BrachyPlan: A fine-grained efficient dose-guided inverse planning strategy for low-dose-rate brachytherapy.

Abstract

Brachytherapy delivers highly conformal doses for malignancies ranging from pancreatic to head-and-neck cancers, yet today's treatment-planning systems still depend on extensive manual manipulation and dose engines of uncertain accuracy. We introduce BrachyPlan, an end-to-end pre-operative framework that automates low-dose-rate (LDR) brachytherapy planning while preserving clinical precision. The system combines: 1) a real-time Monte Carlo dose engine that produces high-fidelity, heterogeneity-aware dose maps, and 2) a dose-guided inverse-planning algorithm that optimally places seeds to meet target coverage and organ-at-risk constraints. In retrospective experiments, BrachyPlan reproduced ground-truth Monte Carlo dose-volume histograms (DVHs) with <5% error across all metrics, achieved V100 >95% for every clinical target, and cut planning time to one-eighth of the manual workflow, with minimal manual intervention and optimal efficiency. To our knowledge, this is the first platform that unites a clinically validated dose calculation algorithm with real-time inverse optimization, delivering both sub-5% dosimetric accuracy and unprecedented efficiency for LDR brachytherapy.