Multi-Criteria Optimization for Efficient Online Adaptive Radiotherapy Planning In Prostate SBRT on a 1.5T MR-Linac
Abstract
Purpose
Daily re-planning based on warm-start optimization is a common strategy for online adaptive radiotherapy. This method is beneficial in cases when differences between the reference and daily anatomy are within a limited range. We quantify the effects of this strategy (adapt-to-position, ATP) on a 1.5T MR-Linac and study whether multi-criteria optimization (MCO) for reference planning enables efficient online IMRT re-planning (adapt-to-shape, ATS) without manual intervention.
Methods
Two prostate SBRT cohorts were analyzed: (A) standard SBRT prescription (N=50) and (B) GTV-boosting with urethral sparing (N=15). ATP dose was evaluated on reference anatomy (ATPCL) and on daily anatomy (ATPDA) after rigid target propagation and re-contouring within 2 cm of the target. ATS was initialized from clinical reference plans (PlanCL) and MCO reference plans (PlanMCO), generating ATSCL and ATSMCO. Online ATS was performed hands-off and timed until all dosimetric goals were achieved or optimization completed. The final plan was considered acceptable if all predefined dosimetric criteria were satisfied.
Results
PlanCL and PlanMCO achieved clinical goals in all patients in both cohorts with comparable dosimetric criteria and modulation factors. In Cohort-A, ATPCL achieved clinical goals in 90% of patients, whereas ATPDA achieved clinical goals in 38%. In Cohort-B, ATPCL and ATPDA were not acceptable in all patients. ATSCL achieved clinical goals in 42% (Cohort-A) and 48% (Cohort-B) of patients, whereas ATSMCO achieved clinical goals in 96% and 93%, respectively. Dose discrepancies in ATSMCO plans were small and primarily attributable to statistical uncertainty. Median(IQR) optimization time for ATPCL versus ATSMCO was 3.1(1.1) versus 2.1(0.9) minutes in Cohort-A, and 1.7(0.6) versus 3.2(1.5) minutes in Cohort-B.
Conclusion
ATP underestimates deviations in dosimetric goals when evaluated on daily anatomy. MCO-based reference planning enables rapid, hands-off ATS that achieves clinical goals within acceptable daily optimization times, making accurate online adaptive prostate SBRT feasible in routine practice.