Mohammad Yusuf Hamadani, Zainul Abidin, M. Fauzan Edy Purnomo
Cardiovascular diseases (CVDs) are leading causes of death worldwide, responsible for over 17.3 million deaths annually. Early detection of heart abnormalities, such as arrhythmia and congestive heart failure (CHF), is crucial in preventing sudden cardiac death (SCD). Although medical advancements have improved treatment outcomes, diagnosing heart conditions through electrocardiograms (ECG) remains complex and often requires expert interpretation. An automated classification system with high accuracy could assist in early detection and improve patient outcomes. This study proposes a two-dimensional Convolutional Neural Network (2D-CNN) for classifying four heart conditions: arrhythmia, CHF, normal sinus rhythm (NSR), and SCD. ECG signals from the PhysioNet database were preprocessed using wavelet transformation for noise reduction before being converted into 2D images. A total of 4,000 ECG images were used, with 3,200 images for training. The proposed 2D-CNN model achieved an accuracy of 97.25%, with precision, recall, and F1-score of 97.32%, 97.25%, and 97.25%, respectively. However, results indicate that while wavelet transformation helps with noise reduction, it may also affect classification accuracy. Further research is needed to assess the signal-to-noise ratio (SNR) after the wavelet transformation process, evaluate the model's performance on larger and more diverse datasets, and validate its effectiveness in real applications. © 2025 IEEE.
Department of Electrical Engineering, Universitas Brawijaya, Malang, Indonesia
