Researchers developed an agent to coordinate radiotherapy tasks

Researchers recently introduced RadOnc-Agent, a system designed to help coordinate the complex steps involved in radiotherapy. Radiotherapy often involves many stages, from the initial decision to treatment planning and follow-up, which are currently handled in disconnected digital environments.

This new framework acts as a digital coordinator. It uses a large language model to understand clinical intent and translates it into specific actions across 26 different digital tools. Think of it as a central hub that keeps track of the patient's information as it moves between phases.

In testing, the system performed well. It managed to select the correct action in nearly 99% of single-task requests. When handling more complex scenarios that spanned different stages of care, it completed over 96% of the tasks successfully.

The researchers used 60 real patient records to verify the system's performance in a controlled environment. The results showed that keeping a consistent track of the patient's state throughout the process was crucial for success.

When the team removed this ability to remember the longitudinal state of the treatment, the success rate for complex workflows dropped significantly to 84%. This demonstrates that coordinating information across the entire clinical path is just as important as the individual tasks themselves.

If you were a clinician managing a patient's treatment plan, would you trust an AI agent to route your requests between different software systems?

It is important to note that this research only confirms technical feasibility. The study does not prove that the system is clinically correct, safe for use, or that it provides any actual medical benefit in a real-world hospital setting.