Prediction Analysis of Dose Verification Based on Plan Complexity Metrics during Radiotherapy

Abstract

Plan validation in radiotherapy includes pre-treatment and in-treatment validation. It is feasible to leverage plan complexity to optimize validation processes, because some research reports that the consistency between planning and measurement or execution decreases as plan complexity increases. Therefore, starting from the plan complexity, this study comprehensively and systematically explores these factors affecting whether the plan verification is passed or not and the strength of their correlations, so as to establish a regression model and optimize the verification process. The IMRT and model-based validation information were retrospectively collected for patients who received tumor radiotherapy at our institution from January 2020 to May 2022. The seventy-eight metrics characterizing the plan complexity were calculated and checked for each plan by an in-house developed computing platform. The comparisons of plan verification results under different tumor species and different verification methods were carried out, and the plan complexity metrics was also compared. Subsequently, Spearman correlation analysis between plan complexity and plan validation under different tumor species, different verification methods and different evaluation indexes was analyzed to provide the basis for regression analysis. One thousand two hundred twenty-six patients were enrolled in the study. The plans in Eclipse TPS were performed by Varian Edge and iX linear accelerators and their verification were executed by EPID, Delat4, and ArcCheck. The gamma index under evaluation index of 3%/3mm, 3%/2mm, 2%/2mm, 1%/1mm were respectively 97.4% (7.1%), 94.8% (10.9%), 86.1% (20.1%), 50.7% (22.6%). The γ and DD of EPID and Detla4 decreased with the increase of TH, and the change of DD was the most significant, while the DTA of Detla4 did not change significantly with TH, and the passing rate of DTA and γ in thoracic and abdominal was the highest in ArcCheck group. Contrast and Variance were the most different between tumor types. The pelvic MIs and MIa were larger than those planned for the other three sites. Except for head and neck CLS and MD, other metrics did not vary significantly among tumor species. The correlation between the metrics characterizing the static characteristics of leaves was more significant for head and neck tumors; the correlation between metrics characterizing the flux complexity were significantly different in pelvic. There was a correlation between plan complexity and γ indicators, and the overall strength was ArcCheck > EPID > Delta4 for head and neck, Delta4 > EPID and ArcCheck for thorax-abdomen, Delta4 and EPID > ArcCheck for pelvic, Delta4 > ArcCheck > EPID for arms and legs. There was a correlation between different plan complexity metrics. Based on this study, it is feasible to predict the verification results based on these complexity metrics, but the regression models were respectively established according to tumor types and validation methods.