A dose analysis framework was implemented for analysis of MC simulation results, and also to calculate MC superposition dose distributions. MC superposition dose distributions are calculated by superposing single MC source distributions across all source co-ordinates and in many cases in later sections are used instead of TG-43U1 (3) for comparison with full MC simulation. This is so that the comparison is not affected by differences between MC source models and TG-43U1 consensus data, or differences
due to the TG-43U1 line source approximation. This section describes how the framework was implemented and validated.
Method
Source co-ordinates, source activity, source dwell times (HDR only) and structure sets were exported from the TPS (for 125I treatments the TPS is
VariseedTM v8.0 (Varian Medical Systems, Inc., Palo Alto, CA, USA) and
for 192Ir treatments the TPS is Oncentra Prostate™ v4.0 (Elekta AB)). For some cases (CT tissue based models) the underlying image set was also exported. The dose analysis framework was implemented in MatlabTM R2010a (MathWorks, Natick, MA, USA) and included code to auto- generate MC simulation input files based on the exported patient data.
The dose and DVH calculation framework involved the following steps: • Extract structure points from DICOM structure data and create 3D
masks for all structures in the structure sets (for example prostate, urethra and rectum).
• Calculate a 3D dose distribution. For MC simulated dose, this is the raw dose converted to Gy as described in 2.1.3. For the MC
superposition dose, the dose at each point in the distribution was calculated by summing the dose from each source at that point. A single source cylindrical dose distribution calculated using MC simulation, with resolution 1mm and extending 10cm along and 10cm away from the source was used to calculate the contribution from each source. Bi-linear interpolation was used to interpolate between points in the cylindrical dose grid (an inverse-square law correction was applied to the dose grid to smooth the interpolation, and then removed after interpolation). The dose calculation
resolution was 1 mm x 1 mm in-plane, with dose calculated on the slices in the original imaging data (2 mm spaced for CT, either 2.5 mm or 5 mm spaced for TRUS).
• For 125I treatments, all sources were assumed to be oriented perpendicular to the plane of the CT/TRUS slices. For 192Ir
treatments, corrections for source orientation were applied using the simplifying assumption that the same rotation could be applied to all sources in a single catheter (Oncentra ProstateTM calculates a
separate rotation for each source (111)).
• For dose calculation from MC simulation results of HDR treatments, for voxels that intersect a catheter, dose was interpolated from surrounding voxels to exclude high dose catheter voxels from the DVH analysis.
• Once dose calculation was completed, the 3D structure masks were applied to the 3D dose distributions to calculate DVH statistics.
The resulting DVH calculations were benchmarked by comparing results to DVH statistics calculated by the TPSes. 15 clinical post-implant CT based plans were compared for the 6711 seed model. 5 HDR prostate treatment plans were compared for the 192Ir source model. For the 6711 seed
comparison, TG-43 source data calculated from single seed MC simulations was added to Variseed and used instead of 6711 consensus data. This is to remove any effects due to differences between the MC seed model and the consensus data, as noted in 2.2. This was not necessary for validation for the
192Ir source, as MC simulation of that source showed good agreement with
consensus data.
Results
Table 11 compares mean DVH parameter values for 15 post-implant 6711 seed implant plans, as calculated by Variseed using the MC source data from 2.2 and the dose analysis framework for the MC superposition dose distribution.
Table 12 compares the mean DVH parameter values for 5 HDR prostate patients from Oncentra Prostate (TG-43U1(3) calculation) and MC superposition dose calculations. Planning target volume (PTV) data is included.
Table 11 Results from benchmarking DVH calculation code comparing Variseed output for the MC simulated source data (MC-SRC) and output from the dose
analysis framework MC superposition dose (MC-SUP). All results are mean values for 15 patients using the 6711 seed.
DVH Parameter Variseed MC-SRC
MC-SUP Difference Variseed MC-SUP – MC-SRC Prostate volume 35.1 cm3 35.2 cm3 0.3% Prostate D90 136.6 Gy 137.5 Gy 0.7% Prostate V100 87.3% 87.6% 0.3% Prostate V150 56.1% 57.0% 0.9% Prostate V200 25.8% 26.7% 0.9% Urethra D10 248.0 Gy 255.4 Gy 3.0% Rectum D2cm3 118.3 Gy 119.2 Gy 0.8%
Table 12 Mean DVH parameter values for 5 HDR prostate patients comparing Oncentra Prostate and MC-SUP dose calculations
DVH Parameter Oncentra Prostate MC-SUP Difference Oncp –MC-SUP PTV volume (cm3) 46.6 46.9 0.7% PTV D90 (Gy) 15.5 15.5 -0.2% PTV V100 (%) 92.4 92.2 -0.3% Prostate volume (cm3) 31.9 32.5 1.8% Prostate D90 (Gy) 17.0 17.0 0.5% Prostate V100 (%) 99.6 99.5 -0.1% Prostate V150 (%) 27.0 27.9 0.8% Prostate V200 (%) 6.87 4.84 -2.0% Urethra D10 (Gy) 17.2 17.2 0.0% Rectum D2cm3 (Gy) 8.86 8.89 0.4% Conclusion
Small differences are observed between DVH parameters calculated by the DVH calculation framework and the TPSes. These differences are not unexpected as there are differences in the dose calculation resolution and TPSes typically calculate DVH parameters by randomly sampling points within volumes for speed, whereas the DVH calculation framework calculates the DVH using all elements in each structure. The differences
observed are smaller than those that have been observed in a comparison of commercial treatment planning systems (112). Therefore the DVH
comparisons demonstrate that the dose and DVH calculation framework is valid.