Utilizing knowledge from prior plans in the evaluation of quality assurance.
Increased interest regarding sensitivity of pre-treatment intensity modulated radiotherapy and volumetric modulated arc radiotherapy (VMAT) quality assurance (QA) to delivery errors has led to the development of dose-volume histogram (DVH) based analysis. This paradigm shift necessitates a change in the acceptance criteria and action tolerance for QA. Here we present a knowledge based technique to objectively quantify degradations in DVH for prostate radiotherapy. Using machine learning, organ-at-risk (OAR) DVHs from a population of 198 prior patients' plans were adapted to a test patient's anatomy to establish patient-specific DVH ranges. This technique was applied to single arc prostate VMAT plans to evaluate various simulated delivery errors: systematic single leaf offsets, systematic leaf bank offsets, random normally distributed leaf fluctuations, systematic lag in gantry angle of the mutli-leaf collimators (MLCs), fluctuations in dose rate, and delivery of each VMAT arc with a constant rather than variable dose rate.Quantitative Analyses of Normal Tissue Effects in the Clinic suggests V75Gy dose limits of 15% for the rectum and 25% for the bladder, however the knowledge based constraints were more stringent: 8.48 ± 2.65% for the rectum and 4.90 ± 1.98% for the bladder. 19 ± 10 mm single leaf and 1.9 ± 0.7 mm single bank offsets resulted in rectum DVHs worse than 97.7% (2σ) of clinically accepted plans. PTV degradations fell outside of the acceptable range for 0.6 ± 0.3 mm leaf offsets, 0.11 ± 0.06 mm bank offsets, 0.6 ± 1.3 mm of random noise, and 1.0 ± 0.7° of gantry-MLC lag.Utilizing a training set comprised of prior treatment plans, machine learning is used to predict a range of achievable DVHs for the test patient's anatomy. Consequently, degradations leading to statistical outliers may be identified. A knowledge based QA evaluation enables customized QA criteria per treatment site, institution and/or physician and can often be more sensitive to errors than criteria based on organ complication rates.
Stanhope, C; Wu, QJ; Yuan, L; Liu, J; Hood, R; Yin, F-F; Adamson, J
Volume / Issue
Start / End Page
Pubmed Central ID
Electronic International Standard Serial Number (EISSN)
Digital Object Identifier (DOI)