Premium
SU‐C‐BRD‐02: Monte Carlo Based VMAT Dose Re‐Optimization for Patient Specific Quality Assurance and Failed Plan Recovery Using Re‐Weighted Aperture MUs
Author(s) -
Folkerts M,
Modiri A,
Ungun B,
Jia X,
Jiang S,
Sawant A
Publication year - 2015
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4923797
Subject(s) - quality assurance , monte carlo method , dosimetry , weighting , nuclear medicine , radiation treatment planning , aperture (computer memory) , computer science , particle swarm optimization , algorithm , medicine , mathematical optimization , medical physics , mathematics , statistics , physics , radiation therapy , radiology , external quality assessment , pathology , acoustics
Purpose: To offer an answer to the question: “What do we do if a treatment plan is shown to be sub‐optimal or unsafe after patient specific VMAT QA based on Monte Carlo (MC) dose calculations?” We attempt to disprove the null hypothesis that re‐weighting the MUs delivered through each aperture of a VMAT plan will not improve the MC delivered dose distribution. Methods: We interface with a GPU‐based patient specific MC quality assurance (QA) system to generate dose distributions for apertures of a multiple arc VMAT head‐and‐neck treatment plan. The distributions were combined into a dose‐per‐aperture matrix. We tested and compared the results of two optimization schemes. To solve the convex problem of re‐weighting the MC dose matrix to match the dose provided by the treatment planning system (TPS), the least‐squares (LSQ) algorithm was used. To meet or improve our DVH goals, a particle swarm optimization (PSO) algorithm was implemented. After normalizing each Result to match the 66.0Gy D95 prescription requirement, we used QUANTEC guidelines to evaluate the clinical impact of each approach. Results: Our non‐convex PSO scheme to re‐weight MUs was able to produce matching if not better OAR DVH results for the one plan we tested when compared to the original plan. For example, the spinal cord was prescribed a Dmax of 42.8Gy, however, the MC QA results reported 44.7Gy; the PSO was able to reduce that Result to 42.0Gy, a 6% improvement; the LSQ optimization produced 45.2Gy. Also, the PSO reduced the max dose of the PTV from 81.7Gy (MC QA) to 75.1Gy, while the LSQ algorithm produced 77.4Gy. Conclusion: We successfully disproved our null hypothesis and showed that it is indeed possible to improve the MC dose distribution of a VMAT plan by re‐weighting MUs delivered through each aperture. This project was partially funded by NIH grant #R01CA169102