Convolutional Neural Networks Architectures for Heartbeat Arrhythmia Classification

The heart is a vital organ that ensures blood circulation within the whole human body through cardiac pulsations; its functioning is characterized by a rhythm describing its beating frequency. This heart rhythm can be influenced by many factors, such as the daily activities and the feelings experienced, and it can be used for diagnosing good cardiac activity as well as it can be considered as an alarm signal. In accordance with the World Health Organization, cardiovascular diseases of which heart arrhythmia is a part are the major cause of death worldwide. Despite the advancement of diagnostic techniques for cardiac arrhythmia, several deaths still occur every year, especially in low and middle-income countries which requires a particular focus from health professionals to develop a preventive diagnostic process easily accessible to all patients. The present study is based on a dataset composed of 404 wav audio files containing the recording of multiple heartbeats. By using various signal transformation, pre-processing and deep learning techniques we have implemented several models allowing the classification of new cardiac recordings. The majority of these models have shown promising achievements in terms of accuracy and speed of execution, which can facilitate the early detection of cardiac pathologies and contribute to life-saving for patients.