Autonomous radiotherapy planning via agentic orchestration using a multimodal TPS-integrated compound AI platform

Abstract

Purpose. Radiation therapy (RT) treatment planning requires iterative, multi-day optimization workflows in which subjective planning strategies produce inter-planner variability in plan quality. Existing computational approaches automate isolated aspects of this workflow, yet none orchestrates an end-to-end pipeline from physician directive to deliverable plan. We developed a compound artificial intelligence (AI) platform for fully autonomous RT treatment planning that combines multi-agent large language model (LLM) orchestration with directive-conditioned three-dimensional (3D) dose prediction, natively integrated with a commercial treatment planning system (TPS). Methods. Seven specialized agents navigated the multi-objective optimization landscape through structured clinical reasoning, iteratively analyzing dose-volume histogram (DVH) metrics and spatial dose patterns, formulating trade-off strategies, and executing validated modifications through the TPS across five fully autonomous iterations per case. A directive-conditioned 3D dose prediction model supplied patient-specific DVH values from which initial optimization objectives were autonomously derived, eliminating the need for curated templates or manual initialization. A retrieval-augmented generation (RAG) system encoded institutional knowledge into the planning workflow. We evaluated 60 retrospective cases across brain, lung, and prostate sites, with 10 intensity-modulated RT (IMRT) and 10 volumetric modulated arc therapy (VMAT) plans per site spanning 20.0–79.2 Gy in 3–44 fractions, scored by the proportion of dosimetric criteria satisfied. Results. Across all 60 cases, AI plans achieved 89.8 ±9.4% of dosimetric criteria versus 85.2 ±10.8% for clinical reference plans ( p<0.001). IMRT plans improved in 25 of 30 cases with none worsened (94.1 ±6.7% vs 84.3 ±11.8%, p<0.001); VMAT plans showed no significant difference (85.6 ±9.9% vs 86.1 ±9.7%, p=0.770). Each plan iteration completed in 20.2±12.7 min, of which agent reasoning consumed 5.2±1.7 min ( 114,429±11,798 tokens, 0.43±0.04). Conclusions. These results established the feasibility of end-to-end, fully autonomous, universal RT treatment planning through compound AI. Integrating dose prediction as an agent-invoked tool for objective initialization resolved the dependency on curated templates and manual specification that constrained prior LLM-based planning systems.