International Journal of Advanced Research in Computer and Communication Engineering

Abstract: Cardiac arrhythmia is a common cardiovascular disorder that results from abnormalities in the electrical conduction system of the heart, leading to irregular heartbeat patterns. Accurate and timely detection of arrhythmia is crucial for effective diagnosis and treatment, yet manual interpretation of electrocardiogram (ECG) signals remains a challenging and time-consuming process due to the complex, dynamic, and non-stationary nature of ECG data. This study proposes a robust automated deep learning framework for the classification of ECG signals into three clinically significant categories: cardiac arrhythmia (ARR), congestive heart failure (CHF), and normal sinus rhythm (NSR). ECG data were obtained from the publicly available MIT-BIH and BIDMC databases on PhysioNet and underwent a comprehensive preprocessing pipeline that included noise removal, normalization, and segmentation to ensure data quality and consistency. Two pretrained convolutional neural network architectures, ResNet-50 and AlexNet, were fine-tuned using transfer learning techniques to leverage their deep hierarchical feature extraction capabilities for ECG classification. The models were trained and validated using a stratified dataset, and their performance was assessed through a multi-class confusion matrix employing evaluation metrics such as accuracy, precision, recall, sensitivity, specificity, and F-measure. Experimental results demonstrated that the proposed deep learning model achieved outstanding performance with an overall classification accuracy of 99.2%, average sensitivity of 99.2%, specificity of 99.6%, and precision, recall, and F-measure all at 99.2%. These results indicate that the model can effectively differentiate between normal and pathological cardiac conditions with high reliability. In conclusion, the proposed system offers a powerful and efficient tool for automated arrhythmia detection, significantly reducing diagnostic time and minimizing errors associated with manual ECG interpretation, thereby supporting clinicians in the rapid and accurate diagnosis of cardiac disorders.

Keywords: Electrocardiogram (ECG), Deep learning (or deep neural network), Convolutional Neural Network (CNN) model, ARRHYTHMIAS, accuracy, Time Frequency Representations, ResNet50, AlexNet and Morse Wavelet.

Downloads: | DOI: 10.17148/IJARCCE.2025.141145