Explainable FA detection in single-lead signals acquired from portable smart-enabled ECG devices

Abstract

Atrial fibrillation is a cardiac condition characterized by an irregular and disorganized heart rhythm, which can be observed on an electrocardiogram (ECG) through a distinctive pattern recognizable by a cardiology specialist. Unlike a normal, coordinated heart rhythm, in atrial fibrillation the atria beat chaotically, increasing the risk of blood clots forming in the heart and leading to serious complications such as stroke. To address this issue, it is crucial to develop highly accurate and sensitive automated AF screening methods. However, many current AI-based approaches lack transparency, making it difficult to understand the underlying reasons behind screening decisions. This work employed explainable artificial intelligence (XAI) to identify AF in a more understandable and detailed way. This approach made it possible to expose the main physiological parameters that influence the detection of positive AF cases. The results obtained provided relevant information on the biological mechanisms involved while offering clear explanations that can help medical professionals better understand the decision-making process in the diagnosis of this cardiac condition.