ESTRO 2025 - Abstract Book

S3396

Physics - Machine learning models and clinical applications

ESTRO 2025

Results: As shown in Table 1, the decision-making assistant provides representative results. ATS dosimetry, derived from the deep-learning dose predictor, guides ATP recommendations. In this example, ATP is recommended due to a 3% reduction in bladder Dmax with ATS compared to the pre-plan, along with reduced OV at the bladder-PTV interface, indicating improved bladder dosimetry. Further, a 5% reduction in rectal Dmax and decreased OV at rectal-PTV interfaces support enhanced rectal wall dosimetry. These findings support choosing the ATP workflow, as it effectively adapts to daily anatomical changes without requiring full re-optimization, thereby optimizing staff resources and reducing treatment time.

Conclusion: This study proposed a clinical decision-making framework that integrates deep-learning-based dose prediction with deterministic dosimetric modeling to guide workflow selection in prostate MRgART. By enabling accurate plan quality estimations for both ATS and ATP workflows, this tool optimizes resource utilization and enhances workflow efficiency while maintaining reasonable treatment quality.

Keywords: SBRT/SAbR, MR-guided ART, oART

References: 1. Wang, S. et al., Efficacy of an Adaptive Plan Quality Predictor for Decision Support in Online MR-Guided Adaptive Radiotherapy for Prostate Stereotactic Ablative Radiotherapy. Int J Radiat Oncol Biol Phys Volume 120, Issue 2, e595 2. Shi, J.J. et al., Dosimetric Comparison between Adaptation Strategies for MRI-Guided Dose-Intensified Prostate Stereotactic Ablative Radiotherapy. Int J Radiat Oncol Biol Phys, Volume 120, Issue 2, e584 - e585