Multi-Center and Federated AI-Driven ECG Analysis for Precision Cardiology

Atrial fibrillation (AF) is the most common heart rhythm disorder and is closely linked with atrial cardiomyopathy (AtCM), a condition where the upper heart chambers undergo harmful changes. These two conditions influence each other and increase the risk of stroke and heart failure. New evidence suggests that it might be AtCM - not AF alone - that primarily drives the stroke risk.

Current tools to detect AtCM are expensive, invasive, and not widely available. This project aims to develop and evaluate a more accessible method using electrocardiograms (ECGs). Specifically, the project focuses on a new ECG marker called amplified P-wave duration (APWD), which better captures electrical signals in the heart’s atria and has already shown promise in predicting AF, stroke, and arrhythmia recurrence after treatment.

The project brings together clinical teams from Basel and Freiburg and a technical team from Karlsruhe. Goals include developing an automated APWD measurement tool, evaluating how well APWD predicts heart and brain outcomes, and comparing simpler, explainable models to advanced AI-based ECG analysis. The teams will also create a federated learning platform that lets hospitals securely train AI models on ECG data without sharing patient records across borders.

The expected outcomes are a validated, AI-powered ECG marker for heart and brain risk and a legal, scalable platform for future clinical AI research-laying the foundation for larger follow-up projects.