Identification of atrial fibrillation predictors on an electrocardiogram using a neural network

Atrial fibrillation (AF) is a common rhythm disorder, a life-threatening complication of which is cardioembolic stroke leading to disability and death. This necessitates the search for early predictors of this pathology. P wave and PR interval abnormalities on electrocardiography (ECG) are associated with the AF risk. Neural networks are considered for rapid ECG analysis in routine practice and identifying the risks of AF occurrence and/or relapse. In recent years, advances in joint projects between medicine and artificial intelligence have made significant progress in the use of open ECG databases for deep machine learning of neural networks. These studies have shown that artificial intelligence makes it possible to identify predictors of AF, which will significantly reduce the risk of mortality due to thromboembolism. This paper reviews in detail the results of published studies that highlight the effectiveness of neural networks to improve AF risk assessment.