A review of automated diagnosis of ECG Arrhythmia using deep learning methods

Arrhythmia is a medical condition in which the heart's normal pumping process becomes irregular. Early identification of arrhythmia is one of the essential phases in diagnosing the disorder. However, due to the relatively low amplitudes, visually assessing the electrocardiogram signals can also be difficult and time-consuming. Using an automation process from a clinical perspective can significantly expedite and increase the accuracy of diagnosis. Conventional machine learning algorithms have gained significant progress. Such methods depend on customized feature extraction, which requires in-depth knowledge. Deep learning (DL) developments have made it feasible to extract and classify high-level features automatically. This study reviewed recent significant progress in DL approaches for automated arrhythmia diagnosis and some critical areas of the dataset used, the application and category of data input, the modeling architecture, and the performance. Overall, this study provides extensive and detailed knowledge for researchers interested in widening existing knowledge in this area.