Sudden cardiac death prediction based on the complete ensemble empirical mode decomposition method and a machine learning strategy by using ECG signals

Abstract

Listen

Translate article

Cardiovascular diseases are a significant global health problem, often culminating in sudden cardiac death (SCD). Approximately 4 million annual deaths worldwide are attributed to SCD. Although the causes are varied, the heart’s electrical activity is always affected prior to an SCD event. This work explores the capabilities of the complete ensemble empirical mode decomposition (CEEMD) method for extracting relevant time-domain features from electrocardiogram (ECG) signals. CEEMD method is compared with EMD and EEMD methods. Additionally, a machine learning approach incorporating statistical indices, the Kruskal-Wallis test, and a support vector machine (SVM) is utilized for automatic analysis and prediction. SVM results are benchmarked against k-nearest neighbors, decision trees, and neural networks. Results demonstrate that the proposed methodology can predict the SCD event 30 min before it occurs with an accuracy of 97.28%, positively impacting on the improvement of timely interventions and, consequently, the survival rates and quality of life of people.