Room: AAPM ePoster Library
Treatment quality for proton therapy patients is monitored using weekly CT scans (reCTs) at our center. Target contours are manually re-delineated on all reCTs which is very time consuming. To increase efficiency, we propose a fully automated workflow for reCT evaluation. This study evaluates if automated contour propagation leads to the same level of plan-adaptation as the manual time-consuming workflow.
So far, 78 patients (13 neuro, 14 head-and-neck, 34 breast, 4 lymphoma, and 14 lung) have finished treatment at our Mevion S250i Hyperscan proton therapy system; all patients were included in this study. Treatment planning was performed in RayStation using Monte-Carlo dose calculation, and robust optimization. Contours were both manually re-delineated (CTVmanual) and propagated (CTVauto; automated workflow) from the planning-CT to each reCT applying deformable image registration in Raystation. Dose recalculations were performed on all reCTs and the clinical goals (target under- and over-dosage for nominal and robust evaluated voxel-wise minimum and maximum doses) were evaluated for both CTVmanual and CTVauto.
Considering all clinical goals (multiple goals per patient), the specificity was 97% and the false-negative rate was 15/865 (1.7%; 1.3% excluding goals failing by less than 0.3% for CTVmanual and therefore not clinically considered). Not all failing goals led to plan-adaption, however, 14 plan-adaptations were performed for eleven patients for dosimetric reasons. All these adaptations were flagged by the automated workflow by at least one clinical goal. Only for two reCTs used for plan-adaptation, a false-negative goal was observed, however, true-positive failing goals were also detected for these reCTs, meaning that the clinical plan-adaptions were appropriately flagged.
A time saving per reCT of 2 hours for RTTs and 1 hour for RTOs is expected. This workflow will streamline the plan-adaptation evaluation procedure ensuring same day results, instead of three days as in our current practice.
Funding Support, Disclosures, and Conflict of Interest: This work was partially financially supported by Mevion Medical Systems