Medizinische Fakul­tät OWL

The diagnosis of intermittent atrial fibrillation (AF) is often a lengthy process, as the arrhythmia only appears episodically on the ECG. Depending on the severity of the suspicion, patients are invasively given a continuous monitoring device or repeatedly monitored for hours to days via a portable, non-invasive solution that is often perceived as disruptive in everyday life.

An improved diagnostic process could lead to fewer stroke patients suffering another stroke, as the two conditions often occur together. That is why we are cooperating with the University Clinic for Neurology at the Protestant Hospital Bethel, as well as the FIND-AF 2 study.

There are already initial approaches that use Deep Learning to detect signs of atrial fibrillation in inconspicuous ECGs. In this project, we want to build on these approaches, reduce the time to diagnosis and limit the effort to routine screening in the clinic.

The project takes a transdisciplinary view, which will be used to develop the basis for a reliable, responsible AI system. Among other things, the following questions will be pursued:

• Can medically based limitations of the system lead to an improvement of its academic achievements?
• How can doctors understand and question the results of the system?
• How can the results of the system be explained to patients?
• What is the reliability of the system for different populations?
• How can the risk to patients from AI-assisted diagnosis be minimised?