Stefan Marciniak

AI-enabled diagnosis of Birt-Hogg-Dubé syndrome

This project develops a federated-learning AI system to improve CT-based diagnosis of Birt-Hogg-Dubé syndrome using multi-institutional data without sharing patient information. By combining transformer-based imaging models with radiomics and molecular data, the approach aims to overcome data scarcity inherent in rare disease research. The goal is to improve diagnostic accuracy in non-specialist settings and establish a scalable, privacy-preserving framework that can be extended to other rare diseases.