Prediction and classification of VMAT dosimetric accuracy using plan complexity and log-files analysis

Abstract

PurposeWe presented different machine learning models based on log files analysis and complexity indexes to predict and classify the dosimetric accuracy of VMAT plans. MethodsA total of 1302 VMAT arcs from 651 treatment plans were analyzed using the modulation complexity score (MCS) and the dynamic log-files generated by the linac. Predicted and measured fluences were compared using γ-analysis in terms of mean γ-values (γmean) and γ-pass rate (γ%). A kernel regression model was developed aiming to predict individual γ% and γmean values. Multinomial logistic regression (LR), Naïve-Bayes (NB) and support vector machine (SVM) models were developed based on MCS values to classify QA results as “pass” (γ%greater than90 % and γmean < 0.5), “control” (80 % < γ% < 90 % and 0.50 < γmean < 0.75) and “fail” (γ% < 80 % and γmean > 0.75). Training, validation and testing groups were used to evaluate the model reliability. A complexity-based traffic light protocol was implemented to flag pass (green light), control (orange light) and failed plans (red light). ResultsPrediction accuracy of residuals for γ% was 2.1 % and 2.2 % in the training and testing cohorts, respectively. For 2 %(local)/2mm, both γ% and γmean classification performances reported weighted precision, recall and F1-values greater than 90 % for all machine learning models. The optimal MCS threshold value for the identification of failed plans was 0.130, with a sensibility and specificity of 0.994 and 0.952, respectively. The optimal MCS threshold for reliable plans was 0.270, with a sensitivity and specificity of 0.925 and 0.922, respectively. ConclusionsMachine learning can accurately predict the dosimetric accuracy of VMAT treatments, representing an efficient tool to assist patient-specific QA.