A Multi-Agent Assistant for Medical Diagnosis: Blood-Test Case Study

Abstract

This project explores the use of Multi-Agent Systems and Federated Learning to automate the diagnosis of blood test results, with a focus on complete blood count analyses. A modular system of autonomous agents was designed to preprocess data, extract features, train federated models, and provide decision support. The model achieved high diagnostic accuracy on both small and large public datasets, demonstrating the potential for privacy-preserving, decentralized medical diagnosis. The project highlights opportunities for future development and real-world integration within healthcare systems.