AI Insight
This study introduces OpenGray, a benchmark for evaluating AI-based radiotherapy planning agents on their clinical decision-making rather than just dose quality. Testing four systems on 30 head-and-neck cancer patients revealed that even high-performing agents missed critical planning contradictions 50-75% of the time and lacked mechanisms to halt planning when goals were unachievable. The benchmark exposes gaps between aggregate performance metrics and specific decision failures that could affect patient safety.
Why it matters
As large language models are increasingly deployed in clinical radiotherapy planning, this work demonstrates that evaluating only dose outcomes is insufficient—agents must also be assessed on their ability to recognize impossible constraints and appropriately escalate cases for human review. The findings suggest current AI planning systems may require explicit "stop and review" actions to prevent unsafe clinical decisions.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
Abstract: Background and purpose: Radiotherapy planning agents based on large language models (LLMs) decide when to submit a plan or request review but are evaluated mainly by dose. This study developed OpenGray, a benchmark of planning tasks with known correct decisions. Materials and methods: OpenGray adds tasks with known answers to the OpenKBP-Opt head-and-neck dataset, including two contradictions with dose goals proven incompatible and a feasible control with a stored plan meeting every hard goal. Four systems, from a rule-based planning script to a full LLM agent, were evaluated on 30 patients, the LLM-based ones with a commercial (Terra) and a local (Ministral 3 8B) LLM. Decisions, achieved dose, problem recognition, and technical failures were recorded and also read from logs of 39 stereotactic radiosurgery runs of SAGE, a clinical LLM agent. Results: With Terra, the script preceded by an LLM review and the full agent had decision accuracies of 95.4% and 92.1%, respectively, yet escalated only 50% (95% confidence interval, 25-75%) and 25% (10-40%) of overlap contradictions, which required arithmetic on the anatomy. All four systems submitted every feasible control, but 24% of these plans missed a hard goal that the stored plan had met. In four SAGE runs, one goal failed at every attempt; the agent identified it in every reply but had no action for stopping. Conclusions: Favorable aggregate scores concealed specific decision failures. Planning agents should be evaluated on their decisions as well as their dose and given an explicit action for stopping and requesting review.
Source: OpenGray: a benchmark for evaluating the decisions of radiotherapy planning agents