Clinical validation of statistical predictive model for patient-specific quality assurance outcomes in stereotactic...

To evaluate Monte Carlo (MC) secondary dose verification for predicting ionization chamber (IC)-based patient-specific quality assurance (PSQA) measurements for stereotactic body radiotherapy (SBRT) plans.

IC-based PSQA is a trusted method for verifying accurate delivery of absolute dose calculated by the treatment planning system (TPS). However, these measurements are often time-consuming and challenging to perform precisely, especially for small-volume SBRT targets. To investigate an MC-based method as a viable alternative, a logistic regression model was developed to predict measurement-based PSQA results utilizing 400 retrospectively collected IC PSQA measurements across our system. Each clinically approved plan was recalculated using a commercially available secondary MC-based dose calculation platform (Rad MonteCarlo, Radformation, NY). The dose to a contoured volume corresponding to the active IC volume was recorded. Additionally, measurement setup uncertainty was modeled by placing equivalent volumes +/- 2 mm in each cardinal direction. The TPS-calculated value was compared to the average MC-simulated values for all contours. Receiver Operating Characteristic (ROC) analysis was performed on an additional dataset of 328 prospective PSQA measurements to determine MC-based QA prediction thresholds for indicating when physical measurements can be safely avoided.

Of the 400 model plans, the percent differences between IC and TPS doses were [Median: -0.06%, Range: -19.6%-4.5%]. The percent differences between IC and MC doses were [Median: 0.17%, Range: -21.8%-5.1%]. When investigating MC against TPS dose for predicting likely PSQA failures, ROC analysis yielded an AUC of 0.76. Based on threshold analysis of the prospective validation dataset, a difference of 1% between MC and TPS calculations resulted in zero false negatives, and would safely reduce the number of required IC measurements by 46%.

This study demonstrates feasibility of and a workflow for implementing MC-based secondary dose calculations to reduce the number of physical measurements required for PSQA without compromising safety and quality.