Contribution to the Development of Intelligent Systems for Multi-Lead ECG Signal Classification Using Convolutional Neural Networks

Accurate and timely interpretation of electrocardiogram (ECG) signals is a very important step in detection of heart disease. Attempts for automated ECG classification based on traditional Machine Learning (ML) methods struggle with complex spatiotemporal nature of multi-lead ECG. In this work, we explore the potential of Convolutional Neural Networks (CNNs) for ECG signals’ classification. The study combines pre-processing methods and a multi-dimensional data-optimized CNN architecture to improve and automatize arrhythmia identification. The results show that our model can successfully distinguish between normal heart rhythm and ventricular arrhythmia with an accuracy of 98.56%, thus transcending the limitations of traditional ML methods. The proposed system has also low runtime, which indicates its potential for real-time clinical deployment.