Medical Imaging and Radiotherapy Journal (MIRTJ) 39 (1) 5
Original article
IMAGE QUALITY IN ABDOMINAL CT: A COMPARISON OF 
TWO RECONSTRUCTION ALGORITHMS IN FILTERED BACK 
PROJECTION (FBP)
Albertina RUSANDU1,* Adrian BECK2, Atle HEGGE2, Gabriele ENGH2
1 Norwegian University of Science and Technology, Department of Circulation and Medical Imaging, Trondheim, Norway
2 Department of Radiology and Nuclear Medicine, St. Olavs Hospital, Trondheim, Norway
* Corresponding author:  albertina.rusandu@ntnu.no
Received: 3. 8. 2022
Accepted: 14. 10. 2022
https://doi.org/10.47724/MIRTJ.2022.i01.a001
Medical Imaging and Radiotherapy Journal (MIRTJ) 39 (1)
ABSTRACT
Objectives: The aim of this study was to evaluate the eff ect 
of the choice of kernel on the image quality in abdominal CT 
images with a focus on liver lesion visibility. 
Methods: In this comparative study, 84 abdominal CT 
examinations of patients with liver lesions that included 
parallel series reconstructed with two diff erent kernels (B30 
and B45) were analysed. A subjective assessment of image 
quality was performed using visual grading analysis based 
on anatomical criteria, liver lesion visibility and perceived 
image quality. Objective image quality was assessed using 
measurements of Hounsfi eld unit (HU) values (average and 
standard deviation) in abdominal organs and calculations of 
contrast-to-noise ratios (CNR).
Results: B30 kernel performed signifi cantly better than B45 
in all criteria except for sharpness. The most considerable 
improvement of the image quality was in terms of subjective 
experienced image noise, overall diagnostic image quality 
and visually sharp reproduction of liver lesions. The physical 
measurements showed that CNR increased by up to 46% 
when using B30. 
Conclusions: Using a B30 kernel algorithm for image 
reconstruction reduces noise and thus improves image quality 
and diagnostic accuracy signifi cantly relative to B45.
Key words: kernel, image quality, CT, noise reduction, liver 
lesions
6 Medical Imaging and Radiotherapy Journal (MIRTJ) 39 (1)
Rusandu A. et al./ Image quality in abdominal CT: A comparison of two reconstruction algorithms in Filtered Back Projection (FBP)
Introduction
In computer tomography (CT) examinations, image quality 
depends on scanning parameters, reconstruction technique and 
parameters, together with scanners particularities. One of the 
factors that aff ect image quality and particularly image noise is 
the image reconstruction algorithm, also referred to as kernel. 
In fi ltered back projection (FBP)-based image reconstruction, 
images are obtained by fi ltering the projection data using a 
reconstruction kernel and then back projecting the fi ltered data 
to the image space (1). The kernels incorporate noise reduction, 
spatial resolution- and edge-increasing techniques that are 
applied to the raw data resulting from CT scanning. The choice 
of kernel always implies a trade-off  between image noise and 
sharpness (spatial resolution) (2). CT images can be reconstructed 
multiple times with no additional radiation dose to the patient. 
Diff erent manufacturers operate with diff erent designations 
for the kernels available on their CT-scanners. For example, 
GE uses more descriptive denominations (with kernels names 
like soft, detail, standard, bone, etc.) while others use codes 
(Phillips uses alphabetic denominations, Siemens uses codes, 
such as B30, B40, B45, B80, etc., while Toshiba uses FC08, FC12, 
FC30, etc.). 
The detection and characterization of small focal lesions in 
parenchymal organs represent a challenge for the diagnostic 
radiologist and can have signifi cant importance for a patient’s 
further treatment. Reconstruction algorithms have an impact 
on image quality, i.e. to determine if adjustments in kernel 
reconstructions can improve the detection of parenchymal 
lesions.
Although iterative reconstruction (IR) is increasingly used as a 
result of its radiation dose reduction potential, FBP is still widely 
applied internationally due to some potential disadvantages 
of IR. These include increased implementation cost due to 
necessary purchases for every scanner or the inability to adopt 
this method at all because of older, incompatible scanners 
(1). Another disadvantage of IR is the usual change in noise 
texture compared to FBP images with which radiologists are 
more familiar, which may alter the radiologist satisfaction with 
the images and diagnostic confi dence (3). Another reason 
FBP is still used is that applying the same reconstruction 
technique makes it is easier to compare with previous images. 
The purpose of this study was to evaluate the eff ect of the 
choice of kernel on the image quality in abdominal CT images 
with a focus on liver lesion visibility. 
Material and methods
The CT scanners used in this study were Somatom Defi nition 
AS+ (128 slice), Somatom Defi nition Flash (2 x 128) and 
Somatom Sensation 64 (Siemens Medical Solutions, 
Forchheim, Germany). For a period of one year, all abdominal 
CT examinations included parallel series reconstructed 
with two diff erent kernels (B30 and B45) in order to make it 
easier to compare the images with previous examinations. 
All examinations that showed liver lesions were included in 
the study (n=84). A post-hoc power analysis confi rmed that 
the sample size was appropriate for detecting diff erences in 
image quality with a power of 80%.
Only the portal venous phases were evaluated. Scan timing 
was individualized using bolus-tracking with a threshold of 
150 Hounsfi eld units (HU) in a region of interest (ROI) in the 
abdominal aorta on an axial image through the middle of 
the liver. The arterial phase was acquired using a delay of 25 
seconds after reaching the threshold, and portal venous phase 
was acquired 30 seconds after the arterial phase. Iohexol 
(Omnipaque 350 mgI/ml, GE Healthcare) followed by 30 ml of 
saline was administered through an 18-gauge cannula placed 
in an antecubital vein. The contrast agent amount and fl ow 
were tailored to patient weight (<50 kg 120 ml and 3.2ml/s; 
50–79 kg 150 ml and 4 ml/s; and >80 kg 180 ml and 4.8ml/s). 
The injection time was 37.5 s for all patients. 
All examinations were performed at 120 kVp using automated 
tube current modulation (CareDose4D, Siemens) with 240 
reference mAs. Pitch was set to 0,6 and the rotation time was 
0.5 s/rotation. Both subjective and objective assessments of 
image quality were performed on images from the portal 
venous phase.
Patients’ gender and age were retrieved from Picture 
Archiving and Communicating System (PACS) and, in order to 
compensate for the lack of information about patients’ height 
and weight, eff ective diameter (eq 1) was used as an indicator 
for body habitus. 
 
(eq 1)
Subjective assessment of image quality
The images were evaluated by two radiologists (with 5 and 12 
years of experience) using relative visual grading analysis (VGA). 
The two image series were randomly displayed on the 
left and right monitor in PACS. A Sectra IDS7 (Linkoping, 
Sweden) PACS workstation with two diagnostic Eizo Radiforce 
MX241W monitors (Cypress, CA, USA) was used for image 
evaluation. The monitors’ luminance was 320 cd/m2, and the 
measurements were performed at a distance of 50–60cm 
from the monitor in an ambient lighting of 40–50lux. The 
radiologists evaluated the images independently, blinded to 
reconstruction kernel and without knowledge of the results 
of the physical measurements performed on the images. 
Radiologists were free to use all the tools available in PACS 
that are commonly used for clinical images (adjustment of 
window/level, magnifi cation, etc.).
Table 1: Quality criteria used for visual image assessment
C1: visually sharp reproduction of the liver parenchyma 
C2: visually sharp reproduction of the intrahepatic vessels
C3: visually sharp reproduction of liver lesions
C4: visually sharp reproduction of the spleen parenchyma
C5: visually sharp reproduction of the pancreas
C6: visually sharp reproduction of the kidneys and proximal ureters
C7: visually sharp reproduction of lymph nodes smaller than 15 
mm in diameter
C8: image noise
C9: overall sharpness
C10: total assessment of diagnostic image quality
Medical Imaging and Radiotherapy Journal (MIRTJ) 39 (1) 7
Rusandu A. et al./ Image quality in abdominal CT: A comparison of two reconstruction algorithms in Filtered Back Projection (FBP)
The criteria used in VGA (Table 1) were visualization of liver 
lesions (C3), perceived image quality (C8–C10) and a selection 
of anatomical criteria (C1–C2, C4–C7) from the European 
Guidelines on quality criteria for computed tomography (4).
The ‘2 / -2’ rating (Table 2) was used when the radiologists 
thought it could have diagnostic consequences, for example 
that one could overlook or not completely evaluate something 
seen on one of the images when looking at the corresponding 
image reconstructed with the other kernel.
Table 2: Scoring 
−2: Images on left monitor are much better than images on right 
monitor
−1: Images on left monitor are better than images on right monitor
0: Images on left and right monitor are equivalent
+1: Images on right monitor are better than images on left monitor
+2: Images on right monitor are much better than images on left 
monitor
The results from the VGA were summarized using VGA scores 
(VGAS) (5) for every criterion calculated using equation 2. 
              
(equation 2)
where Sc represents the given individual scores for observer 
(o) and image (i), Ni represents the total number of images, 
and No represents the total number of observers.
Objective assessment of image quality
Attenuation (quantifi ed as average HU) and noise (quantifi ed 
as standard deviation HU) were measured in ROIs of 
approximately 12 mm in diameter placed on axial slices in 
paravertebral muscle, liver parenchyma, liver lesions, spleen, 
pancreas, aorta and fat tissue (Figure 1). To standardize 
measurements, ROIs were then copied and pasted on 
corresponding images reconstructed with the other kernel. 
CNR values were calculated using the following equation (6):
where CNR represents (HUOrgan - HUMuscle) / SDMuscle, 
CNR for liver lesions were calculated using the following 
equation (6):
where CNR represents (HUliver - HULesion) / SDMuscle
Noise diff erence was calculated using the equation (6): 
Noise diff erence =  noise 45–noise 30   x 100
           noise 45 
where noise 45 and noise 30 is the SD measured in the liver 
on the images reconstructed with B45 kernel and B30 kernel, 
respectively. 
Statistical analysis
Statistical analyses were conducted using SPSS for Windows 
version 27 (IBM Inc., Armonk, NY). The highlighted factors 
related to the distribution of data were: average, standard 
deviation, and lowest and highest value. The Shapiro–
Wilk test was used to determine whether the data were 
normally distributed. Diff erences in physical image quality 
parameters between the groups were evaluated using a 
paired t-test. Diff erences in scores for subjective image quality 
were assessed using the Wilcoxon signed-rank test, while 
correlations between measured image quality parameters 
and criteria-based evaluations were analysed using 
Spearman’s rank order. Inter-rater agreement was assessed 
using the weighted Cohen’s kappa test with the following 
interpretation of agreement: 0.00–0.20 slight; 0.21–0.40 fair; 
0.41–0.60 moderate; 0.61–0.80 substantial; and 0.81–1.00 
almost perfect (7). Detailed analyses of percentage agreement 
were also used.
Ethical considerations
Institutional ethics review board approval was obtained 
(Research Committee of the Department of Medical Imaging 
at St. Olavs Hospital nr. 202012/21.04.2020). Written informed 
consent was waived due to the study’s retrospective design. 
No personally identifi able information was recorded.
Results
A total of 84 examinations were assessed. Patient characte-
ristics are presented in Table 3. 
Figure 1: ROIs for the objective measurements of attenuation (quantifi ed as average HU) and noise (quantifi ed as standard deviation HU)
8 Medical Imaging and Radiotherapy Journal (MIRTJ) 39 (1)
Table 3: Patient characteristics presented as average ± standard devi-
ation (minimum – maximum)
Age
Gender 
(male/female ratio) Eff ective diameter
64.47 ± 13.3 (35-89) 41/43 294 ± 38.3 (205-399)
Subjective assessment of image quality
The image quality diff erences made B30 the most preferred 
kernel option, and that kernel performed signifi cantly better 
than B45 in all criteria except for overall sharpness (C9). These 
results are in line with the VGAS for each criterion that show 
the magnitude of the diff erence between kernels (Table 
4) and the percentual distribution of diff erence evaluation 
scores (ure 2). The diff erence in favour of B30 is consistent and 
statistically signifi cant. 
The VGAS show that the most considerable improvement of 
the image quality when using B30 instead of B45 is in terms of 
subjective experienced image noise, overall diagnostic image 
quality and the visually sharp reproduction of liver lesions, 
while the eff ect on the reproduction of lymph nodes smaller 
than 15 mm in diameter is least signifi cant. The diff erences 
in image quality between the two kernels were statistically 
signifi cant for all criteria (p<0.001 for diff erence analysed 
using the Wilcoxon signed-rank test).
In almost 30% of cases, the images reconstructed with the 
B30 kernel were considered much better than the images 
reconstructed with the B45 kernel (Figure 2).
There was high level of agreement between the two 
radiologists regarding the preferred kernel for all criteria, with 
the exception of the visually sharp reproduction of the liver 
parenchyma and overall sharpness. However, in terms of the 
magnitude of the image quality diff erence between the two 
kernels, there was only fair inter-observer agreement (κ in the 
range of 0.2–0.4).  
Objective assessment of image quality
Noise levels measured in all organs were substantially lower 
and CNR considerably higher for the B30 kernel (Table 5). The 
diff erences were statistically signifi cant and the percentual 
diff erences were around 45% in all organs. 
The correlation between the subjective assessed score for 
image noise and measured noise in the liver, spleen and 
muscle was statistically signifi cant. The correlation between 
the subjective evaluation of the reproduction of liver lesions 
and the measured image noise both in the liver and in liver 
lesions was statistically signifi cant. 
Discussion 
This study compared abdominal CT scans reconstructed with 
two diff erent kernels in routine clinical settings. B30 was the 
preferred kernel in this study for all criteria except for one and 
for the overall image quality. The diff erence in both measured 
image quality parameters and subjective image quality 
assessment between B30 and B45 were statistically signifi cant 
for all criteria. 
As expected, the results show a diff erence in both measured 
and perceived image noise, which was signifi cantly lower 
in B30 images. Image noise reduction is proven to result in 
higher confi dence in lesion detection (8). This is confi rmed by 
the correlation between the assessment of the reproduction 
of liver lesions and measured image noise in the liver in Figure 2: Comparison of the images reconstructed with the two kernels
Table 4: Results of criteria-based image quality comparation for B30 
and B45 reconstruction kernels presented as VGA scores, preferred 
option, and percent agreement between the radiologists regarding 
preferred kernel 
Criteria*
VGAS 
(B30>B45)
Preferred 
kernel
Percent 
agreement
C1 0.345 B30 67
C2 0.601 B30 96
C3 0.880 B30 98
C4 0.655 B30 99
C5 0.423 B30 98
C6 0.524 B30 95
C7 0.196 B30 99
C8 1.036 B30 94
C9 -0.529 B45 39
C10 0.964 B30 95
* C1 visually sharp reproduction of the liver parenchyma, C2 visually 
sharp reproduction of the intrahepatic vessels, C3 visually sharp 
reproduction of liver lesions, C4 visually sharp reproduction of the 
spleen parenchyma, C5 visually sharp reproduction of the pancreas, 
C6 visually sharp reproduction of the kidneys and proximal ureters, 
C7 visually sharp reproduction of lymph nodes smaller than 15 
mm in diameter, C8 image noise, C9 overall sharpness, and C10 
total assessment of diagnostic image quality
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
  B45 much better than B30   B45 better than B30
  equivalent image quality   B30 better than B45
  B30 much better than B45
Rusandu A. et al./ Image quality in abdominal CT: A comparison of two reconstruction algorithms in Filtered Back Projection (FBP)
Medical Imaging and Radiotherapy Journal (MIRTJ) 39 (1) 9
Table 5: Average values and standard deviations for image quality parameters measured for the two kernels and percentual diff erence 
(p<0.001 for all parameters in all organs) 
B30 B45 Percentual diff erence (%)
Noise (SD in HU) CNR Noise (SD in HU) CNR Noise (SD in HU) CNR
Liver 15.43±3.49 4.63±1.52 28.87±6.41 2.51±0.74 46.6 45.78
Liver lesion 16.81±4.50 4.55±2.25 30.30±7.93 2.43±1.19 44.52 46.59
Spleen 15.01±3.01 4.70±1.64 28.66±5.66 2.56±0.86 47.63 45.53
Pancreas 18.41±4.38 3.01±1.44 32.29±8.42 1.64±0.79 42.99 45.51
Aorta 16.71±3.76 8.59±2.64 29.90±7.20 4.69±1.45 44.11 45.40
Muscle 15.55±3.57 - 28.44±6.37 - 45.32
Figure 3: The fi gure shows two diff erent window settings in A and B 
for the B30 kernel reconstruction, C and D for B45 of the CT images of 
this 62-year-old patient with primary neuroendocrine tumour of the 
small intestine (window levels are C:50, W:380 in A and C, C: 120, W: 
200 in B and D). Two small liver lesions are shown in the left liver. The 
one anteriorly (thick arrow) is quite easy to see in all reconstructions. 
The other lesion (thin arrow) more posteriorly and medially is diffi  cult 
to see. A change in window level helps the demarcation in the B30 
algorithm. In the B45 reconstruction, the noise makes it much harder 
to detect it in both window settings.
Figure 4: This 52-year-old patient had a primary thymus malignan-
cy with metastasis to the left kidney. A and B are a B30, C and D a 
B45 reconstruction. A and C are shown with a soft tissue window le-
vel C:160, W:450, B and D in window level C: 120, W: 200. While the 
metastatic lesion is somewhat more sharply demarcated in the B45 
kernel reconstruction (right), the subtle internal structure of both 
tumour tissue as well as kidney parenchyma is much better on B30 
reconstructed images.
our study (Figure 3). The image noise reduction obtained 
using B30 instead of B45 (Table 5) was higher than the value 
obtained by Bhosale et al. (9) when comparing a soft kernel 
and standard kernel.   
B45 performed better than B30 for overall sharpness (C9). 
The importance of sharpness depends on the diagnostic task, 
while the assessment of its clinical relevance is beyond the 
scope of this paper. Sharpness, however, is most relevant for 
demarcation in areas with high contrast, such as parenchyma 
against fat. Internal parenchymal structures, such as lobes or 
subtle contrast heterogeneities, are better depicted in B30 
images (Figure 4). Therefore, the overall diagnostic image 
quality scores show that B30 was much better than B45 
in almost 30% of cases. This, together with a low percent 
Figure 5: This 52-year-old patient had a primary thymus malignancy 
with metastasis to the head of the pancreas (same patient as in Fi-
gure 4, window level C: 120, W: 200). In A, the reconstruction kernel is 
B30, while in B it is B45. The edge of the metastasis and subtle tissue 
structure of the surroundings are blurred by the noise on reconstru-
ctions with B45.
Rusandu A. et al./ Image quality in abdominal CT: A comparison of two reconstruction algorithms in Filtered Back Projection (FBP)
10 Medical Imaging and Radiotherapy Journal (MIRTJ) 39 (1)
agreement between the radiologists when scoring overall 
sharpness and no signifi cant correlation between this 
criterion in either the visually sharp reproduction of liver 
lesions or overall diagnostic image quality, suggests that the 
clinical relevance of the lower overall sharpness when using 
B30 might be negligible. 
The considerable diff erences in CNR and quality assessment 
scores indicate the much better visually detectable 
reproduction of liver lesions in B30 and suggest that the 
increased image noise due to the choice of B45 might 
obscure small low-contrast lesions (Figure 5). At fi rst glance, 
the sharp images often seem better, but when analysing the 
organs in more detail, the demarcation between parenchyma 
and pathology is sometimes blurred by noise on B45 
reconstructions (Figure 6). This is especially true for small 
parenchymal lesions. The reason is the sacrifi ce of low contrast 
resolution due to particular image fi ltering and the post-
processing technique, which increase the image noise when 
choosing a sharp kernel that gives better spatial resolution 
(1). It seems that a sharp kernel makes what is already obvious 
even more obvious. However, fi ne diagnostics are convincingly 
better with a softer kernel that gives better texture at the edge 
of metastases (Figure 5). Other criteria with high VGAS were 
subjective evaluated noise (C8) and overall diagnostic image 
quality (C10) (Table 4).
In the pancreas (C5), delimitation against fat looks better on 
B45 at fi rst glance. However, in patients with low BMI, the 
delineation of organs’ contours can be diffi  cult on images with 
high noise level due to the low amount of intra-abdominal 
fat (10), while blurred lesions become more pronounced on 
B30, which is crucial in severe pathology. That gave a slight 
diff erence in image quality with regard to C5 (a VGAS of 0.423 
out of a maximum possible 2 points). 
The correlation between lower levels of measured image 
noise in the organs on the B30 images and the subjective 
assessed scores was statistically signifi cant. However, not all 
the measurements correlated with the scores given by the 
radiologists, which might be explained by the fact that some 
anatomical structures may be more important than others 
for the anatomical region or pathology being investigated. 
More studies are required in this area to identify the 
weighting factors of the criteria, depending on the clinical 
indication (11).
The kappa values indicate some inter-observer diff erences. 
This diff erence might be caused by the diffi  culty in obtaining 
identical scores when a large scoring scale is used (12) or 
the diff erent use of viewing tools, but it might also be an 
underlying diff erence between the reader’s image quality 
expectancy or the fact that reader’s preference scale might 
also change during the reading session which is described 
in literature as adaptation (13). VGA results when visualizing 
diff erent noise textures might also be infl uenced by the 
experience of the radiologist (5). Another reason for the 
low kappa might be the ambiguity of the criteria, i.e. the 
sharp reproduction of the liver that might be subject to 
interpretation (it is worth noting that the percent agreement 
was also lower for C1) or diffi  culty in scoring normal anatomy 
with regard to diagnostic quality in the absence of pathology 
in the assessed organ. The use of image quality criteria stated 
in European guidelines is recommended for optimizing CT 
protocols based on the assumption that sharply reproduced 
anatomy results in sharply reproduced pathology. However, 
the relationship between the reproduction of anatomy and 
the detection of pathology is still unclear and further studies 
are needed, including an analysis in which pathology is taken 
into consideration to evaluate the relationship between 
image quality and diagnostic effi  cacy (10, 14). Similar kappa 
values were reported in studies using similar image quality 
assessment methods (10). However, the extent of diff erences 
showed by the kappa values is not confi rmed by the percentual 
agreement which was over 90 for most of the criteria, while 
percentual agreement is considered a more informative 
agreement measure for clinicians (15). 
The present study is subject to several limitations. 1. A 
statistically signifi cant diff erence in image quality assessment 
results does not necessarily mean a diff erence in diagnostic 
performance. However, because CNR is considered a 
signifi cant predictor for lesion detection, (16) image noise 
reduction may result in higher confi dence in lesion detection. 
2. Despite the randomization of the images, a truly blinded 
comparison was impossible due to the noticeable diff erences 
in image noise between the images reconstructed with the 
two kernels. 3. Only kernels from one vendor and only portal 
venous phase images were evaluated. 4. VGAS was the only 
scoring system used for quantifying the criteria-based image 
quality assessment. However, VGAS is still widely used to 
demonstrate the magnitude of the diff erence between 
options and providing a context to interpret the physical 
measurements (5) despite their shortcomings (17, 18), while a 
Wilcoxon test value is equal to the area under the curve (AUC) 
in a receiver operating characteristics (ROC) analysis of the 
same data (19).
Conclusion
The comparative image quality assessment demonstrates 
the superiority of B30 over B45 kernel reconstruction in 
abdominal CT examinations. This approach provides a 
statistically signifi cant reduction in image noise, and an 
increase in CNR and higher VGA scores for all criteria except 
Figure 6: An 82-year-old patient with a duodenal malignancy obstru-
cting the papilla vateri, with metastatic liver disease. Air in the intra-
hepatic biliary tree after Endoscopic retrograde cholangiopancre-
atography (ERCP) with stenting. The patient was inoperable due to 
comorbidity. Metastases to the liver are clearly more visible on the 
B30 kernel reconstruction (A) compared to B45 (B), window level C:50, 
W: 380
Rusandu A. et al./ Image quality in abdominal CT: A comparison of two reconstruction algorithms in Filtered Back Projection (FBP)
Medical Imaging and Radiotherapy Journal (MIRTJ) 39 (1) 11
for overall sharpness. With the main goal of achieving the 
highest subjective sensitivity for detecting focal lesions, the 
criterion “sharpness” proved to be a secondary factor in this 
study and is negligible.
References
1. Ehman EC, Yu L, Manduca A, Hara AK, Shiung MM, Jondal 
D, et al. Methods for clinical evaluation of noise reduction 
techniques in abdominopelvic CT. Radiographics. 
2014;34(4):849-62. https://doi.org/10.1148/rg.344135128
2. Bushberg JT. The essential physics of medical imaging. 3rd 
ed. ed. Philadelphia: Wolters Kluwer Health/Lippincott 
Williams & Wilkins; 2011.
3. Fletcher JG, Krueger WR, Hough DM, Huprich JE, Fidler 
JL, Wang J, et al. Pilot study of detection, radiologist 
confi dence and image quality with sinogram-affi  rmed 
iterative reconstruction at half-routine dose level. J 
Comput Assist Tomogr. 2013;37(2):203-11. DOI: 10.1097/
rct.0b013e31827e0e93
4. Jessen K, Shrimpton P, et al., Panzer W. European 
guidelines on quality criteria for computed tomography. 
Luxembourg: Luxembourg: OPOCE; 2000.
5. Verdun FR, Racine D, Ott JG, Tapiovaara MJ, Toroi P, Bochud 
FO, et al. Image quality in CT: From physical measurements 
to model observers. Physica Medica. 2015;31(8):823-43. 
https://doi.org/10.1016/j.ejmp.2015.08.007
6. Eichler M, May M, Wiesmueller M, Saake M, Heiss R, Uder 
M, et al. Single source split fi lter dual energy: Image 
quality and liver lesion detection in abdominal CT. 
European Journal of Radiology. 2020;126:108913. https://
doi.org/10.1016/j.ejrad.2020.108913
7. Landis JR, Koch GG. The Measurement of Observer 
Agreement for Categorical Data. Biometrics. 
1977;33(1):159-74. PMID: 843571
8. Goenka AH, Herts BR, Dong F, Obuchowski NA, Primak 
AN, Karim W, et al. Image Noise, CNR, and Detectability 
of Low-Contrast, Low-Attenuation Liver Lesions in a 
Phantom: Eff ects of Radiation Exposure, Phantom Size, 
Integrated Circuit Detector, and Iterative Reconstruction. 
Radiology. 2016;280(2):475-82. https://doi.org/10.1148/
radiol.2016151621
9. Bhosale PW-B, Nicolaus; Wei Wei; Kundra, Vikas; Tamm, 
Eric. Comparing CNR, SNR, and Image Quality of CT Images 
Reconstructed with Soft Kernel, Standard Kernel, and 
Standard Kernel plus ASIR 30% Techniques. International 
Journal of Radiology. 2016;2(2). http://www.ghrnet.org/
index.php/IJR/article/view/1101/1773
10. Kataria B, Althén JN, Smedby Ö, Persson A, Sökjer H, 
Sandborg M. Assessment of image quality in abdominal 
CT: potential dose reduction with model-based iterative 
reconstruction. Eur Radiol. 2018;28(6):2464-73. DOI: 
10.1007/s00330-017-5113-4
11. Zarb F, Rainford L, McEntee MF. Image quality assessment 
tools for optimization of CT images. Radiography. 
2010;16(2):147-53. https://doi.org/10.1016/j.
radi.2009.10.002
12. McHugh ML. Interrater reliability: the kappa statistic. 
Biochem Med (Zagreb). 2012;22(3):276-82. PMID: 
23092060
13. Helson H. Adaptation-level theory. Oxford, England: 
Harper & Row; 1964. xvii, 732-xvii
14. Jurik A, Petersen J, Jessen KA, Bongartz G, Geleijns J, 
Golding SJ, et al. Clinical Use of Image Quality Criteria 
in Computed Tomography: A Pilot Study. Radiation 
Protection Dosimetry. 2000;90(1-2):47-52. https://doi.
org/10.1093/oxfordjournals.rpd.a033142
15. de Vet HCW, Mokkink LB, Terwee CB, Hoekstra OS, Knol 
DL. Clinicians are right not to like Cohen’s κ. BMJ : British 
Medical Journal. 2013;346:f2125. https://doi.org/10.1136/
bmj.f2125
16. Baker ME, Dong F, Primak A, Obuchowski NA, Einstein 
D, Gandhi N, et al. Contrast-to-Noise Ratio and Low-
Contrast Object Resolution on Full- and Low-Dose MDCT: 
SAFIRE Versus Filtered Back Projection in a Low-Contrast 
Object Phantom and in the Liver. American Journal of 
Roentgenology. 2012;199(1):8-18. https://www.ajronline.
org/doi/10.2214/AJR.11.7421
17. Båth M. Evaluating imaging systems: practical 
applications. Radiation Protection Dosimetry. 2010;139(1-
3):26-36. https://pubmed.ncbi.nlm.nih.gov/20147386/
18. Månsson LG. Methods for the Evaluation of Image Quality: 
A Review. Radiation Protection Dosimetry. 2000;90(1-
2):89-99. https://doi.org/10.1093/oxfordjournals.rpd.
a033149
19. Precht H, Hansson J, Outzen C, Hogg P, Tingberg A. 
Radiographers’ perspectives’ on Visual Grading Analysis 
as a scientifi c method to evaluate image quality. 
Radiography. 2019;25:S14-S8. https://doi.org/10.1016/j.
radi.2019.06.006
Rusandu A. et al./ Image quality in abdominal CT: A comparison of two reconstruction algorithms in Filtered Back Projection (FBP)