Raise — Radiotherapy Accelerated By Intelligent Spatially-Enhanced Segmentation: Development and Multi-Center Evaluation
Abstract
Purpose
To develop and validate RAISE (Radiotherapy Accelerated by Intelligent Spatially-Enhanced Segmentation) across multiple tumor sites and centers.
Methods
RAISE was implemented in the uRT treatment-planning system to analyze axial PTV contours and partition them into longitudinal sub-segments. A beam’s-eye-view analysis optimized rotational segmentation and MLC angles for final dose optimization. A retrospective study assembled a single-center multi-site cohort (89 patients) and a multi-center complex cohort (21 patients). RAISE plans were compared with Tomo and conventional VMAT plans in terms of target homogeneity index (HI), conformality index (CI), OAR metrics, execution time, and patient-specific QA (PSQA).
Results
RAISE produced superior target homogeneity (HI: 0.085±0.05) and consistent improvements in conformity (CI: 0.813±0.10) compared to VMAT and Tomo. Compared with VMAT, RAISE significantly reduced the mean lung dose and the spinal cord maximum dose. Execution efficiency was superior to Tomo, with an average time reduction of 43.35% ± 21.77%. RAISE also demonstrated improved OAR sparing for critical structures, and in the multi-center cohort, CI was significantly better than Tomo. All plans met clinical deliverability criteria with PSQA gamma pass rates exceeding 95%.
Conclusion
RAISE integrates into clinical TPS, offering consistent dosimetric advantages—including improved homogeneity, conformity, and OAR sparing—while substantially shortening execution time compared to Tomo. These results support the clinical feasibility of RAISE for multi-center implementation.