Modified Cardiac Arrhythmia Classification from Electrocardiography Signals Using a Convolutional Neural Network Model

Manual classification of cardiac arrhythmias from electrocardiogram (ECG) signals is a labor-intensive and error-prone process due to the complex and variable nature of cardiac waveforms. Convolutional Neural Networks (ConvNets), widely recognized for their success in image classification, offer a promising solution for automating this task. This study proposes an enhanced ConvNet-based approach for the classification of cardiac arrhythmias, leveraging AlexNet as a feature extractor. The features obtained from the convolutional layers are input into a backpropagation neural network for final classification. The proposed model was evaluated on four distinct arrhythmia conditions using ECG waveforms from the MIT-BIH Arrhythmia Database. Comparative analysis against traditional models revealed the superior performance of the proposed ConvNet architecture, achieving high scores across multiple evaluation metrics, including accuracy, precision, recall, F1-score, and AUC-ROC. The feature extractor demonstrated robust performance, with classification accuracies of 1.00 and 0.99 on training and testing datasets, respectively. These findings underscore the potential of ConvNet-based models to serve as efficient, accurate, and fully automated tools for arrhythmia diagnosis, contributing significantly to advancements in cardiovascular disease detection and clinical decision support systems.