Encrypted login | home

Program Information

Validation of GPU-Based Monte Carlo Code (gPMC) Versus Fully Implemented Monte Carlo Code (TOPAS) for Proton Radiation Therapy: Clinical Cases Study


D Giantsoudi

D Giantsoudi1*, J Schuemann1 , S Dowdell1 , X Jia2 , S Jiang2 , H Paganetti1 , (1) Massachusetts General Hospital and Harvard Medical School, Boston, MA, (2) UT Southwestern Medical Center, Dallas, TX

Presentations

TH-A-19A-11 Thursday 7:30AM - 9:30AM Room: 19A

Purpose: For proton radiation therapy, Monte Carlo simulation (MCS) methods are recognized as the gold-standard dose calculation approach. Although previously unrealistic due to limitations in available computing power, GPU-based applications allow MCS of proton treatment fields to be performed in routine clinical use, on time scales comparable to that of conventional pencil-beam algorithms. This study focuses on validating the results of our GPU-based code (gPMC) versus fully implemented proton therapy based MCS code (TOPAS) for clinical patient cases.

Methods: Two treatment sites were selected to provide clinical cases for this study: head-and-neck cases due to anatomical geometrical complexity (air cavities and density heterogeneities), making dose calculation very challenging, and prostate cases due to higher proton energies used and close proximity of the treatment target to sensitive organs at risk. Both gPMC and TOPAS methods were used to calculate 3-dimensional dose distributions for all patients in this study. Comparisons were performed based on target coverage indices (mean dose, V90 and D90) and gamma index distributions for 2% of the prescription dose and 2mm.

Results: For seven out of eight studied cases, mean target dose, V90 and D90 differed less than 2% between TOPAS and gPMC dose distributions. Gamma index analysis for all prostate patients resulted in passing rate of more than 99% of voxels in the target. Four out of five head-neck-cases showed passing rate of gamma index for the target of more than 99%, the fifth having a gamma index passing rate of 93%.

Conclusion: Our current work showed excellent agreement between our GPU-based MCS code and fully implemented proton therapy based MC code for a group of dosimetrically challenging patient cases.


Contact Email: