RadOnc-Agent: LLM-Orchestrated Framework Streamlines AI Workflows Across Radiotherapy Care Pathway
agents
| Source: ArXiv | Original article
A new arXiv preprint introduces RadOnc-Agent, an LLM-orchestrated framework that integrates AI across the entire radiotherapy care pathway, addressing fragmented tasks and data.
A new arXiv pre‑print (arXiv:2610.06923v1) unveils **RadOnc‑Agent**, an agentic framework that uses large language models (LLMs) to coordinate the disparate AI tools that currently support individual steps of radiotherapy. The authors formalise the entire treatment pathway into four distinct clinical phases and demonstrate that the LLM‑orchestrated system can execute multi‑stage workflows on a cohort of 60 patient records with high success rates.
The contribution matters because AI in radiation oncology has largely remained siloed—image segmentation, dose calculation, toxicity prediction and follow‑up analytics are each handled by separate models and software stacks. By stitching these components together, RadOnc‑Agent promises a more coherent, longitudinal view of patient care. Clinicians could see AI‑generated recommendations that respect the chronological order of diagnosis, planning, delivery and post‑treatment monitoring, potentially reducing manual hand‑offs and accelerating decision‑making.
The paper also releases code on GitHub under the RadOnc‑AI organization, signalling an open‑science approach that may speed adoption in research hospitals. The next steps to watch include external validation on larger, multi‑institutional datasets, integration with commercial treatment planning systems, and any regulatory pathways that might be pursued for clinical deployment. As the oncology community continues to explore LLM‑driven orchestration—mirroring trends seen in multi‑agent frameworks such as AegisFlow—RadOnc‑Agent could become a reference point for end‑to‑end AI assistance across the radiotherapy care continuum.
Sources
Back to AIPULSEN