Early detection of sudden cardiac death using nonlinear analysis of heart rate variability

Background and objective: Sudden cardiac death (SCD) is one of the most widespread reasons for death around the world. A precise and early prediction of SCD can improve the chance of survival by administering cardiopulmonary resuscitation (CPR). Hence, there is a vital need for an SCD prediction system. Methods: In this work, a novel and efficient algorithm for automated detection of SCD six minutes before its onset is proposed. This algorithm uses features based on the nonlinear modeling of heart rate variability (HRV). In fact, after the extraction of the HRV signals, increment entropy and recurrence quantification analysis-based features are extracted. The one-way ANOVA is applied for the dimension reduction of feature space—this results in lower computational cost. Finally, the distinguishing features are fed to classifiers such as the decision tree, K-nearest neighbor, naive Bayes, and the support vector machine. Results: By using the decision tree classifier we have achieved SCD detection six minutes before its onset with an accuracy, specificity, and sensitivity of 95%. These results demonstrate the superiority of the presented algorithm compared to the existing ones in performance. Conclusions: This study shows that a combination of features based on the nonlinear modeling of HRV, such as laminarity (based on recurrence quantification analysis), and increment entropy leads to early detection of SCD. Choosing the decision tree improves the performance of the algorithm. The results could help in the development of a tool that would allow the detection of cardiac arrest six minutes before its onset.