Title:Compact Neural Network Algorithm for Electrocardiogram Classification

Abstract:In this paper, we present a high-performance, compact electrocardiogram (ECG)-based system for automatic classification of arrhythmias, integrating machine learning approaches to achieve robust cardiac diagnostics. Our method combines a compact artificial neural network with feature enhancement techniques, including mathematical transformations, signal analysis and data extraction algorithms, to capture both morphological and time-frequency features from ECG signals. A novel aspect of this work is the addition of 17 newly engineered features, which complement the algorithm's capability to extract significant data and physiological patterns from the ECG signal. This combination enables the classifier to detect multiple arrhythmia types, such as atrial fibrillation, sinus tachycardia, ventricular flutter, and other common arrhythmic disorders. The system achieves an accuracy of 97.36% on the MIT-BIH arrhythmia database, using a lower complexity compared to state-of-the-art models. This compact tool shows potential for clinical deployment, as well as adaptation for portable devices in long-term cardiac health monitoring applications.