Ventricular Arrhythmia Prediction 3-Hours Ahead of Onset for Long-Term ECG Monitoring

Ventricular arrhythmias (VA) are the leading cause of sudden cardiac death. VAs, particularly ventricular tachycardia (VT) and ventricular fibrillation (VF), are potentially life-threatening rhythms that can cause low blood pressure and may lead to sudden cardiac death. The early and accurate prediction of the occurrence of VAs can save the patient's life. The electrocardiogram (ECG) morphologically represents the electrical activity of the heart. In this, work morphological features like interval-based features and the shape of different waves in ECG have been used as the predictor of VA. To predict VA before onset, we developed an approach to use these ECG morphological features-based shallow artificial neural networks (ANN) to predict the heart attacks up to 3 hours before the onset. The proposed approach is validated on the American health association (AHA) ECG database. The sensitivity, specificity, positive predictivity, and accuracy of the VA prediction before its occurrence are 97.5%, 97.3%, 96.2%, and 97.4% respectively.