DZL researchers at the Heidelberg, Berlin and Greifswald sites have used a comprehensive MRI image database to train an AI-assisted method that enables precise analyses of X-ray images in people with cystic fibrosis. The development is based on an imaging dataset compiled over many years, predominantly at the DZL site in Heidelberg, which represents a particularly valuable resource for this rare disease. The method could help make the assessment of relevant changes in the lungs faster and more widely available, including in regional and international healthcare systems where access to complex MRI examinations is limited.

Close interdisciplinary collaboration between radiological expertise and medical informatics was crucial to this advance: while the radiological evaluation of disease-related changes and imaging data provided the clinical foundation, the medical informatics-based preparation and processing of the data enabled the development of the AI-based analytical method. The AI was trained using precise MRI results to identify hidden patterns in X-ray images that correlate with the actual condition of the lungs. In this way, the automated Deep Chest X-Ray Score can reliably assess changes in the lungs based on a rapidly available X-ray image.

Research Letter:

Shengkai Zhao, Lena Wucherpfennig, Yao Kou, Simon M F Triphan, Friedemann G Ringwald, Marcus A Mall, Mirjam Stahl, Olaf Sommerburg, Urs Eisenmann, Petra Knaup-Gregori, Mark O Wielpütz, Artificial Intelligence Improves Chest X-ray Interpretation Employing Magnetic Resonance Imaging as Ground Truth in Patients with Cystic Fibrosis, American Journal of Respiratory and Critical Care Medicine, 2026 aamag146, https://doi.org/10.1093/ajrccm/aamag146