veščimi uporabniki so- različno dolgo življenjsko dobo baterije, ki se giblje od 36 ur dobnih tehnologij lahko precej daljši. V nekaterih primerih pa vse do 30 dni. Oblika omogoča nošnjo v žepu, pripeto na pride celo do tega, da starejši ne zmore priti do mobilnega pas ali kot verižico okrog vratu, nismo pa je našli v obliki ali stacionarnega telefona, s katerim bi poklical na pomoč. zapestnice. V nadaljevanju so predstavljene funkcionalnosti in zahteve za razvoj zapestnice. Z namestitvijo komunikacijske naprave na zapestje navide- 29 zno skrajšamo razdaljo med uporabnikom in potrebno po- pridobiti zadosten vzorec padcev. Ti bodo služili kot učni in močjo, saj se ta nahaja na uporabnikovem zapestju. Da bi testni podatki na podlagi katerih bo v prihodnosti mogoče zmanjšali število neželenih klicev, je potrebno gumb imple- zgraditi algoritme za boljše zaznavanje padcev. mentirati na pravem mestu, oziroma dodati več gumbov, ki šele ob hkratnemu pritisku sprožijo alarm. 2.4 Lokalizacija Pozicioniranje ali lokalizacija uporabnika je uporabna funk- cija, ki močno poenostavi iskanje osebe v primeru kritičnih situacijah. Za enkrat sta predvideni dve funkcionalnosti v povezavi z lokalizacijo: (1) lociranje uporabnika ob klicu na pomoč in (2) lociranje uporabnika, ob izginotju ali nenava- dni odsotnosti. Slika 2: Ročno proženje alarma. 2.3 Detekcija padcev Detekcija padcev nadgrajuje pomanjkljivosti ročnega prože- nja alarma, saj obstaja možnost, da v primeru padca upo- rabnik ni sposoben aktivirati ročnega klica na pomoč. V Slika 4: Globalna in lokalna lokalizacija. takih primerih lahko zapestnica samodejno zazna padec ter samodejno pokliče na pomoč. S to funkcionalnostjo dodatno Prva lokalizacija se proži ob ročnem ali avtomatskem klicu povečamo nabor morebitnih nevarnih dogodkov ob katerih na pomoč, pri čemer klicanemu pošlje sporočilo s koordina- zapestnica nudi pomoč uporabniku. tami uporabnika. Na ta način je mogoče locirati uporabnika, brez njegovega govornega opisa lokacije. To lahko bistveno skrajša čas za iskanje osebe v nevarnosti. Druga funkcionalnost omogoča lociranje uporabnika, v ko- likor se pojavi sum na izginotje, zaradi česar bi lahko bil uporabnik v nevarnosti. Najbolj ranljivi so predvsem de- mentni ljudje, saj lahko pride do tega, da zaidejo in se ne znajo vrniti v svoje bivalno okolje. Obe funkcionalnosti eno- stavno rešujeta precej težke situacije, ki so brez lokalizacije težko rešljive. 2.5 Merjenje aktivnosti Spremembe v aktivnosti uporabnika lahko nakazujejo na zdravstvene težave, zaradi česar je potrebna pomoč upo- rabniku. Z meritvijo uporabnikove aktivnosti je mogoče za- znavati odklone od normalnih okvirov ter pravočasno ukre- pati. Podatki o uporabnikovi aktivnosti lahko služijo tudi za Slika 3: Detekcija padcev z uporabo zapestnice. oceno uporabnikove fizične pripravljenosti ter za svetovanje glede pravilne količine gibanja, ki je primerna za posame- Trenutni sistemi zaznavajo padce s precej nizko točnostjo. znega uporabnika. V [6] poročajo o rezultatih realnih testov, v katerega je bilo vključenih 18 uporabnikov v obdobju 4 mesecev, pri čemer V zadnjem času se je na trgu pojavilo veliko število zape- je bil kar v 8% zaznan padec, čeprav le tega ni bilo, dejan- stnic, ki omogočajo štetje števila korakov, merjenje spanca ski padci pa so bili zaznani le v 25%. Ti podatki kažejo [8], merjenje aktivnosti [9] in podobno. Med raziskavo soro- o velikem deležu napačno proženih alarmih, ki so lahko za dnih naprav pa nismo naleteli na nobeno zapestnico, ki bi uporabnika zelo moteči, ter o majhnem številu pravilno raz- vsebovala zgoraj opisane funkcionalnosti in bi poleg tega nu- poznanih padcih, kar v realnosti nudi le malo pomoči upo- dila merjenje aktivnosti. Prav tako nismo našli zapestnice, rabnikom. ki bi omogočala ugotavljanje sprememb v gibanju. Kljub temu je s hitrim razvojem umetne inteligence mogoče Ob zaznavi odklona uporabnikove aktivnosti, je predvideno pričakovati, da bomo lahko ta problem boljše rešili že v bli- obveščanje skrbnika, ki lahko preko klicne povezave preveri žnji prihodnosti [7]. Glavna težava je pridobitev zadostne ali osebno preveri, če je z varovancem vse v redu. Ta funkcio- količine podatkov o padcih, saj je njihova pojavitev redka, nalnost omogoča zgodnje odkrivanje težav ter njihovo odpra- vendar je z velikim številom naprav in uporabnikov mogoče vljanje, kar lahko pripomore k hitrejšemu okrevanju. Poleg 30 tega je mogoče zaznati tudi prevelike napore uporabnika, ob napolni lahko uporabnika o tem obvesti ter ga s tem spomni, čemer se lahko uporabnika opozori ter vpraša če potrebuje da si zapestnico ponovno nadene na roko. pomoč pri opravilu, ki ga počne. Ta funkcionalnost omogoča personalizirano svetovanje uporabniku s čemer je mogoče na 3.3 Interakcija z uporabnikom dolgi rok izboljšati zdravje in dobro počutje uporabnika. Interakcija uporabnika z napravo mora biti enostavna in hi- tra ter intuitivna za starejše uporabnike. Uporabljeni so 3. ZAHTEVE lahko zvočni, svetlobni ali vibrirajoči efekti, ki uporabniku V tem poglavju so predstavljene zahteve, ki temeljijo na v enoznačno nakažejo stanje naprave oziroma akcijo, ki se vrši. prejšnjem poglavju opisanih funkcionalnostih. Glavne zah- Prav tako je količina nepotrebnih elementov zmanjšana na teve so razdeljene na tri podpoglavja in sicer: (1) območje najmanjšo možno vrednost, s čimer postane uporabniku bolj delovanja, (2) enostavnost uporabe in (3) interakcija z upo- intuitivna in ga različne možnosti ne zmedejo. Zelo po- rabnikom. membno je, da zapestnica nudi hiter odziv ter pridobi upo- rabnikovo zaupanje, da bo v kritičnih situacijah pravilno 3.1 Področje delovanja delovala. V splošnem ločimo naprave, ki so namenjene lokalni (npr. varovani domovi) ali globalni uporabi. V primeru lokalne 4. ZAKLJU ČEK uporabe, sistem navadno potrebuje dodatno infrastrukturo, Pri razvoju zapestnice za starejše je potrebno paziti na mnogo ki omogoča locirati uporabnika znotraj objekta. V primeru dejavnikov, ki lahko močno vplivajo na uporabnost in učin- globalne uporabe je navadno uporabljena satelitska naviga- kovitost le te. Nekatere podatke za načrtovanje je mogoče cija, ki omogoča lociranje kjerkoli na svetu, če naprava pre- pridobiti od uporabnikov samih, v pomoč pa pridejo tudi nji- jema signal z vsaj treh navigacijskih satelitov. V splošnem hovi skrbniki, ki znajo bolj objektivno oceniti in ovrednotiti je globalna uporaba bolj primerna, saj uporabnika ne ome- težave, ki jih imajo starejši. juje, da se giblje le v varovanem območju. Slabost satelitske navigacije je v tem, da ne deluje v zaprtih prostorih. V na- 5. ZAHVALE sprotju z globalnim lociranjem, je slabost lokalnega lociranja omejitev območja varnega gibanja ter potreba po dodatni Raziskava je bila izvedena v okviru projekta ”Ekosistem Pa- infrastrukturi, kar vpliva tudi na ceno sistema. metnega mesta (EkoSmart)” in je sofinancirana s strani Re- publike Slovenije in Ministrstva za izobraževanje, znanost in Mogoča je tudi kombinacija obeh tipov senzorjev. V primeru šport ter Evropske unije iz Evropskega sklada za regionalni ko se oseba nahaja znotraj svojega bivalnega okolja, se to razvoj (ESRR). zazna z uporabo lokalne navigacije, medtem ko se osebo, ki je zunaj, locira z uporabo satelitske navigacije. Še vedno pa so 6. VIRI problematični vsi ostali zaprti prostori, ki nimajo potrebne [1] Safeguardian, carecallersTM. Dostopno na infrastrukture za lokalizacijo. V takšnih primerih celoten https://safeguardian.com/, september 2017. sistem sloni na verbalni komunikaciji med uporabnikom in [2] Life call, fallalertTM system. Dostopno na klicnim centrom. http://lifecall.com/products/, september 2017. [3] Medical guardian, premium guardian. Dostopno na 3.2 Enostavnost uporabe https://www.medicalguardian.com/product/, Enostavnost uporabe je prilagojena starejšim uporabnikom september 2017. z namenom, da je možnih nepravilnosti pri uporabi čim [4] Alert 1, mobile medical alert system. Dostopno na manj. Prva nepravilnost, ki jo želimo preprečiti je, da bi https://www.alert-1.com/, september 2017. uporabnik pozabil nositi napravo. V tem primeru uporab- [5] Philips, gosafe. Dostopno na nik morda ne bo mogel poklicati na pomoč, saj bo naprava https://www.lifeline.philips.com/, september 2017. preveč oddaljena od mesta, kjer se nahaja. Da bi zmanjšali [6] Thompson HJ Chaudhuri S, Oudejans D and Demiris možnost pozabljanja nošenja naprave, smo napravo predvi- G. Real world accuracy and use of a wearable fall deli kot zapestnico, ki se nosi na zapestju. Poleg tega mora detection device by older adults. Journal of the ta omogočati več dnevno, več tedensko ali celo nekaj me- American Geriatrics Society, 11(63):2415–2416, 2015. sečno uporabo, brez da bi jo uporabnik moral polniti. Na ta [7] Yueng Santiago Delahoz and Miguel Angel Labrador. način ima uporabnik manj možnosti, da pozabi napravo, saj Survey on fall detection and fall prevention using jo le redko odstrani z zapestja. Da bi še dodatno zmanjšali wearable and external sensors. Sensors, potrebo po odstranitvi zapestnice, je potrebno to narediti 14(10):19806–19842, 2014. vodoodporno, s čimer jo lahko uporabnik nosi tudi pri ak- [8] Yueng Santiago Delahoz and Miguel Angel Labrador. tivnostih, ki vključujejo vodo. Survey on fall detection and fall prevention using wearable and external sensors. Sensors, Druga nepravilnost je nošenje zapestnice s praznim akumu- 14(10):19806–19842, 2014. latorjem. Tudi če smo rešili težavo z nošnjo zapestnice, je ta [9] Andrea Mannini, Mary Rosenberger, William v primeru ko ima prazen akumulator, neuporabna. V ta na- L Haskell, Angelo Sabatini, and Stephen S Intille. men mora zapestnica samodejno opozarjati uporabnika na Activity recognition in youth using single accelerometer stanje akumulatorja in potrebo po polnjenju. Zelo dobrodo- placed at wrist or ankle. 49:801–812, 04 2017. šlo je brezkontaktno polnjenje, ki zmanjša možnost nepra- vilnega polnjenja ter prenehanja delovanja. Poleg tega je uporabniku veliko bolj enostavno položiti napravo na brez- žični polnilec, kot priključiti polnilni kabel. Ko se zapestnica 31 Aplikacija tehnologije BLE za avtomatično zaznavanje prisotnosti oseb in predmetov Andrej Planina, dr. Luka Vidmar Špica International d.o.o. Ljubljana POVZETEK zmožnosti integracije v razne rešitve ter postavili prototip sistema Vedno več je za avtomatično zaznavanje premičnih objektov v poslovnem prenosnih elektronskih naprav, ki imajo v sebi vmesnik Bluetooth Low Energy (BLE). Te naprave je možno okolju. uporabiti za avtomatsko zaznavanje prisotnosti ljudi ali predmetov v prostorih, kar je uporabno za več aplikacij od poslovnih do 2. Sestavni deli rešitve osebnih. V prispevku prikažemo izkušnje pri uporabi tehnologije 2.1 Prenosni identifikator BLE BLE v ta namen in nekaj možnosti uporabe te tehnologije. Kot prenosni identifikator, ki označuje osebo ali predmet, smo uporabili različne prenosne naprave z vgrajeno tehnologijo BLE Ključne besede in sicer: a) osebna zapestnica za sledenje fizičnih aktivnosti BLE, zaznavanje prisotnosti, zapestnice, pametne ure, spremljanje XiaoMi MiBand, b) značka Chipolo BLE, in c) novejši pametni gibanja, senzorska omrežja telefon Android. V poslovnem svetu je najbolj primerna naprava mobilni telefon z BLE. V privatni sferi pa je najbolj primerna 1. UVOD zapestnica BLE ali pametna ura z vmesnikom BLE. Za Večina raznih informacijskih sistemov trenutno zaznava označevanje predmetov pa je najbolj primerna značka Chipolo. prisotnost oseb ali predmetov preko namenskih sistemov za Vse izbrane naprave delujejo na akumulatorsko baterijo, ki jo je avtomatsko identifikacijo ali preko ročnega vnosa prisotnosti. potrebno občasno napolniti, eno polnjenje pa zadošča za uporabo Večinoma se uporabljajo čipi ali kartice RFID. od enega dneva (pametni telefon) do nekaj let (Chipolo), kar močno vpliva na uporabnost naprav. S pojavom vedno več prenosnih naprav, ki imajo v sebi vmesnike Bluetooth Low Energy (BLE), je za identifikacijo možno uporabiti BLE naprave, ki jih nosijo ljudje ali predmeti na sebi [3]. Domet naprav BLE, padajoča cena in široka dosegljivost teh 2.2 Fiksno senzorsko omrežje BLE naprav še povečujejo možnost takih rešitev. Naloga fiksnih senzorjev BLE je zaznavati prisotnost naprav BLE Za celovito uporabno rešitev za tako vrsto identifikacije je in v nadzorni sistem javljati unikatne naslove MAC in časovne potrebno postaviti omrežje fiksnih senzorjev BLE, ki bodo žige, ko je bila posamezna naprava BLE identificirana. Za namen zaznavali premične identifikatorje BLE in njihovo prisotnost prototipa smo senzor BLE vgradili v Špica Time&Space fiksni javljali v nadzorni sistem. Tako rešitev lahko umestimo v več registrator Zone Touch. različnih področij, tako v poslovno kot v osebno uporabo: Pametni dom, Pametne zgradbe ali Pametne tovarne. Slika 2. Fiksni terminal Zone Touch z vgrajenim senzorjem BLE Izbrali smo tak modul BLE, da se je lahko tehnično vklopil v napravo (USB vmesnik) s hkrati dovolj nizko ceno (velikosti 10 EUR). Prototipna naprava se v zaledni sistem povezuje preko vmesnika Ethernet. Naprava zaznava vse avtorizirane, prisotne in dostopne naprave BLE in v zaledni sistem javlja svojo identifikacijsko oznako, naslov MAC zaznane prenosne značke Slika 1. Shema celotnega sistema BLE in časovni žig. Naprava opravlja osnovno filtriranje, da V okviru projekta EkoSmart smo raziskali tehnologijo BLE za zalednega sistema ne preobremenjuje s preveč podatki. namen identifikacije oseb in predmetov, njeno uporabo in 32 2.3 Baza podatkov natančno razdelili med različen kategorije, na primer priprava na Podatki o prisotnosti objektov iz vseh senzorjev se morajo hraniti delo, hoja med prostori, delo v proizvodnji, odmor, delo v skladišču, administracija, delo na stroju A, delo na stroju B, delo v centralni bazi podatkov. Za prototipne potrebe se podatki hranijo v posebni prototipni bazi podatkov, kjer je zabeležena na nalogi C itd. [6] identifikacija točke, kjer je bil objekt zaznan, naslov MAC zaznane značke BLE in časovni žig dogodka. 4.2 Kontrola pristopa V primeru, ko so določena vrata v poslovnih prostorih zaprta in je V eni od prototipnih postavitev na IBM Innovation centru v prehod dovolj le avtoriziranim osebam, se ta vrata sedaj navadno Ljubljani se ti podatki shranjujejo v IBM Bluemix platformi v odpirajo s pomočjo osebne kartice RFID, podobno kot pri oblaku. registraciji delovnega časa. Identifikacijo pooblaščene osebe pred vrati bi s pomočjo sistema BLE zaznali avtomatsko in vrata 3. Odkriti in odprti problemi avtomatsko odklenili. V okviru raziskav in izdelave prototipa smo odkrili nekaj težav, ki jih moramo v bodoče še nasloviti in poskušati najti primerne 4.3 Evidentiranje prisotnosti pošiljk rešitve. Ena od konkretnih možnosti uporabe je tudi v logistiki, kjer bi značke BLE namestili bodisi na vozičke, palete ali druge 3.1 Kaj je dogodek? transportne enote, bodisi na same pošiljke ali artikle. Uporaba V praksi se dogaja, da senzor BLE zazna časovno zelo kratko tehnologije BLE na pošiljkah ali artiklih je še predraga in zato prisotnost značke BLE. Odločiti se je potrebno, ali lahko tako neprimerna. Uporaba na vozičkih pa je že poslovno sprejemljiva, kratko prisotnost štejemo kot dogodek ali ne. Lahko da gre za saj se z BLE označeni vozički uporabljajo večkrat oz. se vračajo k nehoteno zaznavanje človeka, ki se je samo sprehodil mimo lastniku in gre za t.i. zaprt krog. [2] senzorja in ga niti ne bi bilo potrebno zaznati. Problem je možno reševati na več načinov: s časovnim filtriranjem, z nastavljanjem 4.4 Spremljanje gibanja doma Starejši v domačem okolju dometa senzorja BLE, s postavitvijo več sprejemnikov BLE, s bi uporabljali vsak svojo osebno pokrivanjem celotnega prostora z BLE ali s poslovnimi pravili[4]. zapestnico BLE, vsak prostor v njihovem domu bi bil opremljen z senzorjem BLE, skrbniki bi imeli vpogled v grobo gibanje svojih oskrbovalcev. Nadzorni sistem bi lahko avtomatično ugotavljal 3.2 Kakšna je smer gibanja predmeta? trende pri gibanju in sam alarmiral skrbnike v primeru sprememb Nekatere aplikacije zahtevajo določitev smeri gibanja predmeta. v načinu bivanja in gibanja. [1] Pri kontroli pristopa ali registraciji delovnega časa je pomembno, ali gre zaposleni v prostor ali iz prostora. Podobno je pri sledenju logističnih enot. Tehnologija BLE sama po sebi ne zmore zaznati 5. Zahvala smeri gibanja in bi bilo potrebno smer določiti s pomočjo dveh Projekt EkoSmart, v okviru katerega so nastali opisani rezultati, je točk BLE v prostoru in s pametno interpretacijo časovnih žigov delno sofinanciran s strani Ministrstva za izobraževanje, znanost pri zaznavanju značk BLE. Zaradi časovnih zakasnite in težke in šport. Naložbo sofinancirata Republika Slovenija in Evropska kontrole nad razdaljo zaznavanja značk BLE je metoda časovnih unija iz Evropskega sklada za regionalni razvoj. zakasnitev precej nezanesljiva za ugotavljanje smeri gibanja. 3.3 Sprejemljivost avtomatičnega zaznavanja 6. Literatura oseb [1] Interna dokumentacija projekta EkoSmart, projekt RRP4 Povečana skrb za zasebnost ljudi povzroča zadržanost oseb glede Elektronske in mobilne storitve. avtomatičnega sledenja njihove prisotnosti. Ljudje ne želijo biti [2] Interna dokumentacija projekta EkoSmart, projekt RRP2 avtomatično zaznavani, razen če od tega nimajo izrazitih koristi. Pametna mobilnost. Glavna korist, zaradi katere so se ljudje pripravljeni odreči delu zasebnosti, je varnost. Tehnologijo avtomatičnega zaznavanja [3] Alaa Alhamoud, Arun Asokan Nair, Christian Gottron, oseb smo tako do sedaj uporabili le pri sledenju vhoda oseb v Doreen Bohnstedt, Ralf Steinmetz, Presence detection, rudnik, ki je izrazito nevarno okolje. V tem primeru so se bili identification and tracking in smart homes utilizing upravitelj rudnika in rudarji ter ostali zaposleni v jami pripravljeni bluetooth enabled smartphones, 2014 IEEE 39th Conference odreči zasebnosti, pa še tu je prihajalo po odpora posameznih on Local Computer Networks Workshops. ljudi [5]. [4] Gunther Fischer, Burkhart Dietrich, Frank Winkler, Bluetooth Indoor Localization System, Proceedings of the 1st Workshop on Positioning, Navigation and Communication 4. Možne aplikacije (WPNC 2014) [5] Interna dokumentacija projekta “Avtomatično evidentiranje 4.1 Avtomatična registracija delovnega časa zaposlenih pri vhodu v rudnik, Premogovnik Velenje”, Špica Sistemi za registracijo delovnega časa večinoma delujejo preko International, 2015 izrecne prijave zaposlenih s pomočjo osebnih kartic RFID na razdaljo nekaj cm od čitalca [6] Karel Stanovnik, magistrsko delo “Sistemi sledenja v zaprtih RFID. V primeru pogostih prihodov in odhodov v službo ali pri prehajanju med različnimi prostorih”, Univerza v Ljubljani, Fakulteta za elektrotehniko, 2015 kategorijami delovnega časa (različna stroškovna mesta) bi bilo zaposlenim precej bolj enostavno pripraviti avtomatično registracijo prehodov. Tako bi lahko delovni čas lahko precej bolj 33 Technology for training and assessment of precise movements in persons with Parkinson’s disease Imre Cikajlo Zlatko Matjačić Helena Burger Karmen Peterlin Potisk University rehabilitation University rehabilitation University rehabilitation University rehabilitation institute institute institute institute Linhartova 51 Linhartova 51 Linhartova 51 Linhartova 51 SI-1000 Ljubljana SI-1000 Ljubljana SI-1000 Ljubljana SI-1000 Ljubljana +386 1 475 8 150 +386 1 475 8 150 +386 1 475 8 150 +386 1 475 8 150 imre.cikajlo@ir-rs.si zlatko.matjacic@ir-rs.si helena.burger@ir-rs.si karmen.potisk@ir-rs.si ABSTRACT cube, the pick & place time, the average time required for cube We have developed a system for training and assessment of placement, number of unsuccessful trials and number of cubes precise movements with upper extremities. The system consists of successfully placed in the chest. Besides, the high-frequency and mainly commercially available hardware components and is low amplitude hand movements in the random directions were intended for home use or telerehabilitation in persons with observed and the tremor of the hand was estimated. Additionally Parkinson disease. Small range motion with upper extremity and validated clinical instrument Box & Block Test (BBT) was carried movements with fingers were recorded with miniature 3D camera. out. The person was expected to collect virtual cubes within the virtual environment. The system was preliminarily tested in neurologically intact person and in patient with Parkinson’s disease. The kinematics of the hand and fingers were analyzed off- line and characteristic features were extracted to build the assessment procedure. The major objective of the project is targeted towards the development of novel instrument for the assessment that is closer to the existing validated clinical instrument (e.g. Box & Block Test). Keywords virtual reality, kinematics, upper extremity, Parkinson’s disease, telerehabilitation 1. INTRODUCTION Parkinson’s disease (PD) is a slowly progressive degenerative disease of the extrapyramidal system [1]. The disease may affect Figure 1. 10Cubes, a 2D games, also convenient for people at the age between 35 and 60 years. Typically affects daily home based physiotherapy or telerehabilitation. activities, participation and quality of life. The patients with PD are mainly subject to drug treatment and rarely receive comprehensive rehabilitation including physiotherapy and The system was preliminary tested in one neurologically intact occupational therapy. There are also contradictory reports about person and a person with PD in order to estimate the feasibility of the successfulness of physiotherapy [2]. the approach (Figure 1). However, in the forthcoming project we propose that intensified 3. RESULTS training of upper extremity skills and finger movements may increase the person’s ability to focus on motor function [3] A real-life test demonstrated that the neurologically intact subject keeping the same or even decreased dose of the medicine. grabbed each cube and placed in the chest without failures, while Telerehabilitation as a service may keep the person active and the person with PD occasionally dropped the cube, grabbed it taking less medicine. again or even misplaced the cube (Figure 2,). The healthy participant scored 62 and patient with PD only 42 with the BBT. 2. METHODOLOGY 4. DISCUSSION The hand, palm and finger movements were tracked by small 3D camera (Leapmotion Inc, USA). The entire kinematics of the The preliminary tests were successful, demonstrating that major fingers and palm were used for calculation of the virtual hand in differences in kinematics and strategy exists between the healthy the designed virtual environment (Unity3D, Unity Technologies, person and the person with PD. The person with PD was able to CA, USA). We have designed a simple task, 10Cubes, where the put all the 10 cubes in the chest only in his fifth attempt. During person was asked to pick and place 10 virtual cubes in the virtual the task we recorded several unsuccessful trials like misplacement wooden chest. These cubes were randomly spread over the area, of the cube, cubes falling out of the hand, causing tremendous leaving the participants to choose their own strategy of putting all hand tremor and other measurable components supported by the 10 cubes in the chest within 2 min. literature [5]. We have extracted the kinematics of the hand while moving in the However, in the near future we plan a larger inpatient hospital trial free space and when grabbing the cube; the time of holding the with >20 patients with PD and 1-2 trial on patient’s home. In these 34 trials we plan to perform also clinical test [4] before and after the 6. REFERENCES training tasks and record the medication plan. [1] Melnik M.E. 1995. Basal ganglija disorders. In : Umphred DA ed. Neurological rehabilitation. 3rd ed. St. Louis: Mosby, 606-636 [2] Clarke CE, Patel S, Ives N, Rick CE, Dowling F, Woolley R, et al. Physiotherapy and Occupational Therapy vs No Therapy in Mild to Moderate Parkinson Disease. JAMA Neurol [Internet]. 2016 Mar;73(3):291. Available from: http://www.ncbi.nlm.nih.gov/pubmed/26785394 [3] Barry G., Galna B. and Rochester L. 2014. The Role of Exergaming in Parkinson's Disease Rehabilitation: A+A Systematic Review of the Evidence . J Neuroeng Rehabil. 7, 11- 33. [4] Fisher A.G. 2003. Assessment of Motor and Process Skills: Figure 2. Pick & place of 10cubes: each cube has its Volume I – Development, Standardization, and own color; left: healhty person, right: person with PD Administration Manual Fifth Edition. Three Star Press, Inc.- Fort Collins, Colorado USA [5] Wang C.Y., Hwang W.J., Fang J.J., Sheu C.F., Leong I.F. and 5. ACKNOWLEDGMENTS Ma H.I. 2011. Comparison of virtual reality versus physical The authors would like to acknowledge the financial support to reality on movement characteristics of persons with the Republic of Slovenia and the European Union under the Parkinson's disease: effects of moving targets. Arch Phys European Regional Development Fund (project EKOSMART). Med Rehabil. 8, 92, 1238-1245 35 Smart Dentistry and Smart Tooth Brushing Peter Kokol Matjaž Colnarić Jernej Završnik University of Maribor, Faculty of University of Maribor, Faculty of Dr. Adolf Drolc Health Care Centre, Electrical Engineering and Computer Electrical Engineering and Computer Ulica talcev 9, Science Science 2000 Maribor, Slovenia Koroška cesta 46 Koroška cesta 46 jernej.zavrsnik@zdm.mb 2000 Maribor, Slovenia 2000 Maribor, Slovenia peter.kokol@um.si matjaž.colnaricl@um.si Milan Zorman Domen Verber Marko Turčin University of Maribor, Faculty of University of Maribor, Faculty of Dr. Adolf Drolc Health Care Centre, Electrical Engineering and Computer Electrical Engineering and Computer Ulica talcev 9, Science, Science 2000 Maribor, Slovenia Koroška cesta 46 Koroška cesta 46 marko.turcin@zdm.mb 2000 Maribor, Slovenia 2000 Maribor, Slovenia milan.zorman@um.si domen.verber@um.si Bojan Žlahtič Stanislav Moraus Simon Jurič University of Maribor, Faculty of University of Maribor, Faculty of Pesnica pri Mariboru 32e 2211 Electrical Engineering and Computer Electrical Engineering and Computer Pesnica pri Mariboru, Slovenia Science Science simon.juric@inova.si, Koroška cesta 46 Koroška cesta 46 2000 Maribor, Slovenia 2000 Maribor, Slovenia bojan.zlahticl@um.si stanislav.moraus@um.si Grega Žlahtič Bojan Slemnik University of Maribor, Faculty of Pesnica pri Mariboru 32e 2211 Electrical Engineering and Computer Pesnica pri Mariboru, Slovenia Science bojan.slemnik@inova.si Koroška cesta 46 2000 Maribor, Slovenia grega.zlahtic1@um.si ABSTRACT semantic health records should also be integrated into future/smart In this paper, we describe some smart dentistry tools and dental care [2]. applications developed in the scope of the EkoSmart project. The bibliometric analysis performed with VOSviewer software (Leiden University, Netherlands) [3] of 2567 papers related to General Terms smart dentistry retrieved from the Scopus bibliographical database Smart medicine (Elsevier, Netherlands) using the search string smart or personalized or precision or G4H or "artificial intelligence" or "3D print*" or nanotechnology or robotic* or IoT or "semantic Keywords health record” showed a fast growing trend in research Smart dentistry, smart tooth brush, serious games, kinect endeavors. The research was mainly focused on the following themes: 1. INTRODUCTION  Smart materials and brackets  Latest advances in information, communication and health Computer aided design and 3D printing technologies triggered a paradigm shift in modern medicine – the  Digital impression transition to so called smart medicine. The predicted trends in dentistry like increasing digitization [1] show that key technology  Smart dental implants, abutments and crowns trends emerging in smart medicine like personalized and precision  Periodontal diseases and smart therapy medicine, gamification based treatment, artificial intelligence, 3D printing, nanotechnology, robotics, Internet of things (IoT) and 36 The analysis showed that smart dentistry in general is following smart medicine trends however there are substantial gaps especially concerning smart tooth brushing, gamification and use of artificial intelligence. 2. SMART TOOTH BRUSHING The area of smart toothbrushes has recently had quite some technological leaps forward. In the frontier in the field of smart toothbrushes, Oral-B SmartSeries, has in 2016 allowed us to monitor the duration, location and pressure of brushing, using a smartphone, in 2017 the toothbrush Kolibree Ara has moved the measurement of mentioned physicalities to the toothbrush, in order to store and measure everything even in the absence of a smartphone. They capture the data using a 3D sensor, then they process the data using artificial intelligence. Onvi Prophix went Fig 2 Smart toothbrush prototype with the mobile application even a step further, for around 400 USD (almost two times the price of other smart toothbrushes) they are offering a smart The next prototype will be 3D-printed, the preparatory model is toothbrush with four different adaptors and an HD camera that is depicted in the Fig 3. recording the brushing process using a phone app. Grush has an interesting solution that can be compared with ours. Using Intel’s module, it captures data of this smart toothbrush, this data is than used to change brushing your teeth into a game. All the toothbrushes mentioned before only measure the position and brushing time. Our toothbrush, however, also measures pressure and acceleration, and thus allows us to grade additional parameters for proper cleaning. The other major difference between the before mentioned smart toothbrushes and our solution is that all commercially available toothbrushes have an electrically driven head, while our toothbrush is conventional and therefore its use is more flexible and the price considerably lower. Fig 3 3D model prepared for 3D printing The functionality and usability of the smart toothbrush can be further extended with a Kinect application. The main application functionality is the visual assistance for correct positioning of body, head and arm in the beginning of and during tooth brushing and together with data from the toothbrush to detect “anomalous movements” (Fig 4). The application supports finding correct Fig 1 Smart toothbrush prototype starting position, provides appropriate relativisation of the perception of movement of the brush and head, and enables time synchronization of the smart toothbrush sensors with Kinect. The The prototype toothbrush is shown in Fig.1 (without housing). In exchange of data is performed by Bluetooth connection, with the Fig.2, the complete prototype with the 3D mobile application, time-coordinated recording of the body position and collection of demonstrating the movements of the toothbrush, are given. toothbrush data. 37 Fig 5 The initial scenario of a serious dental game Fig 4 The Kinect application (palm distance from the mouth - 3. REFERENCES violet dot (full circle) means that the palm is positioned correctly in front of the head; the position of the palms in front of the mouth - a light green circle surrounding the mouth indicates the 1. Dowson T. 7 Dental Industry Trends in 2017 & What area in which the palm needs to be positioned (the violet dot must They Mean For Practice Growth [Internet]. [cited 2017 May 13]. be within this circle); position of the wrist in front of the mouth - Available from: https://titanwebagency.com/blog/dental-industry- yellow section of the circle marks the right position of the wrist; trends/ the position of the arm - the green section of the circle indicates 2. Železnik D, Kokol P, Vošner HB. Adapting nurse the right hand position; the position of the head and body; - a competence to future patient needs using Checkland’s Soft green circle section indicates the right position of the head and Systems Methodology. Nurse Educ Today. 2017 Jan 1;48:106– body) 110. PMID: 27744237 3. Završnik J, Kokol P, del Torso S, Blažun Vošner H. Citation context and impact of ‘sleeping beauties’ in paediatric 2.1 Gamification research. J Int Med Res. 2016;44(6):1212–1221. 4. LudoScience - Classifying Serious Games: The G/P/S Model Serious games are games that aren’t designed only for the sake of (Broacasting our studies) [Internet]. [cited 2017 May 9]. entertainment [4]. Serious games for health care can offer new Available from: http://www.ludoscience.com/EN/diffusion/537- and potentially highly effective paradigm for behaviour change, Classifying-Serious-Games-The-GPS-Model.html influence outcome and increase knowledge [5]. In paediatrics, serious games can be used for different purposes. Through 5. Baranowski T, Blumberg F, Buday R, DeSmet A, Fiellin LE, interactive experiences, they can offer children: goals, challenges, Green CS, et al. Games for Health for Children—Current Status problem solving, experience and intense moments that provide a and Needed Research. Games Health J. 2015;5:1–12. high level of motivation [6]. Based on the positive experience in 6. Macías E, García O, Moreno P, Presno MM, Forrest T. using games for health we developed a serious dentistry game Glooveth: healthy living, fun and serious gaming. Stud. Health which based on the outputs from our smart toothbrush configures Technol. Inform. 2012;172:180–4. the game playing platform and motivates children to brush their teeth regularly and in the correct way. 38 Prototipi aplikacij za prenos mobilnih EKG meritev od uporabnika senzorja do zdravnika Denis Pavliha Nataša Planinc Matjaž Depol i Aleš Smokvina SRC SRC Infonet Inštitut Jožef Stefan Marand denis.pavliha@src.si natasa.planinc@infonet.si matjaz.depolli@ijs.si ales.smokvina@marand.si POVZETEK ● postavitev vmesnikov API za dostop in upravljanje z Pri prenosu podatkov od pacienta do informacijskega sistema EKG podatki na Think!EHR platformi zdravstvene ustanove se soočamo z več izzivi. Lokalno (Representational State Transfer Application shranjevanje podatkov meritev, kot je npr. zajem Programming Interface – REST API), ki omogoča elektrokardiogramskega (EKG) signala z uporabo merilnika Savvy EKG, predstavlja namreč dodaten in dostikrat nepotreben shranjevanje, branje in upravljanje z EKG podatki korak. Takšno lokalno meritev je nato namreč potrebno prenesti (Marand); do zdravstvenega osebja, da jo lahko obravnava. Z uporabo samodejnega shranjevanja meritev v pacientov zdravstveni karton ● prototip za zajem podatkov EKG meritev na Savvy (EHR) lahko takšen vmesni korak odstranimo in zdravstvenemu EKG prenosnem merilniku in prenos surovih podatkov osebju poenostavimo dostop do pacientovih meritev. Z razvojem neprekinjenih meritev [1] ter PDF poročil v Think!EHR prototipa za zajem EKG meritev na Savvy EKG prenosnem platformo [2] (IJS, Marand); merilniku in prenos surovih neprekinjenih meritev ter PDF poročil v Think!EHR platformo smo prikazali avtomatski in ● predvsem varen način prenašanja podatkov iz zunanjih sistemov v prototip za prikaz PDF EKG poročil v zdravstvenih centralno platformo. Razvili smo model za shranjevanje EKG informacijskih sistemih ISOZ21 in BIRPIS21, pri čemer podatkov in Representational State Transfer Application je vir podatkov Think!EHR platforma, kjer zdravstveni Programming Interface (REST API) vmesnike za prenos informacijski sistem poizveduje po poročilih za podatkov. Z uporabo prototipa integracije prikaza surovih EKG izbranega pacienta (SRC, Marand); podatkov pa lahko zdravnik na enostaven način išče podatke pacienta v zdravstvenem informacijskem sistemu. ● prototip integracije prikaza surovih podatkov EKG meritev v zdravstvenem informacijskem sistemu., pri KLJUČNE BESEDE čemer zdravstveni informacijski sistem poišče surove Integracija medicinskih naprav, EKG, Savvy, zdravstveni EKG podatke pacienta prek centralne Think!EHR informacijski sistemi, REST API. platforme in jih prikaže s pomočjo VisECG aplikacije (SRC, Marand, IJS). 1. PROTOTIPI Skupina partnerjev na tej nalogi, ki obsega Inštitut Jožef Stefan 2. PREGLED SORODNEGA DELA IN (IJS), Marand, SRC, Saving in Univerzitetni klinični center Ljubljana (UKC) ter Medicinsko fakulteto Univerze v Ljubljani OBSTOJEČE PODOBNE REŠITVE (MF LJ), je pripravila skupino prototipov, ki skrbijo za V zalednih zdravstvenih informacijskih sistemih že obstajajo dostopnost mobilnih elektrokardiogramskih (EKG) meritev, s tem določeni integracijski vmesniki, ki omogočajo komunikacijo z ko jih samodejno prenašajo od pacienta do zdravnika. EKG sistemi. Prav tako obstajajo prenosni merilniki EKG za daljše časovno obdobje, kot je npr. Holter. Prednost predlagane V okviru raziskovalnega dela smo si zadali izdelavo naslednjih rešitve je v tem, da omogoča povezavo med Savvy EKG prototipov: merilnikom [3] in platformo za shranjevanje medicinskih podatkov Think!EHR [2], od koder so podatki lahko brezšivno ● prototip za shranjevanje surovih EKG podatkov in EKG integrirani z zalednimi zdravstvenimi sistemi, ki jih prikazujejo poročil v Portable Document Format (PDF) obliki na zdravstvenemu osebju na enoten način. Ker so meritve na voljo šele po njihovem zaključku, sistemu dodajamo tudi možnost Think!EHR platformi skupaj z vmesniki za shranjevanje aktivne udeležbe pacienta. Slednji lahko med meritvijo doda in dostop do teh podatkov (Marand); komentar svojega počutja ali opombo o zaznani nepravilnosti v delovanju srca. Sistem za komentar avtomatsko ustvari poročilo s ● razvoj podatkovnega modela za shranjevanje EKG kopijo dela EKG signala v časovnem intervalu nekaj minut pred podatkov: analiza in postavitev modela za shranjevanje in po pacientovem komentarju. To poročilo tudi takoj posreduje v EKG podatkov v oblikah S2 in PDF skupaj z Think!EHR platformo, od koder je v skoraj realnem času na voljo metapodatki (Marand); zdravniškemu osebju. 39 3. OPIS PROTOTIPOV način omogočajo definirani vmesniki za dostop do teh podatkov Izdelani prototipi so namenjeni zdravnikom in njihovim enoten, vendar prilagodljiv način za shranjevanje in dostop do teh pacientom, ki potrebujejo spremljanje EKG izven okolja podatkov. zdravstvenih ustanov. Savvy EKG je prenosni medicinski Prototip za prikaz PDF EKG poročil v zdravstvenih pripomoček za trajno in natančno spremljanje srčnega ritma ter informacijskih sistemih povezuje Think!EHR platformo, ki opredelitev morebitnih odstopanj od normalnega [3]. Razvit je bil shranjuje EKG zapise z zdravstvenima informacijskima v sodelovanju z raziskovalno-razvojno skupino Laboratorija za sistemoma ISOZ21 in BIRPIS21. Zdravstveni informacijski vzporedne in porazdeljene sisteme, Odseka za komunikacijske sistem pridobi zapise EKG poročila pacienta iz Think!EHR sisteme na Institutu Jožef Stefan. Ustreza evropskim zdravstvenim platforme s pomočjo poizvedbe in ga prikaže v sami aplikaciji. Na standardom. Ima certifikat CE za celovito zasnovo in za vse ta način ima zdravnik dostop do podatkov EKG v svoji delovni sestavne dele. Savvy je preprost za uporabo. Zaradi brezžičnega aplikaciji, ki jo že uporablja, zato učenje uporabe nove delovanja, majhnosti in uporabljenih materialov je nemoteč tudi programske opreme ni potrebno. med delom, gibanjem in športom. Prototip integracije prikaza surovih EKG podatkov v S prototipom za zajem EKG meritev na Savvy EKG prenosnem zdravstvenem informacijskem sistemu omogoča, da zdravnik na merilniku in prenos surovih neprekinjenih meritev ter PDF enostaven način poišče surove EKG podatke pacienta, ki so poročil v Think!EHR platformo smo vzpostavili pomembno shranjeni v Think!EHR, ti pa se prikažejo s pomočjo zunanjega transportno pot za EKG meritve. Do sedaj so se meritve, prikazovalnika VisECG [3]. Prikazovalnik VisECG je prilagojen pridobljene iz Savvy EKG merilnika, shranjevale na pacientovem za prikaz meritev izmerjenih na Savvy ECG, ki so časovno mobilnem telefonu ali računalniku in niso imele ustaljene poti do neomejene (v praksi so navadno dolge med nekaj ur do enega zdravniškega osebja, ki bi bilo usposobljeno za njihovo tedna) in niso primerne za obstoječe komercialno dostopne pregledovanje in obdelavo. Prototip predstavlja izboljšavo v tem, prikazovalnike. Prikazovalnik podpira tudi prikaz dogodkov, ki da omogoča enostavno in sprotno prenašanje zajetih meritev do jih določi pacient ob boku EKG signal, tj. povezavo med ciljnega zdravniškega osebja. Odvisno od okolja, v katerem se prikazanim EKG signalom in EKG poročilom, ki je prav tako Savvy EKG uporablja, lahko pacient, njegov skrbnik, ali pa dostopno na Think!EHR platformi. medicinsko osebje poveže pacientovo mobilno napravo, ki Z razvojem prototipom smo pokazali, da je zadan cilj sprejema podatke iz EKG merilnika, prek internetne povezave v avtomatskega prenosa EKG podatkov z merilne naprave v pacientov elektronski zdravstveni karton (EHR). Ko je povezava zdravstveni informacijski sistem, ki ga zdravniško osebje že vzpostavljena, meritve samodejno shranimo v elektronski uporablja, zelo smiseln in povsem izvedljiv. zdravstveni karton. Na pacientovo željo (npr. ko se ne počuti najbolje oziroma kako zazna nepravilno bitje srca ali kako drugo motnjo v počutju) sistem izdela tudi avtomatsko poročilo s kratkim, vendar detajlnim izsekom meritve in ga shrani v 4. ZAHVALA elektronski zdravstveni karton. Prispevek je nastal v okviru raziskovalnega dela na projektu EkoSMART [4], ki ga sofinancirata Republika Slovenija in Prototip za shranjevanje surovih EKG podatkov in EKG poročil v Evropska unija iz Evropskega sklada za regionalni razvoj. PDF obliki na Think!EHR platformi omogoča skupaj z vmesniki za shranjevanje in dostop do teh podatkov shranjevanje podatkov 5. REFERENCE EKG meritev zajetih s Savvy EKG merilnikom. V ta namen smo razvili podatkovni model za shranjevanje EKG podatkov. [1] Inštitut Jožef Stefan, "eHealth monitoring system PCARD," Analizirali in postavili smo model za shranjevanje EKG podatkov Ljubljana, 2017. v S2 in PDF obliki skupaj z metapodatki in definirali predlogo [2] Marand, "Think!EHR Platform," Ljubljana, 2017. (angl. template) za EKG, ki vsebuje in strukturira podatke, ki jih dobimo s samega aparata. Predloga bo uporabna tudi za EKG [3] Saving d.o.o., "Savvy EKG," Ljubljana, 2017. meritve iz drugih virov in jo je mogoče razširiti z dodatnimi podatki, ki jih želijo spremljati in jih bodo definirali zdravniki. [4] Konzorcij EkoSMART, "Spletna stran projekta EkoSMART," Postavljen je bil vmesnik REST API za dostop in upravljanje z Ljubljana, 2017. EKG podatki na Think!EHR platformi in dodaten vmesnik REST API vmesnik za njihovo shranjevanje, branje in upravljanje. Na ta 40 Particle Accelerators as Medical Devices Marko Mehle, Luka Kurnjek Cosylab d.d. Gerbičeva ulica 64, 1000 Ljubljana luka.kurnjek@cosylab.com Abstract medical treatment. And even if its purpose as a component of a therapy machine is undeniable, as in Particle accelerators have many applications; one of the case of a dose delivery system, it is not possible these that is gaining popularity is their use in to judge its safety or effectiveness without knowing medicine for cancer therapy. Proton and Carbon Ion the properties of the remaining components of the therapy machines are used to accelerate a beam of system. Nonetheless, quite often manufacturers of particles and to deliver it very accurately into the components and parts for medical accelerators tumor area. This kind of treatment is called Proton, deliver products that are “certification ready”, Ion or simply Particle Therapy, and it presents meaning that the component meets all applicable several advantages over classic radiotherapy. standards and regulations and that all necessary documentation is provided to the manufacturer of the The accelerator machines involved in particle Particle Therapy Machine. therapy are fairly modest either in power and complexity; yet the distance between the lab and the hospital is great – especially in paper work. One of 2. Safety and Effectiveness the tasks that need to be fulfilled is to demonstrate There are two distinct and equally important aspects that the machine is safe and effective, as required by of medical devices that every medical device needs the standards and regulations. The purpose of this to fulfill: article is to present a starting point for understanding what it takes to make a Particle Therapy Machine 1. It must be safe. that is compliant with regulations. 2. It must be effective. It’s easy to understand why a Particle Therapy 1. Medical Devices system needs to be safe: it is of no use to have a A particle accelerator, though complicated by itself, machine that cures cancer, while it is very likely to it is not enough to conduct ion therapy on patients, injure a patient every once in a while. At the same and it cannot be considered a “medical device”. time, it is important for it to be effective, i.e. it is Why? Because an accelerator is a machine that capable of fulfilling its medical purpose. Again, a generates raw beam, without means of controlling or system that it is safe but it does not do what it is quantifying dose and position, and from a medical supposed to do is useless. Imagine a wheelchair (also perspective is useless and even dangerous. The a medical device, though a simple one) that is tightly system that we call a “Particle Therapy Machine” is screwed to the floor: it may be as safe as it gets, but composed of different subsystems; in general: one to it does not allow the patient to move around, as it is generate the beam, one to carry and direct it to the supposed to. treatment area (for example, beamline and gantry), That is why the laws regulating the market of one to measure and control its delivery (dose medical devices take care of these two aspects. These delivery/scanning), one to accurately position the are addressed in different ways. The manufacturer patient (robot and X-ray imaging), one to control the must demonstrate that his device is safe and operation and execution of the treatment (treatment effective, as required by the country where the control system), and various systems for control and device is to be sold. To prove safety, a lot of safety. They altogether constitute a “medical documentation must be provided: extensive test device”, as such it will have to be approved prior to reports, detailed design documents, compliance with its placement on the market. It is clear, for example, applicable standards, and so forth. In contrast, that it does not make sense to market a power supply effectiveness is proven either by doing clinical trials magnet or a “dose control system” alone as medical and investigations involving animals or humans, or devices since they do not have an intended use in by claiming that the device in question is similar to 41 other devices already being used, and demonstrating and the suitability of the chosen risk mitigation it by pointing to scientific articles or other literature. measures is the Risk Management, usually at the system level. 3. Medical requirements A common example from the Particle Therapy The regulatory requirements that a medical device or industry is to build the beam generation subsystem component needs to fulfil are different from country (the particle accelerator itself) as an industrial to country, but quite often they end up being quite device, and then to include in the system safety similar, since it is common that countries around the components (built as medical devices) that world recognize the use of international standards in constantly monitor the beam parameters such as their essence as a valid (and often necessary) energy, current and position, and shutoff the beam if mechanism to comply with their (local) regulations. any of these is out of tolerance. This strategy may All components of a Particle Therapy Machine can help to save resources, but not always, since some be very different in operation and purpose. Some complex industrial subsystems require safety come into contact with the patient or are installed measures sometimes so complex, that it is necessary within the treatment area, some not. Some are to build a whole redundant instance of the original mechanical and have moving parts, others can be system, and this only to monitor the behavior of the only electrical control and processing units. Some original component. As an example, let’s imagine an use high power or voltage, others not. Some contain ion therapy system in which, besides the particle or consist of software, some are fully analog or accelerator, the system that measures and controls discretely digital. And the list goes on. There are lots the delivery of the dose to the patient is also a non- of different regulatory requirements whose medical device. In that case, the only possible way applicability depends on all these factors. But there to mitigate the risk of patient overdose due to a are general things that are required for all “medical failure would be to have an additional medical components”. Regardless of the above mentioned component that measures and monitors the delivered technical factors, a Quality Management System beam. Such redundancy would be unnecessary. (such as [1]) must be established, a Risk That is why we recommend writing the medical Management (as [2]) process must be defined and compliance strategy very early in the design process. executed, and the safety and effectiveness of the (For those of you more interested in this aspect, I component must be proven. recommend taking a look at the standard IEC 60601- 1, section 16, and the informative Annex I). 4. Industrial requirements Do all the subsystems or components of a Particle 5. Medical Software and Verification & Therapy Machine need to be designed and Validation manufactured according to the regulatory requirements for medical devices, or is there any Today, every relatively complex medical device possible exceptions? This is an important question, contains at least some software. We all know that since devices that follow these are much more software is special, very different in nature and expensive than ordinary industrial stuff. The rule is behavior to hardware. Even digital hardware systems that some of the components or subsystems may not composed of discrete logic are not comparable to need to comply with the applicable medical software. The two main reasons for this are the standards, as soon as it is possible to demonstrate complexity of software, with its enormous number that their malfunction can never lead to an of possible internal states, and the usual software unacceptable risk, for the patient and for other people development workflow itself, which is in practice involved. Even in cases where one of the subsystems more organic and less constrained than hardware is inherently critical to the safety or the essential design. The first factor determines that it is performance of the whole system, it is possible to practically impossible to test a software module in mitigate the potential risk by implementing all its internal states, leaving a room for latent bugs. alternative safety measures. The logic behind this The second factor also influences reliability, because approach is that we cannot really trust industrial since the nature of software allows for quick changes (“non-medical”) devices since they were not or last-minute fixes (that can go unnoticed), there is designed and manufactured following quality always a temptation to do it without the necessary processes required by medical standards, therefore it care and to inadvertently introduce bugs. is assumed that they can contain bugs or can fail at If these two aspects are not addressed properly, any time. The mechanism to assess the initial risk software cannot be considered reliable enough for 42 medical devices, especially when there is a risk to people, for example, passing a list of tests people associated with it. That’s way the regulations automatically means that the device has no bugs, or and standards for medical devices take special care that the relentless application of risk management of software: to ensure that, although the possibility can give you a 0-1 type of indication about the of bugs cannot be totally eliminated, their probability degree of safety of a device. As a consequence of the can be reduced and its associated risk mitigated; with previous misconceptions, there are people that do good processes, proper testing, risk management and what is called “safety by paperwork”, which means validation. There is an international standard that tweaking the documentation and doing deals with this: “IEC62304: Medical device software ‘assessments’ until it is proven that the device is safe -- Software life cycle processes.” enough to meet the regulations. It is always possible to fool the system, and it is not in the scope of quality But software cannot exist without the associated systems, processes and technical requirements to hardware, and its proper operation in the context of completely avoid it. Yet, what standards and the whole machine cannot be judged without testing regulations do over anything else is to establish clear them together. Yet, in a system that is as complex as responsibilities and minimum requirements, so – a Particle Therapy Machine, there are many different when things go wrong – no one can say “I didn’t subsystems or components that may contain know”. software, and it is impractical or impossible to execute the full set of tests for every software On the other hand, it is important to understand that module, all integrated and operating in the final a good process helps you to keep organized through machine. That is why this is never done in that way, the development lifecycle. When projects are late or and there is always a strategy to hierarchize and about to go over budget, there is a natural temptation segregate testing into different levels, according to to rush and skip steps. This may appear to save some the system level architecture and the “V-model” time, but can also lead to dangerous mistakes. (Figure 1), in order to make testing effective and Having a well-established process naturally helps to practicable. This is the purpose of the set of activities resist the temptation. Furthermore, processes define usually referred to as “Verification and Validation”: what steps to take if there is an unexpected problem. to Verify, at different system levels, that the build And if the processes are wisely designed, they not system works as specified, and to Validate that it is only enhance safety but even make the development capable of fulfilling its intended purpose. More more efficient by keeping pace, consistency and about verification and validation of Programmable making sure that no aspect or requirement is left Electrical Medical Systems can be found in the neglected. standard IEC60601-1, section 14, and informative annex H. 7. Conclusion Building a Particle Therapy system is a technical challenge by itself. To make it compliant with standards and regulations it is yet another challenge, which is sometimes underestimated. Our experience shows that it is best to start designing the medical compliance strategy together with the machine itself and to already have solid processes established when the development and implementation start. Smart processes not only help to design a safe device and get through compliance; they may also increase Figure 1: V-model productivity and decrease the uncertainties and defects, which may be, at the end of the day, the 6. Purpose of Processes and Standards differentiators for staying in the game. Sometimes I hear, mostly amongst engineers and 8. References physicists, arguments about whether following the standards, regulations and processes can really [1] ISO 13485:2003: Medical devices -- Quality enhance safety and effectiveness, or is it just a management systems -- Requirements for regulatory “placebo” to create an appearance of safety, which purposes (https://www.iso.org/standard/36786.html) somehow relieves the engineering conscience from [2] ISO 14971:2000: Medical devices -- Application the responsibility of thinking. The argument is valid of risk management to medical devices and it has to be taken seriously. I find that for some (https://www.iso.org/standard/31550.html) 43 MOBILNO SPREMLJANJE OKOLJSKIH DEJAVNIKOV IN NJIHOVEGA VPLIVA NA ZDRAVJE Andrej Golija, NELA razvojni center d.o.o 1. UVOD V projektu EkoSmart, ki je sofinanciran s strani Ministrstva za izobraževanje, znanost in šport ter Evropskega sklada za Izbor senzorjev je temeljil na podlagi rezultatov izvedenih regionalni razvoj, bomo v razvojnem centru Nela razvili laboratorijskih testiranj, izvedenih na podlagi tehničnih inteligentni sistem za spremljanje okoljskih dejavnikov. Zavedamo protokolov, skladnih s standardom ISO 17025. Ustreznost se, da danes povprečni človek preživi približno 80-90% svojega delovanja senzorja smo spremljali predvsem z vidika ustrezne časa v notranjih prostorih, kjer pa kakovost zraka ni vedno točnosti in stabilnosti. Za pripravo končnih poročil z rezultati o ustrezna. Zato je z vidika ohranjanja udobja in zdravja človeka, izvedenih meritvah smo uporabili inovativni laboratorijski ključnega pomena spremljanje okoljskih parametrov, katere informacijski sistem. lahko z enostavnimi ukrepi tudi ustrezno obvladujemo. Cenovno dostopni senzorji in mobilna aplikacija, bodo predstavljali dve Tekom projekt smo testirali tudi delovanje različnih načinov ključni komponenti inteligentnega sistema za monitoring komunikacije za posamezne senzorje in sicer I2C, Modbus, RS okoljskih parametrov. Ključna inovativnost projekta – avtomatska 485, RS 232, wi-fi, analogno odčitavanje Pt100/Pt1000, 0-10V, 4- regulacije okoljskih pogojev v prostoru, pa bo še dodatno 20mA, s spremljanjem dometa pri različnih okoljskih pripomogla k obvladovanju in vzpostavitvi idealnih okoljskih obremenitvah. Rezultati testiranj so pokazali, da je Modbus še pogojev za bivanje. vedno najbolj stabilen princip komunikacije. 2. SENZORIKA 3. MOBILNA APLIKACIJA Kot ključne parametre, ki lahko vplivajo na človekovo počutje in Mobilna aplikacija bo pri uporabniku predstavljala enega izmed zdravje, smo določili temperaturo zraka, relativno vlago, hitrost ključnih izbirnih parametrov. Zato smo tekom razvoja, zraka, osvetljenost, koncentracija CO2, koncentracija VOC prilagajanja in nadgradnje sistema, poleg zahtevanih tehničnih (nevarne organske spojine v zraku), radon. Na podlagi definiranih karakteristik, upoštevali tudi enostavno in logično infografiko, ki tehničnih karakteristik smo izvedli testiranja senzorjev različnih bo uporabniku na atraktiven način predstavila le ključne proizvajalcev, ki so dostopni na trgu. informacije. Spodaj so podani primeri izbora grafičnih elementov Tabela 1: Tehnične karaketristike izbranih senzorjev Parameter, veličina Merilno območje Odčitavanje Točnost Temperatura od -20 °C do 60 °C 0,1 °C 0,2 °C Relativna vlaga od 0 % do 100 % 1 % 2,5 % Zračni tlak od 500 mbar do 1150 mbar 1 mbar 2 mbar Osvetljenost do 100.000 lux 0,25 lux Nivo CO2 do 10.000 ppm (vol) 20 ppm 1 % 30 ppm 5% Nivo VOC 450 ppm - 65.000 ppm (equiv.) 1 ppm Hitrost zraka 0 do 2 m/s 0,001m/s 0,001 m/s Senzor odprtih vrat Odprto/zaprto 1/0 / Radon 0.1 ~ 99.99 pCi/l 0.5cpm/pCi/l < 10% 44 Senzor za Pomoč temperaturo Senzor za Nastavitve koncentracija CO2 Senzor za hitrost Alarm zraka Kot ključne lastnosti mobilne aplikacije smo opredelili: izpis on- line meritev in pregled zgodovine, izpis min/max/povprečne vrednosti, drevesna struktura (pregled po lokacijah), alarm (sms, e-mail), izdelava poročil, modularna izvedba. Z definiranjem koncepta povezovanja database – service - client ter z nadgradnjo MS SQL serverja smo vzpostavili ustrezno okolje za beleženje, shranjevanje in obdelovanje podatkov izmerjenih vrednosti. Prikaz on-line meritev, ki je prikazana na spodnji sliki, uporabniku omogoča vpogled trenutnih vrednosti. Slika 1: Meritve v okolju. 4. ZAKLJUČEK Nadaljnji razvoj bo temeljil predvsem na razvoju sistema za pregled zgodovine meritev, izpisu poročil, alarmiranja in načinu uporabniškega dostopa. Poleg tega se bomo osredotočili na procese, ki bodo predstavljali korak k doseganju avtomatske regulacije okoljskih pogojev. ZAHVALA Delo je nastalo v okviru programa EKOSMART. 45 Application for Viral Hepatitis Infection Risk Assessment Alen Ajanović1, Karolina Počivavšek2**, Matic Podpadec1, Andrej Ulčar, Ana Marija Peterlin2,Ana Prodan2, Saša Rink2, dr. Anton Gradišek3, prof. dr. Matjaž Gams3, dr. Gašper Fele-Žorž1*,prof. dr. Mojca Matičič, MD* 1Faculty of Computer and Information Science, Večna pot 113, 1000 Ljubljana *polz@fri.uni-lj.si 2Medical Faculty, University of Ljubljana, Vrazovtrg 2, 1000 Ljubljana **karolina.pocivavsek@gmail.com 3Institut Jožef Stefan, Jamova 39, 1000 Ljubljana, Slovenia 4Clinic for Infectious Disease and Febrile Illnesses, University Medical Centre Ljubljana, Japljeva 2, 1525 Ljubljana, Slovenia *mojca.maticic@kclj.si ABSTRACT symptoms until the disease has already developed to an Globally, an estimated 330 million people are chronically advanced stage, at which point it becomes difficult to treat. The infected with hepatitis viruses. With the aim to provide problematic element of this is that a symptomless infected information on the risks for acquiring infection and educate on person can also transmit the virus to other people. Since it is its signs and symptoms, a web application was developed for possible to successfully treat or at least manage patients with raising awareness in general as well as high-risk populations. viral hepatitis infections, it is important to identify the infected The website contains a questionnaire and informational web individuals as soon as possible and act accordingly. People often pages, with the main focus to help the users disclose their seek medical information online, therefore it is important that all potential exposures to viral hepatitis infection during their past relevant information is available in a centralized location. Here, life and brings about prevention measures to avoid it in the we present a web application that aims to educate users about future. Thus the application serves as an educational as well as a viral hepatitis infections and to assess possible risks of infection. preventative tool, helping the users learn more about the dangers The application is a follow-up of an application to educate about and causes of viral hepatitis infections and, in case of exposure, and assess risks for sexually transmitted infections ASPO [5], refer to proper medical services. It uses a robust system that built on an improved platform with modified goals. We discuss provides anonymity and also delivers the content to the user in a the implementation and the functionality. responsive manner. 2. PROCESS/METHODOLOGY Keywords Creating a website that provides medical information requires Viral hepatitis, infection, web application, questionnaire, use of scientifically proven facts to build user's trust. Website diagnosis, risk level assessment, informative, sexually design and transparency of the developers also play an important transmitted infections contribution in this aspect. Therefore, the site has been built from the bottom up with external libraries that don't collect any 1. INTRODUCTION user information. To address transparency, the project Hepatitis is an inflammation of the liver tissue as a result of description and information is accessible on the main site, with infectious or noninfectious causes. In more than half of the cases references to the individuals that have helped build it. The users it is caused by hepatitis viruses that have currently been can immediately recognize that the site was also officially estimated to affect 330 million of people worldwide [1]. Viral endorsed by reputable sources by seeing logos accompanying hepatitis is caused by one of the five hepatitis viruses: hepatitis the main site. The other part of the initial design is that the site A virus (HAV), hepatitis B virus (HBV), hepatitis C virus had to be simple and clearly indicate from the very start what it (HCV), hepatitis D virus (HDV) and hepatitis E virus (HEV) is offering. Server-side logic is managed by Django [6], a [2]. It is estimated that less than 1% of the population in python based web framework. Another part of achieving user Slovenia is infected with HBV and around 0.4% with HCV [3]. trust is that the site cannot track or use any data that the user has Infections by HBV, HCV and HDV can be transmitted by provided, either voluntarily or involuntarily (e.g. recording IP infected blood, sexual intercourse or from mother to child, address, cookies, tracking data). As a result, the site does not use whereas HAV and HEV are mainly transmitted by contaminated any cookies and only stores information from the questionnaire food or water. Viral hepatitis may be present as an acute or a if the user has agreed to it (for the purpose of potential statistical chronic disease. Infection with each of the five hepatitis viruses analysis in future). can result in acute disease which can be symptomless or presents 2.1 ETHICAL CONCERNS with symptoms of nausea, fatigue, abdominal pain and jaundice When building the question base, we had to make sure to [4]. The natural course of acute HBV and HCV infection can include questions that would include an informative narrative, lead to a chronic one. Due to common lack of signs and but also would not capture any of the personal information symptoms, patients with chronic viral hepatitis infection are the which could be used to identify the user. Even though we don't main source of spreading the disease, mostly by risk behavior. need or use any user data, we believe they could still be useful Furthermore, chronic HBV and HCV infection may lead to for statistical and health studies. We've decided to implement a develop cirrhosis and hepatocellular carcinoma which can lead simple solution, to put a question at the end of the questionnaire to liver failure and death [2]. The main problem with hepatitis which would ask for the user’s permission to anonymously use virus infections is that the patients typically do not exhibit 46 the answers they have provided through solving the countries without proper precautions, eating uncooked food, questionnaire. This required to change our database logic a bit piercings, etc.). The page includes several menus, each being since we thought it would be best if this particular question divided into multiple submenus, based on context. Each page is would append to a questionnaire, as opposed to it just being accompanied by a static background. treated like another question. The only elements that are stored in the actual database are a unique user id (which only translates to the sequence of the users as they’ve taken the questionnaire) and their recorded answers. 2.2 SAFETY/ANONIMITY It is important that the users feel secure while using the website, since many of the questions tackle sensitive personal information. As the site does not use any cookies, none of the user information is stored locally. The website also uses a secure connection (HTTPS) to the server to make sure that all of the traffic is encrypted. The end answers are stored on a server without any identifiable information where only authorized Figure 1 Screenshot of the webpage personnel have actual access to the physical machine. 3.3 BACKEND 3. SYSTEM The majority of the work was done on the backend of the 3.1 TECHNICAL CONSIDERATIONS application. We looked at a few different possible web When we began building the new system we had to consider if it frameworks that could have been used, but ultimately, we've would be wise to keep the old code base from a previous project, decided on Django because of its familiarity, simplicity and ASPO [5]. Due to differences, this project would require a adaptability with other libraries we would be using throughout completely different approach and we had to warrant different the project. The first part was creating the database that would needs that this project will inevitably require. We had ultimately hold all of the necessary information (page layout, questions and decided to keep the frontend largely the same and to focus on answers, unique identifiers, etc.). While the initial premise was addressing the issue of too many connections crashing the relatively simple, as the project evolved we had to address server. different concerns and keep adding more intricacies to the working logic. An example of this would be addressing different We were also dealing with a different rule set for questions, now types of answers. They could've either been a multiple choice, the questions had to display a comment depending on the answer single choice or a completely custom input. While we don't use that the user had picked. There is a possibility of multiple any custom inputs, it was important to assure that the database comments on the same question as well as different answers would be scalable if the need arose. Some of the other elements leading to the same comment. An additional problem arose as that we had to keep in mind were an adaptable quantity of we had to change the functioning algorithm of displaying the questions, different methods of input and most importantly, the questions in the first place. ability to change the underlying system at any time. The most Finally, instead of a generic screen at the end which would only important part was that the system had to be designed in a way tell you an arbitrary risk level, the new website outlines which could be operated without the need of a computer expert. everything you have clicked and presents a list of behavior The way this works is that every answer has a parameter which (based on your answers) that could be considered promiscuous disables a particular question based on its unique identifier. This or risky. We approached this problem by scratching everything means that before the next question is loaded, we check if that we had done for the original project and focused on building a question has been disabled by another answer previous to that new system with these new requirements. and with that, we don’t display it and skip right onto the next one. This system allows anyone to manipulate the rules and the 3.2 FRONTEND order of these questions, provided they have the access to the The frontend consists of a Bootstrap CSS [7] base which is database. mainly used for its ability to work on different devices while maintaining its integrity. The interactive part of the website Regarding the REST (Representational state transfer) interface, (questionnaire, switching menus, loading subpages) is built we've decided to use Django REST framework [9], which serves primarily using AngularJS [8], which serves and loads the files as an intermediary between the information that consists in the as they are requested. The questionnaire itself is transferred once database and a URL, which allows both for a way to access it when it is started and the rest is taken care of by AngularJS, from an outside, secure source, as well as allow the system to which properly switches between questions and comments. This reach it through internal means, after which it can be de- way, even if the user disconnects in between, the questions and serialized and properly displayed as the questions appear. REST their answers are temporarily stored in the scope variable. When is a system commonly used to simplify the connection between the questionnaire is finished, the user locally receives the the database and the program itself. The service itself provides a response and only in the case if they have agreed to store predefined set of operations in order to access, modify or change anonymous answers into the database, the connection to the data that is on the server. server is established again. The response consists of summarized comments that have appeared through the questionnaire, relating The next part of the backend are the static pages. Django allows to possible exposures to an infection. If the user had any risk for a fine distinction between dynamically and statically loaded factors, the background is colored orange or green (risk factors pages. For example, the problem was that both AngularJS and might include not being vaccinated, travelling to different static pages are mainly descriptions of different diseases and 47 guidelines of what to watch out for if you want to avoid an a small margin of the population has to be vaccinated infection. There is also a set of pages containing the information more than three times.) about where to find the proper treatment in case if you want to Q: "Are you or your parent born in the countries where hepatitis get tested or treated for the general hepatitis infections. Static B infections are common?" pages load up as soon as the user requests that particular subpage and they are fetched from the server and transferred to 1. Yes (There exists a higher chance of infection if client. AngularJS then displays them, providing the information you or your parents are born in a country where that the user has requested. hepatitis B is common.) 2. No 4. QUESTIONNAIRE The full questionnaire consists of 28 questions in total with the 5. SUMMARY intent of reducing these questions as much as possible, so the As number of viral hepatitis cases have risen up in the past few user is not required to go through every single one of them in years, we have combined modern technology alongside with order to get the final risk assessment. There are several modern design principles and the latest knowledge about viral questions where the answers trigger/disable sub-questions, hepatitises to create a web application for informing users about therefore reducing the total number of questions that the user has the topic. Awareness and risk assessment are very important as to take. For example, by answering “No.” to the question “Have they influence more people to get tested for viral hepatitis you travelled somewhere in the past three months?” the infections and get appropriate treatment in case they have the questionnaire would omit questions about travel. disease. We have several goals, the first and the utmost important one is to decrease the number of newly discovered Following some general demographic questions, the 4th question patients with end stage diseases. Since viral hepatitis can be asks the user about their specific interests, related to groups of treated if discovered early enough, that can also prevent the possible risks - travel, lifestyle, exposure to hepatitis infection in disease from spreading. Another element of this is information, the past, and medical issues. Based on the answer, the questions as spreading it will allow people to be more aware of the proper from the group of interest are displayed first. precautions, which will hopefully reduce overall exposure to Each question can either be a “checkbox” or a “radio” type hepatitis infections. If enough users agree to it, we also hope to question, where the former allows the user to select multiple collect enough relevant medical data that will allow for further choices and the latter allows the user to only select one. We analysis on addressing this growing problem. With gathering have discussed the possibility of open-ended answers (where the this epidemiological data, we hope to help more easily identify user would input their own values), but have decided against it those with a higher risk factor and urge them to seek treatment as we felt like the constraints work better in terms of providing in the earlier stages of the disease. By introducing this website, useful feedback for analysis. we hope to allow the users to access useful and verified information all in one place. Some of the questions also have image components to them. These provide a visual representation of what the question is 6. REFERENCES currently asking. All of these types of questions have maps [1] WHOhttp://www.who.int/mediacentre/factsheets/http:/ included in them that show where the infection of a particular /www.who.int/mediacentre/factsheets/ disease is most prevalent. For example, a question like “Have Accessed: 18-09-2017 you been born in any of the countries where hepatitis A is prevalent?”, would have an image showing the different [2] A. Wasley, S. Grytdal, and K. Gallagher, “Surveillance for Acute Viral Hepatitis - United infection rates for countries around the world. States, 2006,” MMWR Surveillance summaries, 2008. The final and the most important part of the questionnaire is that [Online]. Available: we wanted it to have both an educational, as well as an https://www.cdc.gov/mmwr/preview/mmwrhtml/ss570 informational component. That would allow for the user to see 2a1.htm. Accessed: 23-Aug-2017. which of their answers/actions have an inherent risk to them. [3] J. Tomažič and F. Strle, Infekcijske bolezni, vol. 1. This is most obviously seen in the comments that show up Ljubljana: Združenje za infektologijo, Slovensko depending on what the user has answered. Sometimes they are zdravniško društvo Ljubljana 2014/2015. purely informational (e.g. “A vaccine for hepatitis B will decrease your likelihood of contracting the disease”), while in [4] N. A. Terrault, N. H. Bzowej, K.-M. Chang, J. P. other cases they can be educational as well (e.g. “In order to Hwang, M. M. Jonas, and M. H. Murad, “AASLD guidelines for treatment of chronic hepatitis B,” make sure you are treated against a particular hepatitis infection, you need to have a sufficient amount of anti-bodies”). These Hepatology, vol. 63, no. 1, pp. 261–283, Jan. 2016. comments are aggregated and at the end shown back to the user. [5] ASPO web application https://aspo.mf.uni- lj.si/Accessed: 15-09-2017 Some question/comment examples: [6] Django https://www.djangoproject.com/ Accessed: Q: "Are you vaccinated for Hepatitis B?" 31-08-2017 1. Yes, I was vaccinated within the national program [7] Bootstrap http://getbootstrap.com/ Accessed: 31-08- for children before entering school (born after 1992) 2017 2. Yes, I was vaccinated individually [8] Angular JS https://angularjs.org/ Accessed: 31-08- 2017 3. No (We suggest to get vaccinated for hepatitis B and checking if the vaccination was successful after [9] Django REST framework http://www.django-rest- the third dose. In order for the vaccine to be effective, framework.org/ Accessed: 31-08-2017 48 Spletno svetovanje študentom v stiskah Tomaž Šef Aleš Tavčar Miha Mlakar Matjaž Gams Institut “Jožef Stefan” Institut “Jožef Stefan” Institut “Jožef Stefan” Institut “Jožef Stefan” Jamova cesta 39 Jamova cesta 39 Jamova cesta 39 Jamova cesta 39 1000 Ljubljana 1000 Ljubljana 1000 Ljubljana 1000 Ljubljana +386 1 477 34 19 +386 1 477 35 64 +386 1 477 36 33 +386 1 477 36 44 tomaz.sef@ijs.si ales.tavcar@ijs.si miha.mlakar@ijs.si matjaz.gams@ijs.si POVZETEK 2. NAMEN PROJEKTA V članku opisujemo študentski projekt »Spletno svetovanje S spletnim svetovanjem študentom v lokalnem okolju se želi študentom v stiskah«. V okviru projekta študenti ob pomoči spodbudili preventivno ukvarjanje z mentalno higieno in tudi pedagoškega mentorja in strokovnega sodelavca iz raziskovalnega omogočili lažji dostop do strokovne pomoči v primeru osebnih zavoda interdisciplinarno proučujejo in razvijajo metode za težav in duševnih stisk. V splošnem bo rešitev prispevala k zgodnje odkrivanje stresa in nudenje pomoči preko spleta. zmanjševanju stigmatizacije ljudi s težavami v duševnem zdravju. Konkretno bo rešitev pomagala študentom naučiti se ustrezno Aplikacija, ki je še v razvoju, bo omogočala enostaven dostop do pristopati k stresnim situacijam in imela pozitivne učinke na strokovno zasnovanih storitev pomoči v primerih, ko se študent zvišanje zadovoljstva in motivacije za študij ter splošno v sooča s stresom ali s telesnimi ali duševnimi težavami kot vsakdanjem življenju. Spodbujanje razvoja konstruktivnih posledicami stresa. Aplikacija bo sprva na razpolago študentom strategij spoprijemanja s stresom bo pozitivno doprineslo tudi k Univerze v Ljubljani, ki pogosto sodelujejo z Institutom »Jožef zaposlitvenim možnostim in tudi zadovoljstvu delodajalcev. Stefan«, kasneje pa bo dostopna tudi za druge visokošolske ustanove in univerze v regiji ter širše družbeno okolje. Tekom projekta se razvija sodelovanje, komunikacija in izmenjava znanj med akademskimi in raziskovalnimi ustanovami Opisane so tehnične specifikacije in druge karakteristike sistema. ter neposreden prenos tega znanja na študente. Študentje Ključne besede pridobljeno teoretično znanje uporabljajo pri reševanju praktičnih in aktualnih družbenih problemov. Pridobivajo dragocene Spletno svetovanje, stres, stresna situacija, prepoznavanje stresa, izkušnje in s tem boljše možnosti za hitro prvo zaposlitev, kar obvladovanje stresa, študent, asistent. prispeval k zmanjševanju brezposelnosti med mladimi. Predvsem pa se vzpostavilo tvorno sodelovanje med tehnično in 1. UVOD družboslovno stroko. Stres je fiziološki, psihološki in vedenjski odgovor posameznika, ki se poskuša prilagoditi in privaditi potencialno škodljivim ali V okviru projekta študenti proučujejo različne inovativne rešitve ogrožajočim dejavnikom, ali drugače stresorjem. Stres prinese za izzive družbenega okolja v katerem živijo in na takšen način spremembe s katerimi se moramo soočiti, to pa počne vsak na osvojijo nov način razmišljanja, ki povečuje njihovo ustvarjalnost svoj način, in kreativnost. To pa je dobra podlaga za kasnejši razvoj novih ki se mu zdi najbolj primeren. [1] inovativnih storitev z visoko dodano vrednostjo. V Evropi trpi zaradi duševnih motenj vsaj enkrat v življenju tretjina prebivalstva. Ta procent se zaradi družbenih sprememb, 3. OPIS VSEBINE PROJEKTA nezaposlenosti, pritiska na srednji razred itd., viša. Uspešne prakse v tujini kažejo, da postaja obravnava psihičnih težav vse Pri obvladovanju stresa si lahko med drugim pomagamo z različnimi tehnološkimi pripomočki ali storitvami kot npr. bolj interdisciplinarna in podprta z novimi informacijskimi tehnologijami, ki se kljub zavedanju možnega tveganja, čedalje zapestnica za stres in spletno svetovanje. bolj širijo na vse ravni podpore. Takšni pristopi so še posebej V okviru projekta se razvijajo računalniški algoritmi za zanimivi za mlade generacije, ki se jih najpogosteje poslužujejo. prepoznavo stresa (npr. na podlagi spremembe srčnega utripa, Počasi se uporaba prenaša tudi na starejše generacije in zajema spremembe v govoru ipd.), kar je zahteven in še ne dovolj dobro celotno družbo. rešen problem. Izdeluje se spletna aplikacija, ki bo predstavljala Stresno življenje je značilnost moderne, hitro razvijajoče se, strokovno osnovano psihosocialno prvo pomoč za študente v uveljavitveno usmerjene družbe. Ko študenti vanjo resno in stiski. S tem bo študentom omogočen prvi strokovno osnovan odgovorno vstopajo, morajo izpolnjevati visoke zahteve iz okolja, svetovalno terapevtski stik, za katerega se jim ne bo potrebno hkrati pa se soočati z lastnimi dvomi, sposobnostmi in interesi. vključiti v socialno zdravstveni sistem. Za to je potrebno Zato so v tem obdobju strategije spoprijemanja s stresnimi analizirati in izbrati ustrezen psihoterapevtski pristop, opredeliti ključne koncepte, način kategorizacije in oblikovati situacijami izrednega pomena. Marsikdaj se prav tedaj izkažejo za ustrezne neučinkovite, kar lahko scenarije dialoga, vaje ter napotke. Načrtovani spletni svetovalec pripelje do blažjih ali hujših duševnih stisk in zdravstvenih težav, ki se bo študente opremil z osnovnimi strategijami za diagnostiko in z neukrepanjem še poglabljajo. obravnavo njihove težave, jim zagotovili ustrezne informacije in [2] jih po potrebi napotili v ustrezne inštitucije. Aplikacija bo Zdravstveni podatki kažejo, da so težave zaradi stresa eden izmed študente tudi preventivno usmerjala k zdravemu načinu življenja s ključnih izzivov prihodnosti. [3, 4] splošnim izobraževanjem o stresu, načinih spoprijemanja, 49 tehnikah sproščanja, pomembnosti prave socialne podpore, vpliva Naslednji korak v pogovornem svetovalcu je klasifikacija v pravo (ne)zdrave prehrane in športne aktivnosti. kategorijo stresa. Sistem poskuša preko zaznavanja ključnih besed in fraz klasificirati tekst v določeno kategorije stresa, saj je Tekom izvajanja projekta bo na osnovi teoretičnih izhodišč mogoče na podlagi te informacije uporabniku posredovati zastavljena klasifikacija študentskih težav povezanih s stresom, specifična vprašanja, naloge ali voden scenarij, ki naj bi pomagal opisi simptomatike in ustrezne obravnave, razvit vprašalnik za pri spopadanju s tovrstno obliko stresa. Pri razpoznavanju smo diagnosticiranje in klasificiranje težav ter pripravljen osnovni uporabljali platformo API.AI [5], ki je namenjena analizi načrt svetovanja z nalogami za uporabnike z različnimi težavami. naravnega jezika (Natural language processing). V sistem se Preizkušeni bodo različni pristopi avtomatske detekcije stresa. S vnese kategorije v katere želimo klasificirati tekst, nato se poda tem bo razvita prva verzija aplikacije, ki bo študente s težavami učne primere pogovorov, ki so značilni za tovrstno kategorijo. V zaradi stresa vodila skozi njihovo problematiko, jih učila, API. AI se kreira novega agenta, za katerega lahko, z uporabo ozaveščala, motivirala, jim ponujala prilagojene domače naloge. strojnega učenja in pravil, ustvarimo enostaven model, ki je Aplikacija bo vključevala tudi povezavo z bazo terapevtov in sposoben klasifikacije podanega teksta v naravnem jeziku. Ta svetovalcev ter osebnih storitev, ki bodo vključeni po potrebi v agent nato lahko program pomoči. Vse bo na uporabniku prijazen in prilagojen način integrirano v spletno okolje. Aplikacija bo testirana in Preko anketnega vprašalnika se je identificiralo vsebinske pripravljena za osnovno uporabo. kategorije stresa pri študentih po pogostosti, od najpogostejše do najmanj: 4. ZASNOVA OZ. SPECIFIKACIJA 1. Študij – Urnik predavanj, stres med izpitnimi obdoblji. SISTEMA 2. Denar – Preslabo plačano delo, odvisnost od staršev ali sistema. 3. Odnosi – Največkrat s starši ali partneji. Groba zasnova sistema je prikazana na sliki 1. Uporabnik lahko 4. Osamljenost – Oddaljenost od doma, prejšnjih prijateljev ali najprej izpolni spletni vprašalnik, ki oceni stopnjo stresa pomanjkanje hobijev. uporabnika. Na podlagi doseženega rezultata se uporabniku, v primeru velikega stresa, priporoči direkten pogovor s terapevtom. 5. Samopodoba/samozavest – Nezadovoljstvo s samim seboj. Če pa se iz vprašalnika zazna blažjo stopnjo stresa, se predlaga 6. Bolezen – Pri študentih ali družini, depresija, anksioznost itd. pogovor s spletnim pogovornim svetovalcem. Ob potrditvi se 7. Družinske tragedije – Smrt starša, sorodnikov. preusmeri pogovor na spletni obrazec, ki omogoča vnos teksta in prikaz pogovora s virtualnim svetovalcem. 8. Selitev v mesto – Predvsem iz manjšega kraja v Ljubljano. Vse kategorije se bo v nadaljevanju izvajanja projekta vneslo v agenta znotraj API.AI in vsaki kategoriji dodalo množico učnih primerov (teksta), ki so značilni za tovrstno kategorijo stresa. Predvidevamo, da pri pogovoru o problemih s študijem bo v tekstu več besed ali zvez, kjer so omenjeni urniki, predavanja, izpiti, naloge ipd. Strojno učenje bi moralo take značilne, ključne koncepte zaznati in jih vključiti v odločitveni model za to kategorijo. 5. PRIMER POGOVORA Uporabniku je ob prihodu na stran ponujena možnost izpolnjevanja vprašalnika, ki oceni njegovo stopnjo stresa. Del pripravljenega vprašalnika je prikazan na Sliki 2. Uporabniku se postavi množica vprašanj, kjer je mogoča izbira enega ali več predpripravljenih odgovorov. Aplikacija ovrednoti odgovore in rezultat prikaže uporabniku v grafični obliki. Uporabniku nato ponudi direkten pogovor s terapevtom ali pa pogovor s spletnim svetovalcem. Slika 1. Shema sistema za spletno svetovanje študentom v stiski Pogovorni svetovalec prejeti tekst najprej posreduje modulu za predprocesiranje teksta, ki naj bi odstranil bolj pogoste besede, ki ne prinašajo dodatnih informacij. Preverili bomo tudi možnost dodajanja funkcionalnosti urejanja sintaktičnih napak v tekstu. 50 Uporabnik: Živjo Svetovalec: Pozdravljeni, kako vam lahko pomagam? Uporabnik: Saj ne vem kako razložiti zadeve, ker so občutki mešani in se počutje spreminja enkrat na dobro, drugič na slabo. Svetovalec: Ali lahko prosim opišete kakšno situacijo, ki vas spravlja v stresno situacijo. Ko sistem klasificira opis uspe klasificirati, zahteva potrditev s strani uporabnika. V primeru, da uporabnik zavrne predlagano napako, ga svetovalec pozove naj še na drug način opiše probleme s katerimi se sooča. Ko uporabnik potrdi ustreznost opisa njegovega razmišljanja sistem uporabi določen scenarij, to je pogovor, ki pomaga uporabniku, da spremeni svoj način razmišljanja. Kratek primer: Svetovalec: Če bi ocenili na lestvici od 1 do 10 (1 pomeni »me sploh ne obremenjuje«, 10 »neizmerno me obremenjuje«). S katero številko bi ocenili vaše trenutno stanje? Uporabnik: 6 Svetovalec: Zapišite sedaj situacijo čez 1 leto, pomislite, da se še vedno ni nič spremenilo. Vse ostaja enako, vaše Slika 2. Del anketnega vprašalnika za ugotavljanje stopnje težave so še vedno tukaj, morda še v večji razsežnosti, stresa. kot so bile pred letom dni. (vzemite si čas preden zapišete). Pogovor med uporabnikom in spletnim svetovalcem poteka v Uporabnik: To bi bilo nevzdržno, saj se že zdaj dnevno naravnem jeziku. Uporabnik postavi vprašanje in sistem mu obremenjujem s tem in si predstavljam, da bi to odgovori, svetuje ali poda naloge glede na določen scenarih. precej vplivalo na moje takratnje počutje. V prvi fazi se najprej na osnovi uporabnikovega opisa problema Svetovalec: In sedaj si zamislite isto situacijo čez 5 let, še vedno klasificira miselna napaka. Primer pogovora: ne naredite nobene spremembe, ker verjamete, da tako ali tako nima smisla. Venomer se sprašujete, kako bi bilo, če bi se odločili in se podali na drugo Uporabnik: Živjo pot, če bi poizkusili narediti nekaj drugače. Stres, ki ga doživljate sedaj je narasel za petkratno vrednost? Svetovalec: Pozdravljeni, kako vam lahko pomagam? Svetovalec: Napišite mi, kako živite, kako se počutite, kako Uporabnik: Kaj znaš? izgledate po toliko letih neprestanega stresa, Svetovalec: Sem virtualni svetovalec, ki vam lahko svetu preko premišljevanja, kaj bi bilo, če bi si dovolili narediti pogovora. Prosim napišite težave ali probleme s drugače? katerim se soočate in poskušal bom zaznati miselno … napako ter ustrezno prilagoditi pogovor in naloge. Sem še v razvoju in zato vsega še ne znam. Uporabnik: Naslednji teden začnem s študijem in se bojim, da se Scenarij pogovora se lahko še nadaljuje in uporabniku posreduje ne bom vklopil in našel novih, pristnih prijateljstev. nasvete ali naloge. Vsak scenarij, zbirka nalog ali pogovorov je za Svetovalec: Iz podanega opisa problema je mogoče razbrati, da vsako kategorijo ali miselno napako različna in se sproži po uspešni klasifikaciji. Ta mehanizem daje upravljavcem širok pri vas obstaja nagnjenost k pretiranemu osredotočanj nabor možnosti spopadanja z različnimi tipi stresa. u na (negativno) prihodnost, temu strokovno rečemo Napovedovanje prihodnosti, kar preprečuje delovanje, sprevrže se v samo 6. ZAKLJUČEK izpolnjujočo prerokbo. Predstavljen je bil projekt za spletno svetovanje študentom v Svetovalec: Ali se vam zdi, da je to značilno za vas? stiskah, ki se še izvaja. S projektom bo nastala osnovna verzija aplikacije za svetovanje študentom. Aplikacija se bo nato Uporabnik: Da integrirala z drugimi obstoječimi spletnimi funkcionalnostmi na področju svetovanja. V primeru, da sistem ne zmore klasificirati opisa v nobeno izmed Spremljalo se bo odzive uporabnikov in po potrebi rešitev napak prosi uporabnika, naj mu podrobneje, oziroma drugače korigiralo. K osnovni rešitvi se bo kasneje razvijalo še dodatne opiše problem. Na primer: funkcionalnosti ter se jo promoviralo v okviru širše regije in celotne države. Ko bo aplikacija dodelana do ustrezne st opnje, se 51 jo bo integriralo tudi v Nacionalni inštitut za psihoterapijo, ki kakovostni karierni orientaciji za šolajočo se mladino na vseh deluje v okviru prijavitelja projekta. ravneh izobraževalnega sistema«. Delo, opravljeno v okviru tega projekta, bo sestavni del Izvajalec javnega razpisa je bil Javni štipendijski, razvojni, obsežnejše, splošno zasnovane spletne aplikacije za zdravstveno invalidski in preživninski sklad Republike Slovenije. svetovanje državljanom. 8. LITERATURA IN VIRI 7. ZAHVALA [1] Podjed, T., 2015. Prepoznavanje in obvladovanje stresa med Projekt »Spletno svetovanje študentom v stiskah« je delno študenti, zaključna naloga, Fakulteta za matematiko, financirala Evropska unija, in sicer iz Evropskega socialnega naravoslovje in informacijske tehnologije, Univerza na sklada, v okviru Operativnega programa za izvajanje evropske Primorskem. kohezijske politike v obdobju 2014-2020 kot neposredna potrditev [2] Baša, L., Jevšnik, M., Domajnko, B., 2007. Dejavniki stresa operacije - programa »Projektno delo z negospodarskim in med študenti, Obzor Zdr N 2007; 41: 77–84. neprofitnim sektorjem v lokalnem in regionalnem okolju – Študentski inovativni projekti za družbeno korist 2016 – 2018«, [3] Gorinšek, K., 2008. Stres med študenti rednega in izrednega 10. prednostne osi »Znanje, spretnosti in vseživljenjsko učenje za študija, diplomsko delo, Fakulteta za organizacijske vede, boljšo zaposljivost«, 10.1 prednostne naložbe »Izboljšanje Univerza v Mariboru. enakega dostopa do vseživljenjskega učenja za vse starostne [4] Colarič, D., Eder, K., 2008. Anksioznost in depresija pri skupine pri formalnih, neformalnih in priložnostnih oblikah študentih Univerze v Mariboru, Medicinska Fakulteta, učenja, posodobitev znanja, spretnosti in kompetenc delovne sile Univerza v Mariboru. ter spodbujanje prožnih oblik učenja, tudi s poklicnim svetovanjem in potrjevanjem pridobljenih kompetenc« in 10.1.3 [5] API.AI, https://api.ai/. specifičnega cilja »Spodbujanje prožnih oblik učenja ter podpora 52 Upgrade of AH-Model with Machine Learning Algorithms Gregor Molan Martin Molan Comtrade Digital Services Faculty of Mathematics and Physics Letališka cesta 29b Jadranska cesta 19 1000 Ljubljana 1000 Ljubljana +38640254724 +38640516254 gregor.molan@comtrade.com martin@molan.net ABSTRACT on workers’ performance. This paper tri es to determine which The goal of this research is upgrading of AH model with data mining algorithm best predicts actual work performance introduction of classification algorithms to explain connection from perceived well-being. between human well-being and efficiency. 2. CLASSIFICATION Three different algorithms (JRip, SVM, decision trees) were Classification process involves training (learning) and test sets. implemented and tested on the same data with samples of 10, Individual instance (in the training set) usually contains one 100 and 2031. JRip predicted correctly about two thirds of target value (which identifies the class – leaf in a data tree) and connections. It is not sensitive on choosing different training several additional attributes. End goal of decision tree algorithm and learning sets and accuracy was achieved at a sample of 100. is to produce a target value (leaf identifier) based only on It, however, does not offer useful explanation about well- attributes. In this sense, classification problem can be examined being’s influence on efficiency. SVM algorithm is not useful (as illustration) on example of two classes where the algorithm for this kind of data; predicted error rate is too high. is tasked with separation of classes based on available Decision tree algorithm offered the best output. It identified examples. Algorithm should produce and identify function relations between particular elements of well-being and (segregation function), based from training data, that works efficiency. Obtained results are reflection of the work place well on unseen examples (test data). Successfulness of characteristics (data was collected in real work places) and segregation algorithm is determined by its ability to perform on identify relation between elements of well-being and efficiency test data [2]. which is understandable to user. 2.1 SVM Algorithm Decision tree algorithm upgrades AH model with description of Based on statistical learning theory, Support Vector Machines relationship between QAA data and workers’ performance- is a method of supervised learning used for classification and efficiency. regression [3]. SVM creates non-overlapping partitions of data employing all attributes of particular instance (e.g. individual). Categories and Subject Descriptors In order to produce linear partitions, data is segmented in a H.1.2 [Models and Principles]: User/Machine Systems - single pass. Although similar to probabilistic approaches (they human factors, I.2.1 [Artificial Intelligence]: Applications and are based on maximum margin linear discriminants), SVM do not take into account possible co-dependencies between Expert Systems - medicine and science, J.4 [Computer attributes. Applications]: Social and Behavioral Sciences - psychology In contrast to traditional neural networks SVM algorithms do General Terms not suffer problems with generalization (SVMs have no tendency to over fit the data as a result of optimization Algorithms, Human Factors, Measurement, Reliability. measures). Consequently, SVMs seem to promise great empirical performance [4]. Structural risk minimization (SRM) Keywords [5], which is the basis of SVMs, seems to be superior to AH-model, classification algorithms, JRip, SVM, decision Empirical Risk Minimization (ERM) [6] (traditional approach trees. employed by neural networks). In contrast to ERM, which optimizes the training data, SRM optimizes the upper bound on 1. INTRODUCTION the expected risk. Consequently SRM has much greater generalization ability. Aim of this paper is to compare different AI approaches to data mining; more specifically to individual behavior modeling. This SVMs present data patterns in higher dimension than the comparison is executed on data collected from standardized original space (defined by attributes of particular entity). Questionnaire of Actual Availability (QAA). The goal of data Segregations of two categories (hyperplane separating two analysis, performed on this data, is to better understand patterns) can be achieved by nonlinear mapping to a sufficiently individual’s behavior – that is to better understand connection high dimension [4]. between different individual’s attributes and influence of availability on prediction of performance on real working 2.2 Decision tree algorithm situation. Decision trees are implementation of divide and conquer principle [7]. Differences between different implementations of Data is collected with Questionnaire of Actual Availability from decision trees are in how to approach and execute division into AH-model (psychometric data) [1]. It is comprised of subjective smaller problems. Decision trees based classifiers divide the self-assessment of workers; workers assessed their data into smaller groups (where each node of a tree represents a psychophysical condition. Besides their assessment of criteria for data segmentation) until desired homogeneity of a psychophysical condition and well-being their actual work subgroup is achieved (this subgroup is represented as a leaf performance-efficiency was also noted. containing class majority). Classification is achieved by The goal of data analysis is to determine the influence of following the classification criteria (nodes in the tree) until a perceived well-being and self-estimation of actual availability desired group is reached (represented as a leaf in the tree) [7]. 53 2.3 JRip propositional rule learning Table 3 Classification correctness with JRip method for JRip is rule-based machine learning method called n=2031 propositional rule learning and is based on association rules Classified as E1 as E2 as E3 as E4 as E5 with reduced error pruning. Data, analyzed by propositional rule learner, is split into growing and pruning set (learning and E1 402 267 3 1 0 testing set). An initial rule is formed on the growing set E2 120 873 46 2 0 (learning set) and then repeatedly optimized (simplified) by the E3 1 109 157 7 0 pruning set (testing set). For each stage of pruning (optimization) the set of rules with greatest error reduction is E4 3 11 6 21 0 chosen as the rule. Optimization ends when additional E5 0 1 0 0 2 alternations (optimizations) of rules yield additional error instead of error reduction. JRip implementation of propositional rule learner is an optimized version of Incremental Reduction The algorithm JRip is not sensitive in choosing different Error Pruning (IREP) proposed by William W. Cohen [8]. training and learning set. The first adequate level of accuracy is achieved at n=100. 3. Results All three algorithms are compared for three different 3.2 SVM Algorithm numerousness: 10, 100 and 2062. For each sample each n=10: algorithms is tested ten times to assure different random The algorithm does not work. Number of parameters is too big generated learning sets. for this sampling size and algorithm fails with error “‘cross’ 3.1 JRip propositional rule learning must not exceed sampling size”. algorithm n=100: n = 10: SVM algorithm returns similar results as JRip propositional rule learning algorithm, it returns classification correctness 100% correctly classified instances. In all ten trials, the result is table. For n=100 are only 17% correct predictions for efficiency the same: 100% correctly classified instances. 1, 2, and 3. Let denote efficiency measured as 1, it means the best Table 4 Classification correctness with SVM algorithm for efficiency, with E1, efficiency measurement value 2 with E2, n=100 and so forth to measurement value 5 denoted with E5. In table 1 are presented results of the AI model generated with JRip Classified as E1 as E2 as E3 as E4 as E5 classification algorithm for the instances of the test set. This E1 7 4 0 0 0 presents expected results of the generated model for real-world data. E2 6 8 1 0 0 Table 1 Classification correctness with JRip method for E3 0 2 2 0 0 n=10 E4 0 0 0 0 0 Classified as E1 as E2 as E3 E5 0 0 0 0 0 E1 5 0 0 E2 0 3 0 n=2031: E3 0 0 2 Only 20% of correct predictions, mostly for efficiency 2. Table 5 Classification correctness with SVM algorithm for n=100: n=100 68% correctly classified instances, 32% incorrectly classified Classified as E1 as E2 as E3 as E4 as E5 instances. The best matching is with efficiently 2. In all ten E1 145 69 1 1 0 trials, results are in the interval between 66-71% for correct classification and in the interval between 29-34% for incorrect E2 59 223 38 3 0 classification. E3 2 16 43 8 1 Table 2 Classification correctness with JRip method for E4 0 0 0 0 0 n=100 E5 0 0 0 0 0 Classified as E1 as E2 as E3 as E4 E1 19 18 0 0 The algorithm SVM is not applicable for analyzed data. Correct E2 3 49 0 0 prediction is poor without useful information for user. E3 0 10 0 0 3.3 Decision tree algorithm E4 0 1 0 0 Problem of decision tree can be complexity of a tree which is a result from decision tree algorithm. To overpass this problem, it is possible to check a visual representation of the cross- n=2031: validation results. It gives a geometric presentation of values of 72% correctly classified instances, 28% incorrectly classified complexity parameter (cp) related with tree size and expected instances. The best matching is with efficiency 2 and 1. In all relative error. The cp value is the input value for tree pruning trials the intervals are between 70-72% for correct classification algorithm that reduce tree size and increase expected and 28-30% for incorrect classification. 54 correctness of proposed decisions, but decisions has less detail argumentations. n=10: Resulting tree is not complex enough to require pruning. There is distinction in identification of psychical well-being on efficiency 1. There is also identification of psychical impact on efficiency 2. Decision tree is simple; the only influential item is psychical well-being. Figure 3 Visual representation of the cross-validation results for n=100 Figure 1 Not pruned decision tree for n=10 After the pruning of the decision tree influential elements are n=100: defined – another important influential factor is stress. Pruned decision tree is clear and offers possibility for identification of In 46% of overall estimation of well-being below 1,649 is influential elements of well-being on efficiency, mostly on 1 related mostly with efficiency 1 and some of them with and 2. efficiency 2. In 54% of overall estimation of well-being are above 1,649 and they are related mostly with efficiency 2, only a little of them with 1, 3, and 4. There is a complex decision tree presenting influence of other elements of well-being. Figure 4 Pruned decision tree for n=100 n=2031: In 36% there is psychical well-being below 1.4295 determining efficiency 1. In 49%, psychical well-being below 2,4295 predict efficiency 2. In 15% psychical well-being above 2,4295 predict efficiency 3. The same as for n=100, tree pruning is Figure 2 Not pruned decision tree for n=100 implemented to decrease relative error and to simplify decision Tree pruning is implemented to decrease relative error and to tree. simplify decision tree. The cp=0.051 is used for pruning parameter according to adequate size of tree and adequate relative error 55 predicts efficiency 1 and sometimes 2 but does not identifies influential elements of well-being. SVM algorithm is not applicable for this kind of data. The decision tree algorithm is the most suitable for implementation of QAA data from AH. It offers identification of the most influential well-being element on efficiency with distinguished limits. From the aspect of end user – psychologist or human factor specialist it offers an upgrade to the AH model. Upgrade of AH model with decision tree algorithm offers clear limits of expected values of efficiency depend on well-being perception. Due to the nature of data (QAA data was collected at workplaces with psychical workloads) obtained results reflect workplace characteristics, perception of well-being and their influence on efficiency. 5. REFERENCES [1] G. Molan and M. Molan, "Formalization of expert AH model for machine learning," Knowledge-based intelligent Figure 5 Visual representation of the cross-validation results information engineering system and allied technologies, for n=2031 pp. 110-114, 2002. Similar as for n=100, value cp=0.051 is used for pruning parameter according to adequate size of tree and adequate [2] I. Kononenko, Strojno učenje, Ljubljana: Založba FE in relative error. After pruning of decision tree psychical well- FRI, 2005. being below 1,429 is crucial in prediction of efficiently 1. Efficiency 2 is predicted with psychical well-being below [3] C. Cortes and V. Vapnik, "Support-vector networks," 2,429. If psychical well-being is above 2,429 the efficiency is 3. Machine Learning, p. 273–297, 1995. The correct prediction for 3 is 15%. [4] T. Joachims, "Text categorization with Support Vector Machines: Learning with many relevant features," European Conference on Machine Learning, pp. 137-142, 1998. [5] J. Shawe-Taylor, "Structural risk minimization over data- dependent hierarchies," IEEE Transactions on Information Theory, pp. 1926 - 1940, 1998. [6] C. M. A. D. S. Kamalika Chaudhuri, "Differentially Private Empirical Risk Minimization," Journal of Machine Learning Research, p. 1069−1109, 2011. [7] M. J. P. S. Bogumił Kamiński, "A framework for sensitivity analysis of decision trees," Central European Journal of Operations Research, pp. 1-25, 2017. Figure 6 Pruned decision tree for n=2031 The decision tree algorithms identifies relations between [8] W. W. Cohen, "Fast Effective Rule Induction," Twelfth elements of well-being and efficiency. User gets possibility for International Conference on Machine Learning, pp. 115- implementation and interpretation. 123, 1995. 4. Conclusion Comparison of three algorithms answers a question: Which algorithms is the most suitable? According to the analysis JRip is too dependent on sampling of learning sets. It correctly 56 Wireless Sensor Prototype for Industrial Harsh Environments Marko Pavlin Špela Poklukar, Gregor Papa, Franc Novak HYB d.o.o. Jozef Stefan Institute Levičnikova cesta 34 Jamova cesta 39, 8310 Šentjernej, Slovenia 1000 Ljubljana, Slovenia marko@pavlin.si {spela.poklukar, gregor.papa, franc.novak}@ijs.si ABSTRACT systems represents a great challenge. In the near future, maintenance of industrial systems will change from traditional The overall goal of the ECSEL project MANTIS is to provide a monitoring, based on the detection of malfunctions, to advanced proactive maintenance service platform architecture based on techniques that prevent malfunctions by predicting the faults. To Cyber Physical Systems. Proactive maintenance can be regarded this day, four different maintenance plans are used: reactive as upgrade of conventional preventive and predictive maintenance maintenance, preventive maintenance, predictive maintenance and and goes further by focusing on problem causes. In this way, proactive maintenance [2]. problems are settled before they actually occur. The MANTIS project comprises eleven distinct industrial partners and deals In the case of reactive maintenance, the equipment is replaced or with maintenance use cases in different environments (e.g., repaired only after it breaks. This approach has the advantage of industrial machines, vehicles, renewable energy assets). An minimizing the manpower to keep things running. Disadvantages important issue of the MANTIS project is provision of reliable are unpredictable production capacity and high overall communication. In this paper we present a solution of wireless maintenance costs. pressure sensor developed for possible replacement of the existing In preventive maintenance, maintenance tasks are performed cable-connected sensors in a harsh industrial environment. periodically, based on specific time period or the amount of working hours of machine use. The drawback is that the production is stopped during the maintenance. On the other hand, Categories and Subject Descriptors the equipment lifetime is prolonged and the probability of C.2.1 Network Architecture and Design; Wireless malfunction is reduced [3]. communication, C.3 Special-Purpose And Application-Based Predictive maintenance or condition-based maintenance, relies on Systems; Real-time and embedded systems physical measurements of the equipment conditions, such as temperature, vibration, noise, lubrication and corrosion [4]. When General Terms these measures reach a certain threshold, preventive maintenance Measurement, Performance, Reliability task is applied. Proactive maintenance benefits from the preventive and Keywords prediction methods and goes further by focusing on problem Cyber-Physical Systems, proactive maintenance, sesnsors causes. In this way, problems are settled before they actually occur. Proactive maintenance is a constant process of operation 1. INTRODUCTION improvement that starts at the early design phase and comprises the whole periodic life cycle analysis. By employing prediction Cyber-Physical Systems (CPS), which represent the next methods it relies on constant condition monitoring and evaluation generation embedded intelligent ICT systems, are characterized to avoid machine failures. Condition monitoring is achieved by large numbers of tightly integrated heterogeneous components, through extensive sensor data collection and analysis [5]. which may dynamically expand and contract with each other. Multiple sensors and actuation units that gather, process, The overall goal of the MANTIS project (http://www.mantis- exchange and use information bring together the world of project.eu/) is building a maintenance service platform that will computing and communications with the physical and biological enable proactive maintenance strategies in different environments worlds [1]. CPS components are collaborative, autonomous and (e.g., industrial machines, vehicles, renewable energy assets). For provide computing and communication, monitoring/control of this purpose, advanced data monitoring, communication and physical components/processes in various applications. The analytics is required. Since the maintenance service platform will concept of CPS is being widely applied in indistry, energy operate in different environments including harsh conditions, economy, health care, to mention just the few most prominent ensuring reliable communication is one of the major issues. In the examples. following we present a solution of wireless pressure sensor developed for possible replacement of the existing cable- While CPS are known to be difficult to analyze due to their connected sensors in a harsh industrial environment. increasing complexity, the maintenance of CPS-based industrial 57 2. WIRELESS SENSOR PROTOTYPE 2.2 Power management 2.1 Sensor Sensor is powered by small Lithium battery charged via standard USB connector, commonly used as a mobile phone charging Sensor prototype is based on HYB pressure transducer for device. The electronics is supplyed at 3,3V. The Lithium battery differential wet-wet applications. It is a new generation of has own protection circuit to avoid over-charge, over-discharge, ceramic pressure sensors made with low temperature cofired over-current and short circuit conditions which may permanently ceramic (LTCC) technology using piezo-resistive principle to damage the battery cell. The cell voltage could be over 3,3V and detect the pressure. The LTCC material is compatible with many below 3,3V during the discharge cycle. This requires Buck-Boost types of aggressive media like water, hydraulic oils, diesel and DCDC converter. Due to low power overall consumption, the others, which makes the sensor suitable for pressure synchronous buck boost converter was selected with maximum measurements in harsh environments. Special protection of the efficiency 97% at lower currents in the range of 100mA. The piezo-resistors also makes this sensor suitable for wet-wet device operates from 0.65V to 4.5V input supplying max. 200mA applications. High performance and accuracy are achieved with of current from single battery cell. No other power management the special sensor construction, which allows this sensor to be was implemented on the prototype. The power switching was used in many applications, and with its compact and convenient done with usual mechanical SPST switch. design it is very suitable for OEM users requiring use in harsh environment. The output signal from the sensor is analog and 2.3 Wireless interface digital. The HPSD 7000 analog output signal is amplified and The wireless part of the sensor is based on ESP8266 from temperature compensated from 0 to 70°C with signal conditioning Espresiff in the form factor of small WiFi Module. It is a self- electronics. contained SOC with integrated TCP/IP protocol stack with additional interfaces to give the device access to WiFi network. The module has a powerful enough on-board processing and storage capability that allows it to be integrated with the HPSD7000 sensor through its GPIOs. Its high degree of on-chip integration allows for minimal external circuitry and occupying minimal PCB area. 2.4 Software WiFi module operates as an Access Point by setting up a network of its own, allowing other devices to connect directly to the sensor. The WiFi client connects to the SOC and exchange packets via User Datagram Protocol (UDP). This represents Figure 1: HPSD 7000 pressure sensor lowest possible load to the sensor, client and allows low latency. The drawback is lack of control mechanisms when packets are not delivered. The tradeoff between data loss and latency seems to be The digital output signal is available via standard I2C optimal for such short range peer-to-peer communication. communication with default slave address 0x78 (1111000b). Pressure and temperature output signals form HPSD7000 pressure sensors are 15 bit values from the data acquisition output register. Data transfer is initiated by I2C master with the start condition. followed by 7 bit slave address (factory default is 0x78) and data direction bit R/W (for read data R/W=”1”). Slave confirms this address with acknowledge (A) bit followed by pressure data as 2 byte value, MSB first and temperature data as 2 byte value, MSB first. Master must confirm each received byte with acknowledge bit and terminate the data transfer by sending the stop condition. Master receives pressure data as a 15 bit values which can be converted to actual pressure value with pressure units in mbar using simple linear transformation using data from the datasheet for Pmin, Pmax, Dmin and Dmax, where valuses are min pressure (mbar), max pressure (mbar), max digital pressure (counts) and min digital pressure (counts), respectively. Figure 2: Communication start dialog 58 Prototype was tested using smart phone with installed “UDP terminal” application. First, the phone is connected to the access point using default password. Figure 5: Wireless pressure sensor prototype Preliminary prototype was assembled inside 3D-printed plastic housing as shown in Fig. 6. Figure 3: UDP terminal application Then UDP terminal application is started. Packets are sent from phone to sensor on port 4096 and received by phone on port 11000. When the letter “P” is sent with UDP packet on port 4096, the sensor returns pressure readout on port 11000. It is up to the client application to calculate the pressure from the readout. Figure 6: 3D model of the housing, which was printed using 3D printer and PLA plastic material 2.6 Extending the range of the wireless module The on-board antenna has potentially low range, especailly indoors. Some preliminary experiments have shown the indoor range of about 20m when sensor and client were placed in the same room. When obstacles, like human body, wall, doors or other objects were placed in the signal propagation path, the range was significantly lower. Possible improvement is additional antenna. The module used in the prototype was ESP8266-12E with PIFA (Planar Inverted »F« Antenna) integrated on the module itself. Modules with the same functionality and connector for external antenna exist. Most widely used is module ESP8266-05, which has »u.Fl« type of antenna connector. Such small connector is not suitable for direct antenna connection and requires some adapter. The adapter has u.Fl connector on one side and SMA or similar connector on the other side of the coaxial cable. The SMA connector is more suitable for integration on the sensor housing and sealed against external environment. Figure 4: Pressure sensor readouts One example is shown in Fig. 7. Advantage of such adapter is the 2.5 Hardware prototype possibility to attach external antenna for 2,4GHz or connect The prototype was developed and tested using multi-module stack remote antenna with coaxial cable between SMA connector and shown in Figure 5. antenna location. 59 Another option to improve the wireless sensor range is to use 4. ACKNOWLEDGMENTS larger patch antenna, which is placed outside the housing. This work was partially funded from the Electronic Component Advantage of this lies in easier sealing against environment (dust, Systems for European Leadership Joint Undertaking under grant moisture, water). The main disadvantage is larger dimension. agreement No 662189 (MANTIS). This Joint Undertaking receives support from the European Union’s Horizon 2020 research and innovation programme and Spain, Finland, Denmark, Belgium, Netherlands, Portugal, Italy, Austria, United Kingdom, Hungary, Slovenia, Germany. The authors also acknowledge the financial support from the Slovenian Research Agency [research core funding No. P2-0098]. 5. REFERENCES [1] Panos Antsaklis. Goals and Challenges in Cyber-Physical Systems Research Editorial of the Editor in Chief, IEEE Transactions on Automatic Control, VOL. 59, NO. 12, DECEMBER 2014, 3117 – 3119. [2] Gregor Papa, Urko Zurutuza, Roberto Uribeetxeberria. Cyber Physical System Based Proactive Collaborative Maintenance, International Conference on Smart Systems and Technologies (SST), Osijek, Croatia , 2016, DOI: 10.1109/SST.2016.7765654 [3] Swanson, L., “Linking maintenance strategies to Figure 7: External antenna options: adapter cable (above), larger performance”, International Journal of Production patch antenna (middle) and SMD solderable antenna (below). Economics, 18 April 2001, Vol. 70, Issue 3, pp 237-244. [4] Eade, R., “The importance of predictive maintenance”, Iron 3. CONCLUSIONS Age New Steel 13 (9), 1997, pp 68-72. The presented pressure sensor prototype has been developed for the proof of concept for possible replacement of wired sensors in [5] Dhillon, B.S., “Engineering Maintenance: A Modern existing industrial use case installations. Initial test is planned to Approach”, CRC Press, Boca Raton, 2002. be carried out at Philips shaver production plant. 60 Napovedovanje časovnih vrst za podporo energetski optimizaciji stavbe Dr. Domen Zupančič Robotina d.o.o. OIC Hrpelje, 38 6240 Kozina domen.zupancic@robotina.com POVZETEK Drugo poglavje opisuje energijsko shemo sistema ter na kratko opiše energijske komponente sistema. Tretje poglavje opisuje V naslednjem sestavku predstavljamo komponento sistema za optimizacijo upravljanje energije, ki napoveduje proizvodnjo in postopek modeliranja sončne elektrarne in izvedbo napovedi ob porabo električne energije. V ta namen je bil izveden predpostavki, da je znana napoved sončnega obsevanja. Četrto demonstracijski projekt za sistem večstanovanjske zgradbe, ki (i) je poglavje opisuje postopek za izvedbo napovednega modela za priključen na električno omrežje in sistem daljinskega ogrevanja, porabo električne in toplotne energije ob predpostavki, da je (ii) je priključen na fotovoltaično elektrarno, (iii) vsebuje toplotno poznana zgodovina porabe za obdobje zadnjih nekaj dni in napoved črpalko z zalogovnikom za ogrevanje stanovanj in (iv) vsebuje zunanje temperature. Peto poglavje je namenjeno prikazu baterije. Namen celotnega projekta je demonstracija sistema za rezultatov in vrednotenju izvedenih modelov, v zadnjem poglavju optimizirano usmerjanje energetskih tokov, kjer želimo doseči ali je podan zaključek. maksimirano lastno rabe energije ali minimizacijo stroškov. 2. ENERGIJSKA SHEMA SISTEMA Medtem ko je izvajanje prve strategije razmeroma enostavno, pa je Slika 1 prikazuje energijsko shemo sistema. Zelena barva za drugo potrebno planiranje delovanja sistema v prihodnosti na osnovi predvidevanj. V ta namen smo razvili komponento za predstavlja električne 𝑃, rdeča pa toplotne 𝑄̇ moči. Puščice napovedovanje proizvodnje električne in porabo električne ter označujejo smer tokov, pri čimer so dejanske smeri 𝑃𝐺, 𝑃𝐵, 𝑄̇𝑇 toplotne energije. lahko tudi nasprotne. Kot vidimo iz slike, je stanovanjska zgradba porabnik tako električne, kot tudi toplotne energije. Sončna Ključne besede elektrarna predstavlja vir energije. Električno omrežje lahko predstavlja vir lahko pa tudi porabnik energije. Sistem daljinskega Poraba energije, proizvodnja energije, toplota, elektrika, ogrevanja predstavlja vir energije. Na koncu so še baterije kot napovedovanje, strojno učenje, sončna elektrarna hranilniki električne energije in zalogovnik kot hranilnik toplote. 1. UVOD Napredno upravljanje energije v stanovanjskih objektih z možnostjo shranjevanja energije pomeni vključevanje posebnih metod, ki omogočajo izvedbo kratkoročnega in srednjeročnega planiranja razporejanje energije. Planiranje je pomembno predvsem zaradi nadzora upravljanja energijskih tokov, katere je mogoče upravljati skladno s cenovno ali energetsko učinkovitimi strategijami. V ta namen smo za demonstracijski projekt [1] izdelali modul za napovedovanje determinističnih in stohastičnih časovnih vrst, ki predstavljajo proizvodnjo in porabo električne energije, ter porabo toplotne energije. Pri zasnovi modela napovedovanja smo si pomagali s simulacijo delovanja sistema na osnovi podatkov, pridobljenih na primerljivem objektu. Namen izvedbe je priprava napovedanih vrednosti, ki predstavljajo vhod v fazo optimizacije [2], ki pa ni tema tega prispevka. Sorodne raziskave opisujejo različne metode napovedovanja časovnih porabe energije, na primer primerjava različnih implementacij z nevronskimi mrežami [3] ali kratkoročno napovedovanje z poznavanjem vremenskih podatkov [4]. Drugi so uporabili različne podporne podatke, kot na primer podatki o delovanju električnih naprav [5] in aktivnosti uporabnikov, medtem ko so v [6] potrdili, da se napovedljivost porabe zelo razlikuje glede Slika 1: Energijski tokovi v sistemu na stanovanja in uporabnike in da je kratkoročno napovedovanje na urni ravni bolj natančno z napovedovanjem porabe posameznih V povezavi s planiranjem nas zanima predvsem, kakšna bo poraba naprav, medtem ko je napovedovanje za nekaj dni v naprej energije stanovanj in kakšna bo proizvodnja elektrike v bližnji natančnejše na agregiranem sistemu celotnega objekta. prihodnosti. Primer: če želimo zagotoviti maksimalno samozadostnost zgradbe, potem je potrebno napolniti baterije in 61 zalogovnik ravno toliko, da bo zadosti energije za potrebe zgradbe, obsevanja na enoto horizontalne površine, 𝑃𝑖𝑛𝑠𝑡 pa je nazivna moč čim več ostale energije pa želimo prodati. Poraba stanovanjske elektrarne. Če označimo koeficienta učinkovitosti obeh strani zgradbe je stohastičnega tipa, za katerega nimamo fizikalne elektrarn kot 𝑘𝐸𝐻𝑑,15 in 𝑘𝐸𝐻𝑑,30, pri izračunu pa upoštevamo formulacije, medtem ko je proizvodnja elektrike odvisna od samih vrednosti, izračunane z aplikacijo PVGIZ za 𝐸𝑑,𝑡𝑖𝑙𝑡,15 , 𝐸𝑑,𝑡𝑖𝑙𝑡,30, dimenzij elektrarne in od sočnega obsevanja. 𝐻𝑑,ℎ𝑜𝑟 ter nazivni moči posameznih strani 𝑃𝑖𝑛𝑠𝑡,15 in 𝑃𝑖𝑛𝑠𝑡,30 kot V naslednjih dveh poglavjih bosta opisana postopka za 15 kWp in 30 kWp, potem je skupna učinkovitost vsota, podana z napovedovanje porabe energije in proizvodnje električne energije. Enačbo (2). 𝑘𝐸𝐻𝑑 = 𝑘𝐸𝐻𝑑,15 + 𝑘𝐸𝐻𝑑,30 (2) 3. MODELIRANJE SONČNE Napoved moči proizvodnje glede na napovedano sončno obsevanje ELEKTRARNE na horizontalno ploskev 𝐼𝑠,𝑛 napovemo v skladu z Enačbo (3). Model za napovedovanje proizvodnje energije smo izdelali na 𝑃 osnovi nazivnih parametrov sončne elektrarne in podatkov o 𝑛 = 𝑘𝐸𝐻𝑑 × 𝐼𝑠,𝑛 (3) geolokaciji elektrarne. Pri razvoju modela napovedovanja smo Napoved sončnega obsevanja 𝐼𝑠,𝑛 [𝑘𝑊ℎ] je za dotično lokacijo privzeli naslednje dinamične podatke, ki jih bo model potreboval 𝑚2 mogoče pridobiti preko vremenskega servisa, na primer za izvedbo napovedi: Meteomedia. 1. napoved sončnega obsevanja na horizontalno površino 𝐼𝑠 [𝑊] 4. MODELIRANJE PORABE 2. povprečna dnevna proizvodnja energije sistema 𝐸 ELEKTRIČNE IN TOPLOTNE ENERGIJE 𝑑 [𝑘𝑊ℎ] 3. povprečna dnevna količina sončnega obsevanja na enoto Napoved porabe električne in toplotne energije v stanovanjskih površine, ki jo prejme sistem 𝐻 objektih predstavlja problem, ki ga ni mogoče enostavno zapisati 𝑑 [𝑘𝑊ℎ/𝑚2] kot enoznačen matematični zapis. Poraba električne in toplotne Točko 1 - napoved sončnega obsevanja na horizontalno površino je energije je odvisna od človeških zahtev, ki pa se med posamezniki mogoče pridobivati preko vremenskega servisa, na primer razlikujejo. Meteomedia1 na urni ravni. Podatke za točki 2 in 3 pa pridobimo za celo leto s pomočjo spletne aplikacije PVGIZ2, kjer vnesemo Za potrebe optimizacije planiranja shranjevanja energije v baterijah geolokacijo elektrarne in podatke o azimutu ter naklonu strehe. in zalogovniku smo problem napovedovanja porabe energije zasnovali na osnovi stohastičnega modela, generiranega s pomočjo Shema sončne elektrarne je prikazana na Sliki 2. Desna stran slike metod umetne inteligence. Postopek izvedbe napovednega modela kaže satelitski posnetek strehe. Elektrarna je razdeljena na vzhodno in izvajanja napovedi je prikazan na Sliki 3. stran z nazivno močjo 15 kWp in zahodno stran z nazivno močjo 30 kWp. Leva stran slike prikazuje azimut vzhodnega in zahodnega dela elektrarne ter naklon strehe. Slika 2: Koti sončne elektrarne Koeficient učinkovitosti posamezne strani elektrarne izračunamo kot razmerje med sončnim obsevanjem na elektrarno in prejeto energijo, pomnoženo z nazivno močjo elektrarne Enačba (1). 𝐸 𝑘 𝑑,𝑡𝑖𝑙𝑡 𝐸𝐻𝑑,𝑖𝑛𝑠𝑡 = × 𝑃 𝐻 𝑖𝑛𝑠𝑡, (1) 𝑑,ℎ𝑜𝑟 kjer je 𝐸 Slika 3: Proces stohastičnega napovedovanja časovnih vrst 𝑑,𝑡𝑖𝑙𝑡 povprečna dnevna proizvodnja energije za elektrarno z naklonom tilt, 𝐻𝑑,ℎ𝑜𝑟 je povprečna dnevna količina sončnega 1 Meteomedia - spletna storitev za napoved vremenskih podatkov 2 PVGIZ - Spletna aplikacija za izračun proizvodnje električne za geografsko lokacijo (http://wetterstationen.meteomedia.de/), energije (http://re.jrc.ec.europa.eu/pvgis/apps4/pvest.php), zadnjič pridobljeno: 20.9.2017 zadnjič pridobljeno: 20.9.2017 62 4.1 Priprava učnih podatkov komunikaciji in v teh obdobjih ni bilo pridobljenih podatkov, ali pa so bili napačni. Med delovanjem sistema je potrebno z logiranjem pridobivati Slika 5 prikazuje primer urnih napovedi za obdobje zgodovino z datumom in časom opremljenih podatkov o porabi med 1.9.2016 in 5.9.2016. toplotne in električne energije ter podatke o zunanji temperaturi ter Rezultati kažejo na zadovoljivo napovedovanje, pri čimer je jih shranjevati v bazo podatkov. Ker so podatki o električni in potrebno upoštevati geografsko razliko med podano vrednostjo toplotni porabi pridobljeni iz električnih števcev in kalorimetrov v sončnega obsevanja za lokacijo vremenske postaje in lokacijo obliki energije, jih je potrebno filtrirati, kar obsega: elektrarne. Na proizvodnjo vplivajo tudi dimniki in drevo, ki  mečejo senco na elektrarno, predvsem pa je sončna elektrarna izbor zapisov z enakimi časovnimi koraki,  močno občutljiva na stopnjo oblačnosti. izračun diferenc za prevod energije v moč,  skaliranje za zapis v kW, 300  brisanje ekstremnih vrednosti in drugih nepravih Napoved podatkov. 250 Postopek priprave učnih podatkov Meritev obsega tudi dodajanje atributov, 200 ki določijo dan v tednu, mesec, tip dneva (delavnik / prost dan), čas h] Napaka v dnevu (dopoldan / popoldan). [kW 150 4.2 Učenje, napovedovanje in vrednotenje rgija 100 Učenje napovednega modela smo izvedli s pomočjo m ne odula za e napovedovanje časovnih vrst v sklopu programskega okolja Weka 50 vna [3]. Pri tem smo preizkusili delovanje naslednjih algoritmov: neD 0  metoda podpornih vektorjev  k-ti najbližji sosed z vrednostmi k: 3, 5, 7 in 10 -50  linearna regresija  Gaussovi procesi -100  maj jun jul avg sep okt nov dec jan metoda naključnih gozdov Za potrebe vrednotenja delovanja algoritmov smo izvedli Slika 4: Prikaz vsote dnevnih vsot napovedi, dnevnih vsot simulacijski postopek, ki zajame celotno zgodovino meritev in meritev proizvodnje in napake (maj 2016 – januar 2017) izvede postopek priprave učno množico podatkov. Za vsakega od zgoraj navedenih algoritmov strojnega učenja v prvem koraku izbere podatkovne instance za prvih 14 dni in testne podatke na 30 ] naslednji dan, izvede strojno učenje, izdela napovedi in jih shrani. h Potem za vsak naslednji dan poveča učno množico za en dan, 20 izvede učenje, izdela napovedi ter jih doda dotedanjim na [kW povedim. Postopek tako iterativno ponavlja, dokler ne naredi napovedi za 10 zadnji dan podatkov. ergija en Po končanem postopku smo vse napovedane vrednosti porabe 0 a primerjali z merjenimi vrednostmi in izračunali še srednjo rnU absolutno napako, srednjo relativno napako ter koren srednje -10 kvadratne napake. Na ta način smo pridobili informacijo o čet pet sob ned pon tor uspešnosti posameznih algoritmov napovedovanja in jo uporabili Napoved Meritev Napaka kot podlago za izbor primernega algoritma za delovanje na realnem Slika 5: Prikaz urnih napovedi proizvodnje, meritev objektu. Rezultati bodo predstavljeni v naslednjem poglavju. proizvodnje ter napako za obdobje med 1.9. in 5.9.2016 5. Rezultati in diskusija Sledeče poglavje predstavlja rezultate vrednotenja modela za Napake dnevnih vsot napovedi so razmeroma nizke in ker sistem napoved proizvodnje sončne elektrarne. Za tem sledi primerjava vsebuje baterije, se urne napake kompenzirajo tekom dneva in algoritmov za napovedovanje porabe električne in toplotne posledično natančnost urnih napovedi ne pomeni zadovoljivo energije. Nazadnje bo na kratko predstavljeno še delovanje sistema delovanje, ki bo potrebno pri kasnejši izvedbi planiranja nakupa, na realnem objektu, kjer smo simulator priredili za delovanje v prodaje in hranjenja električne energije. produkcijsko verzijo. 5.2 Vrednotenje in primerjava algoritmov za 5.1 Vrednotenje modela sončne elektrarne napoved porabe energije Model sončne elektrarne smo vrednotili tako, da smo primerjali proizvodnjo, napovedano z našim modelom na os Primerjava algoritmov je prikazana v Tabeli 1. Posamezni stolpci novi vsebujejo različne izvedbe alg napovedanega sončnega obsevanja in izmerjeno proizvodnjo oritmov. Za vsako od množic napovedi, ki so bile izvedene z različnimi algoritmi smo izračunali sončne energije. Slika 4 prikazuje rezultate napovedovanja proizvodnje sončne elektrarne. Napovedane in merjene vrednosti naslednje tipe napak: koren srednje kvadratne napake, srednjo kvadratno napako, srednjo napako, srednjo absolutno in srednjo so prikazane kot dnevne integrale moči v času – torej dnevne vsote odstotno absolutno napako ter standardni odklon napake. Stolpci so energije. Napaka je prikazana kot razlika med napovedano in obarvani tako, da zelena barva prikazuje boljše rezultate, rdeča pa dejansko proizvedeno energijo. Prazni prostori na grafu pomenijo slabše. izpad meritev zaradi vzdrževalna ali druga dela, ali zaradi napak v 63 Tabela 1: Primerjava algoritmov za napovedovanje električne in toplotne energije Metoda podpornih Linearna Naključni Gaussovi vektorjev regresija kNN, n=3 kNN, n=4 kNN, n=5 kNN, n=6 gozdovi procesi koren sr. kvadratne napake 1.838 1.883 1.978 1.958 1.929 1.919 1.685 1.766 sr. kvadratna napaka 3.379 3.545 3.914 3.835 3.723 3.681 2.841 3.119 sr. napaka 0.060 -0.154 0.043 0.012 -0.001 -0.001 -0.124 -0.112 sr. absolutna napaka 1.403 1.430 1.507 1.493 1.473 1.465 1.282 1.350 standardni odklon napake 1.837 1.877 1.978 1.958 1.929 1.919 1.681 1.763 sr. odstotna abs. napaka 28.80% 31.39% 31.35% 31.50% 31.36% 31.43% 28.15% 28.97% Najboljši algoritem je označen z poudarjenimi črkami v tabeli. 6. ZAKLJUČEK Podatki, ki smo jih uporabili za vrednotenje algoritmov, so bili pridobljeni v obdobju med 26.6.2016 in 24.10.2016. V prispevku je opisan postopek modeliranja in uporabe napovednih Že prvi pogled na tabelo pokaže, da najboljše rezultate izkazuje modelov za proizvodnjo in porabo električne in porabo toplotne algoritem naključnih gozdov po skoraj vseh kriterijih. Ta algoritem energije. Modeliranje proizvodnja električne energije je izvedeno se v primerjavi z drugimi, predvsem 5. in 6. najbližjim sosedom, na osnovi matematičnega modela preko logičnih fizikalnih relacij izkaže kot slab le v kategoriji srednje napake. Dober rezultat kaže saj je odvisna od sončnega obsevanja. Poraba električne in toplotne energije pa je poleg stanja vremena – zunanje temperature odvisna tudi algoritem Gaussovi procesi, kar je ugodno zaradi hitrega učenja modela v primerjavi z naključnimi gozdovi. tudi od uporabnikov stanovanja, kar je razlog za izvedbo stohastičnih napovednih modelov, ki uporabljajo metode strojnega Kljub temu, da je kategorija srednje napake pomembna, saj se učenja. V prispevku so poleg modeliranja prikazane tudi metode za podobno kot pri proizvodnji tudi poraba energije tekom časa izbor algoritma strojnega učenja. Rezultati prikazujejo uspešno kompenzira zaradi hranilnikov energije, smo se za implementacijo izvedbo na realnem objektu. na realnem modelu odločili za implementacijo algoritma naključnih gozdov na realni sistem. Slika 6 prikazuje urne napovedi 7. ZAHVALA porabe električne energije za obdobje med 10.8. in 14.8.2016. Slika Izvedbo raziskav in razvoja so omogočili podjetje Hitachi Chemical, ltd. v sklopu projekta »Self-consumption HEMS with 10 PV and battery demonstration project«, projekt EcoSmart in ]h Agencija Republike Slovenije za Raziskovalno dejavnost v sklopu programa »Spodbujanje zaposlovanja mladih doktorjev znanosti« [kW 5 in projekta ARRS-MDR-ZP-2017-02. ergija 8. REFERENCES en 0 [1] Hitachi Chemical Co., Ltd. , http://www.hitachi- a chem.co.jp/english/information/2016/n_160531f2a.html, rnU Spletna stran, zadnji dostop: 25.9. 2017 -5 [2] Kurosawa Y., Establishing an “Energy Self-Consumption sre čet pet sob ned pon Model” for Local Production and Consumption of Solar Napoved Meritev Napaka Power in German y, NEDO - New Energy and Industrial Slika 6: Prikaz urnih napovedi porabe električne energije Technology Development Organization., za obdobje med 10.8. in 14.8.2016 http://www.nedo.go.jp/english/news/AA5en_100067.html, Spletna stran, zadnji dostop: 25.9. 2017 [3] H. R. Khosravani., in drugi, A Comparison of Energy 20 Consumption Prediction Models Based on Neural Networks ]h of a Bioclimatic Building, Energies 2016, 9, 57; 15 doi:10.3390/en9010057 [kW 10 [4] M. Becalli in drugi(2008), Short-term prediction of 5 household electricity consumption: Assessing weather ergija sensitivity in a Mediterranean area, Renewable and en 0 Sustainable Energy Reviews 12(8):2040-2065 a rn -5 [5] Gajowniczek K, Ząbkowski T (2017) Electricity forecasting U on the individual household level enhanced based on activity -10 patterns. PLoS ONE 12(4): e0174098. pon tor sre čet pet sob https://doi.org/10.1371/journal.pone.0174098 Napoved Meritev Napaka [6] D. Lachut, N. Banerjee, S. Rollins, Predictability of energy Slika 7: Prikaz urnih napovedi toplotne energije za obdobje use in homes, Green Computing Conference (IGCC) 2014 med 21.11. in 25.11.2016 International. , pp. 1-10, 2014. [7] Eibe Frank, Mark A. Hall, and Ian H. Witten (2016). The WEKA Workbench. Online Appendix for "Data Mining: Practical Machine Learning Tools and Techniques", Morgan Kaufmann, Fourth Edition, 2016 64 Avtomatizacija in digitalizacija vrta Luka Colarič Fakulteta za matematiko in računalništvo Večna pot 113, Ljubljana, Slovenija luka.colaric@gmail.com POVZETEK 2. PARAMETRI ZA NADZOR Prispevek opisuje razvoj sistema digitalizacije in avtomatizacije AGRIKULTURNIH POVRŠIN agrikulturnih oziroma vrtnih površin. Sistem je sestavljen iz treh glavnih delov. Oddaljeni moduli imajo priključene senzorje za Pri izbiri parametrov, ki smo jih želeli nadzorovati na temperaturo in vlago zraka, vlažnost zemlje ter osvetljenost agrikulturnih površinah, smo se odločali na podlagi dveh okolice. Oddaljeni moduli imajo tudi funkcijo avtomatičnega kriterijev pomembnost [1]: za rast in razvoj rastlin in obstoju senzorjev za merjenje tega parametra. Npr. parameter pH (kislost zalivanja. Glavni modul sprejema podatke iz oddaljenih modulov in jih posreduje strežniku. Na strežniku je podatkovna baza za ali bazičnost zemlje) je pomemben za zdrav razvoj rastline, a senzor, ki bi ga meril, ni preprost za uporabo. Potrebno ga je shranjevanje podatkov in spletna stran za pregled teh podatkov. namreč vsake toliko časa kalibrirati z različnimi raztopinami \cite{ph}, nezanemarljiva pa je tudi izrazito višja cena v Splošni izrazi primerjavi z ostalimi senzorji. Izbrani parametri, ki so bili Agrokultura, meritve, pošiljanje podatkov, testiranje. smiselni pri našem projektu, so: - svetloba; spada med najpomembnejše dejavnike za rast in zdrav razvoj vsake rastline. Rastlina s pomočjo Ključne besede svetlobe ustvarja fotosintezo, biokemijski proces, ki ji Digitalizacija, avtomatizacija, senzorika, baza podatkov. omogoča, da le-to pretvori v energijo za lastno rast - temperatura; prav tako eden izmed nepogrešljivih 1. UVOD dejavnikov za rastlino. Večina rastlin uspeva med 0°C in 50°C. Optimalna temperatura skozi dan, skozi noč in V dobi hitro spreminjajočega sveta se spreminjajo tudi najbolj najvišja še sprejemljiva temperatura osnovne človekove dejavnosti, kot je pridelava sadja in zelenjave. za rastlino pa so različne za posamezne rastline S tem povezani postopki se poskušajo avtomatizirati, okoljski parametri pa zajeti z računalniki oziroma digitalizirati. Podatki - vlaga zraka; je količina razpršene vode v zraku, ta digitalizacije se selijo na splet. Tam se spremljajo in iz njih količina pa je povezana s temperaturo zraka. Toplejši razbirajo razmere zraka, sonca in zemlje, kjer se vzgajata sadje in zrak ima zmožnost zadržati več vode kot hladnejši zelenjava. Iz teh podatkov je razvidno, kako se razmere - vlaga zemlje; je količina vode v zemlji, prav tako spreminjajo preko dneva, skozi letne čase in kako se skozi leta pomembna za obstoj rastline. V primeru, da je vlaga spreminjajo podnebne razmere. Na začetku je bil namen zemlje visoka dlje časa, obstaja nevarnost, da rastlina avtomatizacije poljedelskih in vrtnih površin zmanjšati in zgnije. Če je vlage premalo, pa rastlina ne more vsrkati poenostaviti delo pridelovalcev in oskrbnikov. Sistemi so bili dovolj vode in odmre zaradi pomanjkanja preprosti in nepovezani, namenjeni so bili predvsem intervalnemu zalivanju površin, kar pa ni bilo vedno usklajeno s pomembnimi 3. DELOVANJE SISTEMA zunanjimi dejavniki, kot sta trenutna namočenost zemlje in osončenost zemljišča. Vpliv enega in drugega je lahko za rastline 3.1 Oddaljeni moduli uničujoč, v prvem primeru povzroči gnitje, v drugem učinek Oddaljeni moduli delujejo na Arduino mikrokrmilniku in so povečevalnega stekla na moč sončnih žarkov. Kasneje so se namenjeni temu, da se postavijo na vrtne površine in brezžično sistemi izboljšali in začeli upoštevati zunanje dejavnike ter se jim glavnemu modulu pošiljajo podatke, ki jih beležijo s pomočjo prilagajati. Nadzorovali so izvajanje različnih potrebnih aktivnosti vgrajenih senzorjev. Prav tako omogočajo samodejno zalivanje - poleg zalivanja so ogrevali tople grede, optimizirali vlago v vrtov z vgrajenim relejem, ki preko ventila sproža dovod vode do zraku, zastirali sonce s premičnimi strehami, … namakalnih sistemov. Uporabljeni senzorji na oddaljenih modulih: Agrikultura je največja in za človekov obstoj ena izmed najbolj - senzor FC-28; za merjenje vlage zemlje pomembnih panog. Pod besedo agrikulturo spadata kmetijstvo in - senzor DHT-11; za merjenje vlage zraka in temperature poljedelstvo. Z namenom kvalitetne avtomatizacije in okolice digitalizacije smo izdelali sistem, ki je prilagodljiv glede na velikost in število agrikulturnih površin, prav tako pa sistem - foto upor; za merjenje osvetljenosti uporabniku na prijazen način ponuja podatke, ki mu omogočajo - modul nRF24l01; za brezžično komunikacijo z glavnim kvalitetno vzgojo zdravega sadja in zelenjave. modulom 65 3.2 Glavni modul Glavni modul deluje na mini računalniku Raspberry Pi [2], na katerem teče programska koda, napisana v programskem jeziku Python. Ta sprejema podatke iz oddaljenih modulov, nato pa jih posreduje naprej na strežnik. Podatke iz oddaljenih modulov pridobiva s pomočjo serijske komunikacije, saj je preko USB vodila povezan na Arduin-a, ki s pomočjo modula nRF24l01 [3] komunicira z oddaljenimi moduli. Glavni modul uporablja: - senzor DHT-11; za merjenje vlage zraka in temperature okolice - senzor BMP085; za merjenje zračnega pritiska 3.3 Strežnik in spletna stran Tretji del predstavlja strežnik, na katerem je podatkovna baza, ki shranjuje podatke, katere dobiva iz glavnega modula. Na strežniku se nahaja tudi spletna stran, ki končnemu uporabniku na preprost način prikazuje relevantne podatke iz njegovih vrtov. Ti podatki, ki jih lahko spremlja od kjerkoli, mu omogočajo pregled nad trenutnim dogajanjem na njegovem vrtu. Ti uporabniku pomagajo Slika 2: Posplošen prikaz algoritma za zalivanje pri odločitvah za možne izboljšave na vrtu. Poenostavljen prikaz vidimo na sliki 1. 4. STREŽNIK IN SPLETNA STRAN 4.1 Strežnik Glavna funkcija našega strežnika je pridobivanje podatkov iz oddaljenih modulov, jih shraniti v podatkovno tabelo in nato posredovati spletni strani, ki jih prikaže. Naš strežnik temelji na ASP.NET [4] platformi, programski jezik, ki smo ga uporabljali je C#, za programersko okolje smo uporabili Microsoft-ov program Visual Studio. 4.2 Spletna stran Spletna stran je namenjena prezentaciji podatkov iz vrtov uporabniku. Želeli smo, da je spletna stran: - preprosta za uporabo - pregledna in nenasičena - daje uporabniku le pomembne podatke - uporabna tudi na mobilnih napravah Kot že omenjeno, naša spletna stran pridobiva podatke iz modulov preko GET ukaza iz strežnika. Ta ukaz se sproži ob zagonu spletne strani. Od podatkov iz glavnega modula spletna Slika 1: Poenostavljen prikaz delovanja stran uporabi le zadnji vnos v podatkovno tabelo. To dobimo s pomočjo SQL ukaza TOP in jih sortiramo po datumu. Ti podatki so le informativne narave. Iz oddaljenih modulov se uporabi za grafični prikaz le nekaj zadnjih podatkov (ob zagonu spletne 3.4 Samodejno zalivanje strani zadnjih 500 podatkov), število le-teh lahko določimo s Naš oddaljen modul ima tudi možnost namakanja zemlje, ki se pomočjo znaka + in - ob grafu ter s tem spreminjamo časovni zgodi pod določenimi pogoji, ki so prikazani na sliki 2. Sistem razpon na grafu. Preberejo se še podatki kot so koordinate in ime deluje s pomočjo solenoidnega ventila. To je vodni ventil, ki ga vrta. Prikaz grafa enega izmed vrtov, med zalivanjem, na sliki 3. lahko po želji vključimo ali izključimo z dovodom električnega toka. Ta ventil deluje na 12V napetost, zato je bila potrebna baterija s takšno napetostjo. Dovod napetosti, da se ventil odpre, se regulira s 5V relejem, ki smo ga lahko upravljali direktno iz Arduina-a. Na začetek in konec ventila smo namestili nastavek za povezavo na navadno vrtno cev, katere en konec gre na dovod vode, drugi konec cevi pa se priključi na nastavek za zalivanje, npr. škropilnik, porozna cev za namakanje. Kot alternativa solenoidnega ventila za manjše vrtove oziroma lončnice je priključitev na majhno vodno črpalko. 66 Slika 3: Graf enega izmed vrtov Spletna stran je narejena v AngularJS, ki je odprtokodno ogrodje - skupina A - namakanje je bilo samodejno s pomočjo za izdelavo dinamičnih spletnih strani. V našem projektu smo našega algoritma zalivanja uporabili precej knjižnic in en API za pridobivanje podatkov. API - skupina B - ročno zalivanje glede na podatke, ki jih je je namenjen pridobivanju trenutnih vremenskih razmer. API uporabnik izvedel iz naše spletne strani deluje tako, da spletni naslov spremenimo na naše koordinate vrta. Ta pa nam nazaj v obliki XML sporoča trenutne razmere. V - skupina C - ročno zalivanje vsak tretji dan ob isti uri našem projektu uporabimo trenutno temperaturo, ki jo lahko - skupina D - ročno zalivanje "po občutku" primerjamo z izmerjeno s strani naših modulov in trenuten opis vremenskih razmer, npr. oblačno, megleno, ... 6. REZULTATI Rezultate testiranja smo analizirali po dveh mesecih, tako da je 4.3 Elektronsko obveščanje uporabnika vsak lonček imel dovolj časa za rast in razvoj. Rezultati pa so V naš projekt smo dodali še elektronsko obveščanje skrbnika vrta sledeči: v primeru, ko vlažnost zemlje pade pod določeno kritično - skupina A; Vse štiri rastline so razvile podobno vrednost. To se lahko zgodi v treh primerih: velikost, kar nam sporoča, da je bila voda enakomerno - možnost mehanske napake - ni dovoda vode, ker se le-ta razporejena po celotnem lončku. Vse rastline so zdrave, nekje v sistemu prekine kar nam pove, da so bili pogoji za rast dobri. Zemlja ni - nepovezanost oddaljenega modula z ventilom, ki bila nikoli preveč razmočena in nikoli presuha. S to sprošča dovod vode ali pa se ventil pokvari skupino smo bili zadovoljni, saj nam je v praksi pokazala, da je naš algoritem zalivanja pravilno in - izsušitev zemlje, kadar ni vode v vodovodnem sistemu dobro deloval. Sistem obveščanja se sproži, kadar senzor pokaže vrednost med - skupina B; Razvile so se vse štiri rastline, vendar so bile 50 in 10. To pomeni, da je zemlja suha. Ne more pa se sprožiti, če dokaj različne glede višine. Ena izmed njih je izstopala, pade vrednost pod 10, ker to z veliko verjetnostjo pomeni, da je ostale so bile precej nižje, a še vedno nekoliko manjše bil modul iztaknjen iz zemlje. Ko dobimo podatek iz modula, da kot rastline v skupini A. Večina rastlin iz te skupine je je na vrtu zemlja suha, se na elektronski naslov pošlje sporočilo, bila zdravega izgleda, le najvišja je, zaradi hitre rasti, da je na točno določenem vrtu (npr. tomato garden) močno padla imela tanko in krhko steblo. Predvidevamo, da je do vlažnost zemlje. To elektronsko sporočilo se pošilja na največ opisane razlike prišlo iz dveh razlogov. Po eni strani triurne intervale, da ne poplavimo elektronskega nabiralnika. ročno zalivanje lahko privede do neenakomerne Elektronski naslov je lahko drugačen za vsak vrt posebej, kar je navlaženosti zemlje. Tako smo včasih zalili vse rastline uporabno, če imajo vrtovi različne oskrbovalce, saj s tem vsakemu enako, včasih pa je bila kakšna bolj, druga manj zalita. posebej pošlje sporočilo samo za njegov vrt. Po drugi strani pa je ročno zalivanje manj optimalno glede na dejansko vlažnost zemlje, kot če je avtomatsko. Vendar smo tudi s temi rezultati bili zadovoljni, saj so 5. TESTIRANJE nam sporočali, da sistem dobro deluje tudi, če nas samo Projekt smo testirali na čiliju, vrsti paprike, ki je občutljiva opozarja, kdaj moramo zalivati rastline. rastlina in za zdrav razvoj zahteva ugodne pogoje. Odločili smo se za vrsto Carolina Reaper, ki je v času izdelave tega projekta - skupina C; Razvile so se vse štiri rastline, vendar precej različno. Ena rastlin najmočnejši čili na svetu. Semena smo sprva posadili v skupen a je na robu preživetja, dve rasteta počasi. Le ena od štirih rastlin je dokaj primerne rasti. lonček, ko pa so rastlinice razvile po dva para listkov, smo jih po Samo dve od štirih rastlin izgledata zdravi. Tretja ima štiri do pet razdelili v štiri različne testne lončke. Vsi lončki so rumene listke, četrta je izrazito zakrnela. Vzrok izraziti imeli isti vir svetlobe in čas osvetlitve - žarnica, ki je gorela 12 ur na dan, s čimer smo simulirali idealno osvetlitev, ki jo ta vrsta raznolikosti v izgledu rastlin je verjetno le močno nihanje namočenosti zemlje, ki se je lahko v treh dneh rastline potrebuje. Lončki so se med seboj razlikovali le glede na režim namakanja. popolnoma izsušila ali pa je bila še vlažna, ko je bil čas Poimenovali smo jih s črkami od A do D: ponovnega zalivanja. Torej rastline v bistvu nikoli niso bile primerno zalite. Ta test nam je nazorno pokazal, da 67 avtomatično časovno namakanje slabo zadosti potrebam 7. ZAKLJUČEK rastline po vodi. V našem projektu smo si zadali nalogo izdelati sistem, ki bi - skupina D; V tej testni skupini sta od posajenih petih omogočal digitalizacijo in avtomatizacijo vrtnih površin. Sistem rastlin dve propadli. Preostale tri so se v višino razvile deluje na principu glavnega modula, ki sprejema podatke različno. Preostale rastline izgledajo zdrave, vendar so v oddaljenih modulov. Oddaljeni moduli so postavljeni po vrtovih, povprečju precej nižje kot v testih A in B. Le ena od iz katerih želimo pridobivati podatke, pomembne za zdravo rast in rastlin je dosegla podobno velikost kot v prvih dveh razvoj rastlin. Sistem omogoča tudi avtomatično zalivanje vrtov. skupinah. Vzrokov za takšen rezultat je več - Zalivanje se prilagaja trenutnim razmeram na vrtu in dejanski neenakomerno zalivanje, različni časovni intervali, potrebi rastlin po vodi, s principom logike čakanja pa se voda včasih je oseba na rastline pozabila, včasih pa jih je enakomerno razporedi po zemlji pred ponovnim namakanjem. premočno zalila. S tem testom smo dobro ponazorili Sistem uporabniku nudi tudi spletno stran, preko katere lahko običajno zalivanje rastlin. Tudi rezultati so temu spremlja svoje vrtove, prav tako omogoča obveščanje preko primerni. spletne pošte v primeru, da se zemlja na vrtu preveč izsuši. Na sliki 4 so prikazani lončki po dveh mesecih rasti. V splošnem pa smo z delovanjem projekta zadovoljni. Sistem deluje podobno, kot smo si ga že na začetku zamislili. Najbolj nas je prepričalo testiranje našega sistema, ki smo ga implementirali v razvoj sadik čilija. Rezultati so zanesljivo govorili v prid našega projekta, saj so tako občutljive rastline, zalivane z našim sistemom zalivanja, zrasle najvišje in bile med najbolj zdravimi sadikami. 8. REFERENCE [1] K. Benyovsky Šoštarič. Zeleni kvadrat, zdravje iz organskega vrta. 2012, str. 45-47. [2] E. Upton, G. Halfacree. Raspberry Pi User Guide, 4th edition, 2016, str. 120-153. [3] Brezični modul nRF24L01. [Online]. Dosegljivo:\\ https://http://www.nordicsemi.com/eng/Products/2.4GHz- RF/nRF24L01 [Dostopano 18. 9. 2017]. [4] A. Freeman. Essential Angular for ASP.NET Core MVC. 2017, str. 112- 303. Slika 4: Lončki po dveh mesecih 68 Indeks avtorjev / Author index Ajanović Alen .............................................................................................................................................................................. 46 Bajec Marko ................................................................................................................................................................................. 27 Bon Jure ..................................................................................................................................................................... 10, 13, 16, 20 Bratko Ivan ................................................................................................................................................................................... 24 Burger Helena .............................................................................................................................................................................. 34 Cikajlo Imre ................................................................................................................................................................................. 34 Colarič Luka ................................................................................................................................................................................. 65 Colnarić Matjaž ............................................................................................................................................................................ 36 Cukjati Iztok ................................................................................................................................................................................. 10 Depolli Matjaž .............................................................................................................................................................................. 39 Fele Žorž Gašper .......................................................................................................................................................................... 46 Gams Matjaž ...................................................................................................................................................................... 7, 46, 49 Golija Andrej ................................................................................................................................................................................ 44 Gradišek Anton ............................................................................................................................................................................ 46 Janković Marko ............................................................................................................................................................................ 27 Jurič Simon................................................................................................................................................................................... 36 Kališnik Jurij Matija ..................................................................................................................................................................... 10 Kokol Peter ................................................................................................................................................................................... 36 Kompara Tomaž ........................................................................................................................................................................... 29 Kurnjek Luka ............................................................................................................................................................................... 41 Matičič Mojca .............................................................................................................................................................................. 46 Matjačić Zlatko ............................................................................................................................................................................ 34 Matkovič Andraž .......................................................................................................................................................................... 20 Mehle Marko ................................................................................................................................................................................ 41 Mlakar Miha ................................................................................................................................................................................. 49 Molan Gregor ............................................................................................................................................................................... 53 Molan Martin ............................................................................................................................................................................... 53 Moraus Stanislav .......................................................................................................................................................................... 36 Novak Franc ................................................................................................................................................................................. 57 Papa Gregor .................................................................................................................................................................................. 57 Pavliha Denis ............................................................................................................................................................................... 39 Pavlin Marko ................................................................................................................................................................................ 57 Perellón Alfonso Ruben ............................................................................................................................................................... 16 Peterlin Ana Marija ...................................................................................................................................................................... 46 Peterlin Potisk Karmen ................................................................................................................................................................. 34 Pileckyte Indre ............................................................................................................................................................................. 13 Pirtošek Zvezdan ........................................................................................................................................................ 10, 13, 16, 20 Planina Andrej .............................................................................................................................................................................. 32 Planinc Nataša .............................................................................................................................................................................. 39 Počivavšek Karolina ..................................................................................................................................................................... 46 Podpadec Matic ............................................................................................................................................................................ 46 Poklukar Špela ............................................................................................................................................................................. 57 Prodan Ana ................................................................................................................................................................................... 46 Rink Saša ...................................................................................................................................................................................... 46 Šef Tomaž .................................................................................................................................................................................... 49 Slemik Bojan ................................................................................................................................................................................ 36 Smokvina Aleš ............................................................................................................................................................................. 39 Tavčar Aleš .............................................................................................................................................................................. 7, 49 Turčin Marko ............................................................................................................................................................................... 36 Ulčar Andrej ................................................................................................................................................................................. 46 Verber Domen .............................................................................................................................................................................. 36 Vidmar Luka ................................................................................................................................................................................ 32 Završnik Jernej ............................................................................................................................................................................. 36 Žibert Janez .................................................................................................................................................................................. 10 Žitnik Slavko ................................................................................................................................................................................ 27 Žlahtič Bojan ................................................................................................................................................................................ 36 69 Žlahtič Grega ................................................................................................................................................................................ 36 Zorman Milan ............................................................................................................................................................................... 36 Zupanc Andrej .............................................................................................................................................................................. 24 Zupančič Domen .......................................................................................................................................................................... 61 70 Konferenca / Conference Uredila / Edited by Delavnica za elektronsko in mobilno zdravje ter pametna mesta / Workshop Electronic and Mobile Healt and Smart Cities Matjaž Gams, Aleš Tavčar Document Outline A - Naslovnica-SPREDNJA - I B - Naslovnica - notranja - I C- Kolofon - I D-E - IS2017 - skupni zacetni del Blank Page F - Kazalo - I G - Naslovnica podkonference - I H - Predgovor - I I - Programski odbor - I J - PDF - I 01-EMZ-Gams 02-EMZ-Cukjati 03-EMZ-Pileckyte 04-EMZ-Perellon 05-EMZ-Matkovic 06-EMZ-Zupanc 07-EMZ-Jankovic 08-EMZ-Kompara 09-EMZ-Planina 10-EMZ-Cikajlo 11-EMZ-Kokol 12-EMZ-Pavliha 13-EMZ-Mehle 14-EMZ-Golija 15-EMZ-Ajanovic 16-EMZ-Sef 17-EMZ-Molan 18-EMZ-P-Pavlin 19-EMZ-P-Zupancic 20-EMZ-P-Colaric K - Index - I L - Naslovnica-ZADNJA - I Blank Page Blank Page Blank Page Blank Page Blank Page Blank Page Blank Page