Radiol Oncol 1999; 33(2): 163-8.
Intensity-modulated radiotherapy with a multileaf collimator
Arthur J. Curtin-Savard*, William Parker, Te Vuong, and Ervin B. Podgorsak
Departments ofMedical Physics and Radiation Oncology, Montreal General Hospital, McGill
University, Montréal, Canada
In intensity-modulated radiation therapy (IMRT), a small number of spatially-modulated beamports is used to deliver a uniformly high dose to the target region, while simultaneously producing a better sparing of normal tissues than is possible with conventional, non-modulated beam delivery. When intensity-modulation is carried out luith a multileaf collimator rather than with physical compensators, any required intensity pattern can be produced with a minimal planning effort on the part of the physicist ar dosimetrist, and with reduced manipulation on the part of the therapists. The paper describes the recent implementation of multileaf collimator-based intensity-modulated radiation therapy at McGill University in Montreal.
Key words: radiotherapy-methods; radiotherapy planning, computer-assisted; radiotherapy dosage; multi-leaf collimator,
Introduction
Intensity modulated radiation therapy (IMRT)1-3, carried out with multileaf collimators (MLC)4-6 in combination with 3D planning systems which have the capability far inverse planning7-9, represents the current farefront in radiotherapy. Since MLCs can be retrofitted to existing linear accelerators and controlled with readily available manufacturer hardware and software components, it is
Received 7 November 1998 Accepted 12 January 1999
Correspondence to: Dr. Ervin B. Podgorsak, Department of Medical Physics, Montreal General Hospital, McGill University, 1650 avenue Cedar, Montreal, Quebec, Canada H3G 1A4. Tel: (514 9348052; Fax: (514) 934-8229 *Current addres:
Lowa Linda University Medical Center, Department of Radiation Medicine 11234 Anderson St. Lowa Linda, CA 92354 USA
neither difficult nor excessively expensive to make the transition from conventional to intensity-modulated radiation therapy.
In our department, the MLC-IMRT treatment procedure is typically separated into the. fallowing six distinct steps: (a) definition of beam geometry far a prospective patient; (b) calculation of intensity-modulation matrices far each radiation field; (c) determination of MLC leaf sequencing far each field; (d) calculation and evaluation of the dose distribution; (e) verification of the beam delivery sequence with respect to machine outputs and the MLC sequence; and (f) patient treatment.
In this paper we describe each of the six steps involved in the MLC-IMRT process, and, as an example, present a typical head and neck treatment with the technique.
164
Curtin-Savard A. ]. et al. / Intensity modulated radiation therapy
Materials and methods
Patient simulation and treatment planning
In preparation for treatment planning, the patient is scanned on a dedicated radiotherapy CT-simulator (Picker-5000; Cleveland, OH). The complete set of axial CT images covering the tumour volume and neighbouring structures is transferred to a virtual simulation and image processing workstation (Picker VoxelQ/AcQSim; Cleveland, OH) for the localization and delineation of critical organs and target structures. The beam geometry (treatment isocenter, field sizes, gantry angles, special shielding, etc.) is determined using anatomical landmarks visible on digitally reconstructed radiographs (DRRs) and beam's- eye-views (BEVs) of the previously delineated target and critical organ data.
The axial CT slices together with the beam geometry are then transferred via a computer network to our treatment planning system (CadPlan, version 3.12; Varian, Palo Alta, CA) for calculation of the appropriate isodose distributions. For all our IMRT treatments to date we have used beam configurations common to standard techniques (e.g., three fields for head and neck, four-field box technique for prostate, etc.) in order to facilitate the transition from conventional, non-modulated therapy to MLC-IMRT.
Calculation of the intensity modulation matrix
Once the beam parameters have been established, we instruct the CadPlan system to create intensity-modulation matrices for each field of the treatment plan. Currently, these intensity matrices serve to compensate for irregularities in the patient's contour; however, in the future a complete inverse-planning package will provide the appropriate intensity modulation matrices to optimise the preplanned sparing of critical structures and dose homogeneity within the target volume.
Radiol Oncol 1999; 33(2): 163-8.
Calculation of the leaf sequence
A leaf sequence algorithm10-11 translates an intensity-modulation matrix into a set of MLC leaf configurations delivered at the treatment machine. The leaf sequence algorithm that we use has been developed in-house; however, most modern commercial treatment planning systems now have the ability to write leaf sequences. Our algorithm is designed to write the step-and-shoot type leaf sequences, implying that the leaves move to a given configuration, a prescribed number of monitor units is delivered for this configuration, the leaves then move to the next configuration, followed by the delivery of the prescribed number of monitor units, and so on.
The algorithm for the calculation of the leaf sequence proceeds as follows. At 1 cm intervals throughout the intensity-modulation matrix, profiles of the intensity along the direction of leaf motion are extracted. These profiles represent the intensity to be delivered under each of the 26 leaf pairs of the MLC. Each 2D intensity profile is then partitioned, or "sliced", into a number of regularly spaced intensity levels. The positional coordinates of the intercepts of a given intensity level with the intensity profile form a set of positions at which that particular MLC leaf pair must be placed during beam-on to deliver the given intensity level. This process is repeated for all intensity levels and then for all leaf pairs. Once completed, the resultant set of leaf pair positions is sorted to establish a sequence of segments that can be delivered efficiently. A percentage monitor unit setting (or "dose index") is calculated for each segment from the knowledge of the particular dosimetric characteristics of the linear accelerator. The leaf sequence is then transferred to the Varian MLC Dynamic Beam Delivery (DBD) Toolbox workstation, which controls the 52-leaf MLC of our linear accelerator (Clinac-2300 C/D; Varian, Palo Alta, CA).
165	Curtin-Savard A. ]. et al. / Intensity modulated radiation therapy

Generation and verification of isodose distribution
In order to ensure accurate treatment planning, we reintroduce the MLC segments, determined by the leaf sequence program, into the CadPlan system in the form of separate blocked fields which are weighted according to the calculated "dose index" and calculate a new dose distribution. If there are no errors in the leaf sequence calculation, the dose distribution calculated from the "doseindex" weighted fields will agree closely with the dose distribution calculated with the original intensity matrices, leading to the conclusion that the proposed MLC sequence is adequate for the treatment. The actual treatment is carried out only after full agreement between the two calculated dose distributions is found and the beam delivery verified.
Verification of beam delivery
Prior to patient treatment the leaf sequence for each field is subjected to two dosimetric tests. The first test is a film measurement of the planar dose distribution resulting from the delivered leaf sequence for each individual field consisting of multiple subfields. This measurement confirms the proper orientation of all fields and allows us to verify that the beam intensity delivered to selected points is consistent with the intensity matrix.
The second test is a central axis measurement of the cumulative dose resulting from the leaf sequence for all fields. This test is carried out with a calibrated ionization chamber at the treatment isocenter depth in a tissue equivalent phantom.
Patient treatment
The actual patient treatment is similar to conventional type treatments. The main difference is that, instead of placing a physical compensator in the beam path for each field, the therapist first loads a leaf sequence file into the MLC control computer and then pro-
grams in the total monitor unit setting pertaining to that leaf sequence. The Varian DBD ToolBox software apportions the appropriate percentage of monitor units according to the "dose-index" to each segment of the leaf sequence.
The MLC segments are delivered at the nominal operating dose-rate of the linac (-400 MU/minute). During the time that the MLC leaves are changing position, however, the dose-rate drops to zero allowing the leaf movement under "beam off" conditions before returning to the nominal dose-rate for the delivery of the next MLC segment. Typically, the delivery of an MLC intensity-modulated leaf sequence requires 20%-50% more beam time than would the same treatment produced with physical compensators. However, this loss of beam time is amply made up by the radiation therapists not having to enter the treatment room to change the compensators, blocks, or wedges for each individual field used in the treatment plan.
Results and discussion
The first treatment with MLC-IMRT at the McGill University Hospital Centre was carried out in April 1998 and within six months 15 patients were treated with the technique. We now give an example of a typical MLC-IMRT treatment. The patient was treated for a base of tongue carcinoma with a dose of 44 Gy using an isocentric set-up consisting of two half-blocked lateral parallel-opposed 6 MV photon beams with a collimator setting of llxlO cm2. An additional 16 Gy boost (sparing the spinal cord) was delivered with the same beam geometry. The lower neck region was treated with a single anterior half-blocked field to 50 Gy prescribed at a depth of 3 cm. Intensity-modulation was desirable for the lateral fields to compensate for the uneven surface of the neck region, which under standard conditions would have
Radiol Oncol 1999; 33(2): 163-8.
166
Curtin-Savard A. ]. et al. / Intensity modulated radiation therapy
required the construction of custom-built compensators. Before calculating the intensity matrices for each field, the BEV and DRR modes were used to position shielding for the protection of the base of skull as well as the eyes and nose.
Figure 1 shows the digitally-reconstructed radiograph (DRR) generated for the patient from a series of axial CT slices and used for field placement and delineation as well as for comparison with portal images in the verification of patient positioning prior to dose delivery.
Dose distributions for the initial portion of the IMRT treatment are shown in Figure 2; part (A) for the midseparation sagittal plane, part (B) for an axial slice through the target volume as indicated by the dashed line in part
(A).	The distributions were calculated with the CadPlan 3D treatment planning system and are normalized to 100% at the dose maximum in the target volume. The total dose of 44 Gy, delivered in 22 fractions, was prescribed to the 96% isodose surface which covers the target volume. The dose uniformity in the treated volume is considerably better than that achievable without the use of intensity-modulated beams.
Figure 3 displays the four irregular sub-fields which were used to produce the intensity-modulation for the right lateral beam for the treatment of Figure 2. The irregular fields were produced and the dose was delivered with a multileaf collimator and a step-and-shoot method using the dynamic MLC option on our linear accelerator. For the right lateral field the basic irregular subfield shown in part (A) delivered the 85% of the total dose, while the other three subfields shown in parts
(B),	(C), and (D) each delivered 5% of the total dose.
The patients treated with the IMRT technique have an unusually mild skin reaction compared to reactions normally observed in patients treated with standard techniques. We believe that the Jack of wedges, blocking
Figure l. A digitally reconstructed radiograph (DRR) for a typical patient treated with intensity-modulated beams at McGill University.
trays, and compensators, all of which may increase the skin dose, is to be credited for the reduction in skin reaction.
Summary
Intensity-modulated radiation beams produced with a multileaf collimator are rapidly becoming an essential component of modern radiation therapy. With a wise purchase of equipment, the investment in time and money required for the successful implementation of intensity-modulated radiotherapy into the clinic can be maintained at a reasonable level and yield a substantially improved patient treatment when compared to standard techniques.
Radiol Oncol 1999; 33(2): 163-8.
Curtin-Savard A. J. et al. / Intensity modulated radiation therapy
167
Figure 2. Dose distributions for a typical patient treated for a base of tongue carcinoma with intensity-modulated 6 MV beams. Part (A) is for the rnidseparation sagittal plane, part (B) for an axial plane defined with the dashed line in part (A). The distributions are normalized to 100% at the dose maximum in the target volume.
References
1.	Fraass BA. Clinical use of intensity modulated radiotherapy using multi-segment treatment delivery techniques. Med Phys 1997; 24: 974. (abstract)
2.	Leibel SA, Ling CC, Kutcher GJ, Mohan R, Cordon-Cardo C, Fuks Z. The biological basis of conforrnal three-dimensional radiation therapy. Int J Radiat Oncol Biol Phys 1991; 21: 805-11.
Figure 3. The four irregular subfields which were used to deliver the intensityrnodulated right lateral beam for the dose distribution of Figure 2. The basic irregular subfield of part (A) delivers 85% of the total dose, the other sub-fields deliver 5% each.
Radiol Oncol 2999; 33(2): 163-8.
168
Curtin-Savard A. ]. et al. / Intensity modulated radiation therapy
3.	Ling CC, Burman C, Chui CS, Kutcher GJ, Leibel SA, LoSasso T, et al. Conformal radiation treatment of prostate cancer using inversely-planned intensity-modulated photon beams produced with dynamic multileaf collimation. Int J Radiat Oncol Biol Phys 1996; 35: 721-30.
4.	Galvin JM, Smith AR, Lally B. Characterization of a multileaf collimator system. Int J Radiol Oncol Biol Phys 1993; 25: 181-92.
5.	Jordan TJ, Williams PC. The design and performance characteristics of a multileaf collimator. Phys Med Biol 1994; 39: 231-51.
6.	Klein EE, Harms WB, Low DA, Willcut V, Purdy JA. Clinical implementation of a commercial multileaf collimator: dosimetry, networking, simulation, and quality assurance. Int J Radiol Oncol Biol Phys 1995; 33: 1195-208.
7.	Oldham M, Neal A, Webb S. A comparison of conventional forward planning with inverse planning for 3D conformal radiotherapy of the prostate. Radiother Oncol 1995; 35: 248-62.
8.	Rosen II, Lam KS, Lane RG, Langer M, Morrill SM. Comparison of simulated annealing algorithms for conformal therapy planning. Int J Radiat Oncol Biol Phys 1995; 33: 1091-9.
9.	Wang X-H, Mohan R, Jackson A, Leibel SA, Fuks Zvi, Ling CC. Optimization of intensity-modulated 3D conformal treatment plans based on biological indices. Radiother Oncol 1995; 37: 140-52.
10.	Bortfeld TR, Kahler DL, Waldron TJ, Boyer AL. X-ray field compensation with multileaf collimators. Int J Radiat Oncol Biol Phys 1994; 28: 723-30.
11.	Geis P. Boyer AL. Use of a multileaf collimator as a dynamic missing-tissue compensator. Med Phys 1996; 23: 1199-205.
Radiol Oncol 1999; 33(2): 163-8.