A Deep Learning Approach to Multi-Class ECG Interpretation

The application of artificial intelligence and deep learning in cardiology has great potential for diagnostics and prevention of cardiovascular diseases. This work contributes to the development and implementation of intelligent electrocardiography (ECG) interpretation systems using Convolutional Neural Networks (CNNs). We applied a series of data filtering and preprocessing techniques to improve the performance of the proposed model. Based on predefined metrics, our model shows high performance in distinguishing between normal ECG rhythm (N) and various types of arrhythmias, such as ventricular (V), supraventricular (S), fusion (F), and those of unknown origin (Q). The accuracy of 96.19%, with F1 score of 0.67 and specificity of 0.99, demonstrates the high effectiveness of the implemented methodology. It also highlights the significant potential of CNN in diagnostics of cardiac diseases in general.