Exploring the Efficacy of Deep Residual Networks and CNN-ELM for Arrhythmia Classification in Lead II ECG

The increasing demand for portable ECG devices stems from their lightweight and user-friendly nature, offering a convenient alternative to traditional hospital-grade ECG systems. This trend is particularly noticeable among fitness enthusiasts and health-conscious individuals who seek wearable ECG monitors integrated into smartwatches or patches. These devices require effective signal acquisition methods, and a key reason for the widespread use of Lead II in portable ECG devices is its alignment with the natural electrical axis of the heart. Lead II provides a strong, clean, and interpretable signal, making it ideal for accurate ECG monitoring in wearable devices. In this study, we compare two deep neural networks to classify arrhythmias from Lead II ECG signals. Specifically, we assess the performance of a 1D deep residual network and a CNN-ELM model. Our results demonstrate the effectiveness of these models of five groups of arrhythmia detection in Lead II from MIT-BIH datasets, offering a promising approach for healthcare applications in wearable technologies.