AI in Cardiology
Cardiovascular Risk Prediction Through Intelligent ECG Analysis

When it comes to the early detection of cardiovascular diseases, there is a "diagnostic gap": people with elevated risk factors who do not exhibit any symptoms are often not identified as high-risk patients early on. However, optimized screening could improve the prevention of cardiovascular diseases. This is precisely where the "AI Cardiology" project steps in.

The goal is to develop an artificial intelligence (AI)-based ECG analysis for detecting cardiovascular risk factors in 12-lead and single-lead ECGs, in order to enable such effective screenings for early detection of cardiovascular risk.

The Approach: From Clinic to Everyday Life

The project follows a two-step plan to further develop cardiovascular prevention:

1. In the first step, AI models will be researched and developed based on 12-lead ECGs to identify relevant cardiovascular biomarkers and risk factors directly from ECG data. These include laboratory parameters such as HbA1c (long-term blood glucose measurement), LDL cholesterol, and systolic blood pressure. With current knowledge, it is not possible for medical personnel to derive such values solely from an ECG. Using AI methods, the not yet fully explored wealth of information of ECGs can now be utilized, thereby expanding their diagnostic application.
2. Scalability through wearables: In a second step, this approach will be adapted for use with single-lead ECGs, such as those found in smartwatches or other wearables. This would enable a cost-effective and easily accessible continuous screening in people´s daily lives – long before they need to visit a clinic.

As with any other project in this field, AI-driven decisions in medicine must be trustworthy. The role of the Fraunhofer Institute for Cognitive Systems IKS in the project is to design, develop, and evaluate the AI models. This reflects two of the institute´s core competencies

• Robustness: The focus is on ensuring that AI functions reliably even when dealing with varying data quality and external datasets.
• Explainable AI (XAI): To ensure that medical professionals can trust the results, researchers at Fraunhofer IKS are using explainable AI techniques. These methods intend to provide transparency on why an AI-model arrives at a particular prediction. The goal is to transform AI from a "black-box" into a comprehensible assistance toolfor doctors.

Through close cooperation with the University Heart Center Freiburg-Bad Krozingen, it is ensured that research is driven by clinical practice and reaches the people who need it most: the patients.

The "AI Cardiology" project demonstrates how research initiatives can improve cardiovascular prevention for the general population and strengthen medical research in Germany. In this regard, we would like to extend our special appreciation to the Hector Foundation for its valuable support of our work.