Multi-Modal Multi-Agent Reinforcement Learning for Radiology Report Generation: Radiologist-Like Workflow with Clinically Verifiable Rewards
This work addresses the problem of generating accurate and clinically relevant radiology reports for medical professionals, representing an incremental advance by building on prior reinforcement learning methods with a novel multi-agent approach.
The paper tackled radiology report generation by proposing a multi-modal multi-agent reinforcement learning framework that coordinates region-specific and global agents, optimized with clinically verifiable rewards. It achieved state-of-the-art clinically efficacy performance on MIMIC-CXR and IU X-ray datasets, improving metrics like RadGraph, CheXbert, and GREEN scores.
We propose MARL-Rad, a novel multi-modal multi-agent reinforcement learning framework for radiology report generation that coordinates region-specific agents and a global integrating agent, optimized via clinically verifiable rewards. Unlike prior single-model reinforcement learning or post-hoc agentization of independently trained models, our method jointly trains multiple agents and optimizes the entire agent system through reinforcement learning. Experiments on the MIMIC-CXR and IU X-ray datasets show that MARL-Rad consistently improves clinically efficacy (CE) metrics such as RadGraph, CheXbert, and GREEN scores, achieving state-of-the-art CE performance. Further analyses confirm that MARL-Rad enhances laterality consistency and produces more accurate, detail-informed reports.