AI-MED: Agent-Based LLM-Systems for Interdisciplinary Support in Medical Diagnostics

Overview of the AI-MED project

Radiological imaging plays a central role in modern diagnostics, yet accurate decisions often require more than images alone. Additional information such as patient history, laboratory results, and clinical context can be decisive in reaching the correct diagnosis, but current multimodal approaches remain limited in the amount of context they can process. To address this gap, AI-MED aims to develop an agent-based AI system powered by Large Language Models (LLMs) that dynamically integrates medical images with supplementary information on demand.

This project explores whether such agent-driven architectures can provide more nuanced and holistic diagnostic insights compared to conventional AI. A proof of concept focuses on thoracic X-rays, where distinguishing between conditions such as pneumonia and pulmonary congestion requires combining imaging data with laboratory values and patient records. By orchestrating specialized modules for image analysis, laboratory interpretation, and clinical recommendations, AI-MED aims to advance diagnostic precision while ensuring privacy-preserving workflows.

The project unites radiological expertise from the University Hospital RWTH Aachen with computer science strengths in distributed systems, privacy-aware data management, and advanced AI architectures at RWTH Aachen University. Together, the team works toward a secure, efficient, and clinically validated proof of concept, with the long-term vision of extending the approach to additional imaging modalities and clinical use cases.

Involved PIs