ECG Heartbeats Classification Using Two-Dimensional Deep Learning Convolutional Neural Network

The prevalence of heart-related illnesses, particularly arrhythmias, has surged due to lifestyle factors, making accurate diagnosis crucial. This article delves into the transformative potential of deep learning in ECG signal classification, addressing the challenges posed by noisy, multi-frequency signals. By integrating continuous wavelet transform (CWT) with 2D convolutional neural networks (2D-CNN), the proposed method converts one-dimensional ECG signals into two-dimensional scalograms, enhancing feature extraction and classification robustness. The study utilizes the MIT-BIH Arrhythmia Database, focusing on five types of arrhythmias, and employs a 20-layer 2D-CNN architecture to achieve adaptive, precise, and noise-resilient classification. Experimental results demonstrate superior performance metrics, including an overall accuracy of 99.59%, highlighting the method's potential as a supplementary clinical diagnostic tool. The article also discusses the implications for future research, including the integration of additional ECG features, optimization of feature selection, and validation across diverse patient populations.