Impact of the number of discrete angles used during dosecomputation for TomoTherapy treatments

Verboomen, Claude [UCL] Vynckier, Stefaan [UCL] Sterpin, Edmond [UCL]

I. Introduction Until recently (up to treatment planning software version 4.0.x), the TomoTherapy treatment planning system (TPS) used 51 angles for dose computation. As shown by various quality assurance studies of TomoTherapy treatments, this approximation works for most of the configurations, especially with usual modulation factors (around 2.0), small off-axis distances and relatively large tumors. However, treatment plans with small modulation factors display leaves open a larger fraction of the discrete angle. Off-axis positioning breaks the cylindrical symmetry of the system. These two conditions make the hypothesis of a static gantry for each 51 angles less realistic. Accuracy is also likely to be more affected proportionally for a small target. II. Material and methods The effect on accuracy of changing the number of angles was quantified with the following framework: 1. Qualitatively predict the impact of the discretization of gantry rotation for various modulation factors, target sizes, and off-axis positions using a simplified analytical algorithm, which assumes a perfect multileaf collimator and no scatter when photons interact; 2. Analyse difference between dose computation from version 4.0.x and 4.1.x of the TPS [Up to version 4.0.x included, TomoTherapy TPS approximates gantry rotation by computing dose from 51 discrete angles. In versions 4.1.x and later, TomoTherapy oversamples the projections to better account for gantry rotation (153 discrete angles), but only during full scatter optimization and final calculation (i.e., not during optimization in “beamlet” mode)]; 3. Perform regular quality assurance using measurements with EDR2 radiographic films; 4. Isolating the effect of changing the number of discretized angles only (51, 153, and 459) using a previously validated Monte Carlo model (TomoPen1). The diameters of the targets were 2, 3, and 5 cm; off-axis central positions of target volumes were 5, 10, 15, and 17 cm (when accepted by the treatment unit); planned modulation factors were 1.3 and 2.0. These configurations are illustrated on Fig 1. FIG. 1. Schematic diagram (not to scale) of the experimental setup. III. Results · Analytical model predictions The maximum dose deviation equaled 12.9% (for a 2 cm target size placed 17 cm offaxis). Deviations larger than 2% were observed for all off-axis positions simulated with the 2 cm target. For larger targets, deviations higher than 2% were observed for offaxis distances superior to 10 cm. The analytical algorithm showed also the effect of varying modulation factor, with an additional deviation up to 7% for low modulation factor compared to results obtained with a high one. · Comparison measurements-TPS Differences were quantified using gamma analysis of film dose versus computed dose for two sets of criteria: 2%/2mm and 3%/3 mm. For extreme configurations (3 cm tumor, 1.3 modulation factor, 15 cm off-axis position), effects on dose distributions were significant with 89.3% and 95.4% of the points passing gamma tests with 2%/2 mm and 3%/3 mm criteria, respectively, for TPS software version 4.0.x (51 gantry angles). For the 4.1.x version (153 gantry angles), excellent agreement was observed for all cases: the passing rate was 100% for both gamma criteria. · Comparison TPS-TomoPen Figure 1 shows a comparison for a particular case between dose profiles computed with Tomo-HD unit (version 4.1.x) and TomoPen. Differences between Tomo-HD and TomoPen for 153 and 459 gantry angles are negligible (within 1%/1 mm). However, significant differences are observed when comparing Tomo-HD and TomoPen for 51 gantry angles. With a smaller modulation factor, agreement is worse for dose distributions computed with 51 gantry angles with an additional deviation of up to 3% of prescribed dose within the target. TomoPen for 51 gantry angles reproduces the behavior of software version installed on Hi-Art machine (4.0.x, 51 gantry angles). The differences showed by TomoPen confirmed the qualitative predictions of the analytical model. Those results confirmed that the measured effect comes from the discretization of gantry rotation only. FIG. 2. Comparison between computed dose profiles acquired perpendicular to off-axis direction using Tomo-HD software (version 4.1.x) and TomoPen for 51, 153 or 459 gantry angles. This configuration is 3 cm target size, 15 cm off-axis for actual modulation factors of 1.3 for (a) and modulation factor of 1.7 for (b). IV. Conclusion A sufficient number of gantry angles is necessary during dose computation to ensure planning robustness versus target off-axis position, low modulation factors, and small target sizes, as it may occur in SBRT treatments. This study, published in medical physics2, showed that the use of 51 gantry positions could lead to significant deviations for off-axis positions of more than 10 cm. However, the use of 153 gantry angles was showed sufficient to cope with all configurations tested during this study, even the most extreme. Although versions 4.1.x and above of TomoTherapy TPS software do implement 153 gantry angles in “get full dose” and “full scatter” iterations, the effect can only be corrected by iterating in “full scatter” mode, which is unpractical in current systems. New releases of TomoTherapy TPS, with GPU implementation3, might solve the problem by allowing iterations including 153 gantry angles at practical speed.