Comparative Study of Deep Learning Models to Classify Arrhythmia

An irregular heart rhythm that can result from various heart-related conditions are arrhythmia. Globally an estimate of 3% to 5% of population is affected by arrhythmia. If it is not diagnosed and treated quickly, it can lead to sudden cardiac death (SCD). Hence early detection of SCD is essential for patient safety. Detection of arrhythmia has improved with the recent advancements in deep learning (DL). This research paper examines arrhythmia classification based on the AAMI EC57 standard, using several prominent deep learning models applied to the lead II Electrocardiogram heartbeat data from the well-known MIT- BIH Arrhythmia dataset. The models trained and tested in this research work are Artificial Neural Network (ANN), Long Short- Term Memory (LSTM), Convolutional Neural Network (CNN), Residual Network (ResNet), and Temporal Convolution Network (TCN). The results showed that both the ResNet and TCN models performed the best, achieving an impressive accuracy of 96% and precision of 97% and 98% respectively. These models could be further fine-tuned to achieve better generalization and improve performance