Automated prospective quality assurance of tumor segmentation and treatment planning in the hypo-FLAME 3.0 trial.

Abstract

Iso-toxic focal boosting in prostate cancer radiotherapy is associated with improved disease-free survival but faces challenges due to tumor segmentation variability and unconventional treatment planning. The aim of this study was to develop an automated system of prospective quality assurance (QA) on tumor segmentation and treatment planning in a clinical trial setting and to implement it in the study arm of the multi-center hypo-FLAME 3.0 trial (NCT05705921) on iso-toxic focal boosting. We developed an automated QA system to process data from participating centers, apply tumor segmentation evaluation models and plan evaluation models and generate reports for centralized review. Tumor segmentation was evaluated by a model generating tumor probability maps from MRI scans. Focal boost quality was evaluated with a model predicting achievable near-minimum dose for each tumor based on anatomical features. The QA system was applied to 134 patients who participated in the study arm of the hypo-FLAME 3.0 trial between June 2023 and July 2025. They were recruited from 11 centers in the Netherlands and Belgium. Three and 10 out of 134 cases were recommended for resegmentation and replanning, respectively. We responded to external institutes within 12 business hours in 77% and 81% of the cases for the segmentation and the planning QA, respectively. We developed an automated system of prospective QA in a clinical trial setting and implemented it for all patients in the study arm of the phase III hypo-FLAME 3.0 trial.