ECG Signal Classification Using MODWT and CNN for Early Detection of Cardiac Abnormalities

Accurate classification of ECG signals is crucial for the early detection of cardiac abnormalities. This study proposes a method that integrates Maximal Overlap Discrete Wavelet Transform (MODWT) for feature extraction with a Convolutional Neural Network (CNN) to enhance classification performance. Unlike previous research that applied a custom 1D CNN directly to raw ECG signals, this approach preprocesses the data using MODWT to extract QRS complex features, improving the model’s ability to distinguish between different heart conditions. The classification includes four categories: Atrial Fibrillation (AF), Congestive Heart Failure (CHF), Normal Sinus Rhythm (NSR), and Ventricular Fibrillation (VF). The model’s performance was evaluated using key metrics, achieving precision of 96.96%, recall of 96.88%, F1-score of 96.86% and an accuracy of 98.75% on test data and 96.88%, on validation data. These results indicate that the proposed approach provides competitive classification performance, demonstrating the potential of combining wavelet transform and deep learning techniques to support ECG-based cardiac abnormality detection and diagnosis.