Beyond the Individual: Virtualizing Multi-Disciplinary Reasoning for Clinical Intake via Collaborative Agents

A research team from Zhejiang University and collaborating institutions has introduced Aegle, a novel framework designed to bring multi-disciplinary team (MDT) reasoning to real-time clinical consultations. The system addresses a critical bottleneck in healthcare: initial outpatient visits are often conducted by a single, time-pressured physician, leading to potential cognitive biases and incomplete evidence gathering. While traditional MDTs mitigate these issues, they are resource-intensive and difficult to scale. Aegle virtualizes this process using a graph-based multi-agent architecture that formalizes the consultation state into a structured SOAP (Subjective, Objective, Assessment, Plan) representation, separating evidence collection from diagnostic reasoning for better traceability.

The core of Aegle's operation involves an orchestrator that dynamically activates various specialist AI agents—simulating different medical experts—to perform decoupled, parallel reasoning on a patient's case. Their outputs are then integrated by an aggregator into a coherent and comprehensive clinical note. In rigorous experiments on the ClinicalBench and a real-world RAPID-IPN dataset spanning 24 clinical departments, Aegle was evaluated across 53 metrics. The results showed it consistently outperformed current state-of-the-art proprietary and open-source AI models, not only in the quality of clinical documentation and consultation capability but also in improving the accuracy of the final diagnosis. This demonstrates a significant step towards scalable, AI-augmented clinical decision support.