Program Information
Accuracy Assurance of Computed Lung IMRT Dose Distributions for Treatment Response Analyses: Two Commercial Algorithms Vs Monte Carlo
O Vassiliev*, Z Liao , R Mohan , UT MD Anderson Cancer Center, Houston, TX
Presentations
SU-I-GPD-T-413 (Sunday, July 30, 2017) 3:00 PM - 6:00 PM Room: Exhibit Hall
Purpose: To identify accurate and efficient dose calculation algorithm for IMRT treatment response analyses by comparing lung dose distributions calculated with a commercial collapsed cone convolution (CCC) algorithm, a deterministic Boltzmann solver (AXB) and Monte Carlo.
Methods: Dose distributions were calculated for 10 lung cancer patients treated previously with IMRT using CCC, AXB, and an EGSnrc code-based MC algorithms. Dose distributions were compared in terms of DVH parameters, absolute point doses and distance to agreement.
Results: Mean PTV doses agreed well between the three algorithms, the median differences CCC-MC and AXB-MC were within ±0.2%, and the respective interquartile ranges were less than 1%. Three outliers were identified for DVH parameters of esophagus and heart where differences exceeded 10%. In each of the three cases more than one adverse factors were present very close to the organ: PTV, bone, air passage, lung or field edge. In all three cases the dose remained well within the tolerance limits and there was no risk to the patients. CCC consistently underestimated dose to the lung in low dose regions, e.g. the median difference CCC-MC for V5 was -5.8%. Point-by-point dose and distance to agreement analysis was performed only where dose exceeded 0.1 Dmax. About 6.5% / 10.2% of voxels (CCC/AXB) failed the 2% / 2 mm criterion, and only 1.5 / 2.5% failed the 3%/3 mm criterion. Discrepancies were localized in low dose lung (CCC), near interfaces (bone/lung, tissue/air, etc.), and beam edges (especially AXB).
Conclusion: Both CCC and AXB algorithms achieve a high level of accuracy. It also suggests that for challenging cases where a significant dose error is likely, the dose should be additionally calculated with Monte Carlo, as a safety measure. In general, the higher speed CCC or AXB may be effectively used for analyzing treatment response data.
Contact Email: