Towards Offline Adaptive Therapy for Prostate Patients using Velocity

Abstract

The goal of this project is to assess the feasibility of three offline adaptive workflows for prostate radiotherapy (RT) through a retrospective study, potentially allowing future offline dose assessment and adaptive replanning for prostate bed RT. We investigated an offline adaptive therapy workflow previously developed for head-and-neck patients on a small cohort of seven prostate patients over their first five fractions. Because this workflow was previously shown to have poor deformations in the pelvis, a lower-bound threshold contour was used to reduce bladder-volume variation. Three workflow variations were assessed for accuracy in estimated bladder volume and delivered dose-to-bladder: Velocity’s “Navigator” using full region-of-interest (fROI), Navigator with a reduced ROI (rROI), and structure-guided deformation (SGD). DICE similarity coefficients (DSC) were compared between deformed bladder volumes and true day-of-treatment volumes. Workflow-estimated dose-to-bladder and NTCP were compared to ground-truth calculations. The average ratio of day-of-treatment bladder volume compared to simulation was 82±17% for 6 of the 7 patients. For these patients, the fROI and rROI workflows tended to over-estimate bladder volume, with average DSCs of 0.786±0.055 and 0.757±0.069 respectively, whereas the SGD workflow agreed with the true volume (0.902±0.034). Correspondingly, the fROI and rROI workflows tended to underestimate bladder dose (fractional dose differences up to 94%) and normal tissue complication probability (NTCP), with improved agreement using SGD. The first patient’s bladder tended to be extremely over-filled, with day-of-treatment volume up to 548% simulation volume. Due to the large volume difference, poor deformations resulted in unreliable day-of-treatment bladder volume, dose, and NTCP estimates for all workflows. The SGD workflow with a lower-bound threshold for bladder fill at treatment is feasible for offline bladder-dose assessment for prostate treatments and may enable future adaptive replanning. All workflows were inaccurate for drastically over-filled bladder volumes, but since over-filling does not tend to result in increased bladder dose/NTCP, it is not of great concern for offline dose assessment. Recommendations for clinical use were derived from our findings.