Deep Learning Based Cardiac Arrhythmia Detection Using Wavelet Transform

Electrocardiogram (ECG) plays a crucial role in mortality rate reduction caused by Cardiovascular Diseases (CVDs). However, the accuracy and speed of computer-aided diagnostic systems related to cardiac arrhythmia diagnosis using ECG signals remain challenges such as the presence of noise, and instability. Correct and prompt diagnosis of CVD is vital for the survival and well-being of patients, contributing significantly to their treatment and recovery. In this research, aiming for precise diagnosis of various CVDs based on the condition of arrhythmias on ECG rhythms, we implement an automatic ECG rhythm classification using discrete wavelet transform, followed by optimizing the window setting to segment the data into beats. Different deep learning algorithms namely Convolutional Neural Network (CNN), Recurrent Neural Network, and Long Short-Term Memory network layers are considered and optimized for the classification of CVDs. CNN is selected as the proposed method with relatively high-performance results including an accuracy of 99.44%, Precision of 99.40%, Recall of 99.42%, F1-score achieved 99.44%, and AUC of 99.98%. Our proposed classification system can serve as an automated diagnostic tool for cardiac arrhythmias, aiding CVDs experts in the initial screening of patients with rhythm disorders.