Classification of ECG signals using deep neural networks

Abstract

The electrocardiogram (ECG) is an essential tool in the field of cardiology, as it enables the electrical activity of the heart to be measured. It involves placing electrodes on the patient's skin, facilitating the measurement and analysis of cardiac rhythms. This non-invasive and painless test provides essential information about the heart's function and helps in diagnosing various cardiac conditions. The classification of ECG signals using deep learning techniques has garnered substantial interest in recent years; ECG classification tasks have exhibited promising outcomes with the application of deep learning models, particularly convolutional neural networks (CNNs). The GoogleNet, AlexNet, and ResNet Deep-CNN models are proposed in this study as reliable methods for accurately diagnosing and classifying cardiac diseases using ECG data. The primary objective of these models is to predict and classify prevalent cardiac ailments, encompassing arrhythmia (ARR), congestive heart failure (CHF), and normal sinus rhythm (NSR). To achieve this classification, 2D Scalogram images obtained through the continuous wavelet transform (CWT) are utilized as input for the models. The study's findings demonstrate that the GoogleNet, AlexNet and Resnet models achieve an impressive accuracy rate of 96%, 95,33% and 92,66%, in accurately predicting and classifying ECG signals associated with these cardiac conditions, respectively. Overall, the integration of deep learning techniques, such as the GoogleNet, AlexNet, and ResNet models, in ECG analysis holds promise for enhancing the accuracy and efficiency of diagnosing and classifying cardiac diseases, potentially leading to improved patient care and outcomes.