Learning to Optimize Radiotherapy Plans via Fluence Maps Diffusion Model Generation and LSTM-based Optimization

2026-05-13

AI Summary

Volumetric modulated arc therapy (VMAT) planning is typically time-consuming due to repeated re-optimization and the need to balance multi-objective consistency with mechanical deliverability. This work proposes an end-to-end VMAT planning framework that uniquely integrates a distilled diffusion model with an LSTM-driven learning-to-optimize (L2O) architecture: the diffusion model generates high-quality, deliverable fluence maps in a single forward pass, while the L2O component rapidly converges toward dose objectives through learned optimization dynamics. Evaluated on both clinical and publicly available prostate cancer datasets, the proposed method significantly enhances planning efficiency, flexibility, and machine deliverability, outperforming existing end-to-end VMAT approaches.

Abstract

Volumetric Modulated Arc Therapy (VMAT) is a cornerstone of modern radiation therapy, enabling highly conformal tumor irradiation and healthy-tissue sparing. Yet, its planning solves inverse and nested optimization for multi-leaf collimators, monitor units and dose parameters, while enforcing their consistency to ensure mechanical deliverability. Nevertheless, this process often requires repeated re-optimization when treatment configurations change, resulting in substantial planning time per patient. To address these problems, we present a diffusion-driven Learning-to-Optimize (L2O) method for end-to-end VMAT planning. A distribution-matching distilled diffusion model learns a clinically feasible manifold of fluence maps, enabling their one-shot generation. On top of this, an LSTM-based L2O module learns gradient update dynamics to swiftly refine fluence maps toward prescribed dose objectives during inference. Experimental results on clinical and public prostate cancer cohorts demonstrate improved planning efficiency, flexibility, and machine deliverability over currently available end-to-end VMAT planners.

Problem

• VMAT planning
• re-optimization
• treatment configuration changes
• planning efficiency
• mechanical deliverability

Innovation

• diffusion model
• Learning-to-Optimize
• fluence map generation
• LSTM-based optimization
• VMAT planning