Automated stereotactic radiosurgery planning using a human-in-the-loop reasoning large language model agent
🤖 AI Summary
The clinical adoption of black-box AI in stereotactic radiosurgery (SRS) is hindered by its opacity and lack of interpretability. Method: We propose SAGE—the first large language model (LLM)-based agent system integrating chain-of-thought (CoT) reasoning for fully automated single-fraction 18-Gy SRS planning for brain metastases. SAGE employs dose-constraint modeling, prospective constraint validation, multi-objective trade-off analysis, and a human-in-the-loop feedback loop to enable auditable, traceable, human-like reasoning and generate comprehensive optimization logs. Results: Experimental evaluation shows no statistically significant difference between SAGE and manual plans for key dosimetric metrics—including PTV coverage and maximum dose (p > 0.21)—while significantly reducing cochlear dose (p = 0.022). The system executed 457 constraint validations and 609 trade-off analyses, directly addressing the core clinical trust barrier for AI in radiotherapy planning.
Application Category
📝 Abstract
Stereotactic radiosurgery (SRS) demands precise dose shaping around critical structures, yet black-box AI systems have limited clinical adoption due to opacity concerns. We tested whether chain-of-thought reasoning improves agentic planning in a retrospective cohort of 41 patients with brain metastases treated with 18 Gy single-fraction SRS. We developed SAGE (Secure Agent for Generative Dose Expertise), an LLM-based planning agent for automated SRS treatment planning. Two variants generated plans for each case: one using a non-reasoning model, one using a reasoning model. The reasoning variant showed comparable plan dosimetry relative to human planners on primary endpoints (PTV coverage, maximum dose, conformity index, gradient index; all p > 0.21) while reducing cochlear dose below human baselines (p = 0.022). When prompted to improve conformity, the reasoning model demonstrated systematic planning behaviors including prospective constraint verification (457 instances) and trade-off deliberation (609 instances), while the standard model exhibited none of these deliberative processes (0 and 7 instances, respectively). Content analysis revealed that constraint verification and causal explanation concentrated in the reasoning agent. The optimization traces serve as auditable logs, offering a path toward transparent automated planning.
Problem
Research questions and friction points this paper is trying to address.
Automates stereotactic radiosurgery planning with AI transparency
Compares reasoning vs non-reasoning AI models for dose optimization
Generates auditable logs to improve clinical trust in automation
Innovation
Methods, ideas, or system contributions that make the work stand out.
LLM agent with chain-of-thought reasoning for planning
Automated SRS planning with auditable optimization logs
Prospective constraint verification and trade-off deliberation
🔎 Similar Papers
No similar papers found.
