A Machine Learning Method to Predict Sudden Cardiac Death Using Heart Rate Variability Signals
Open AccessSudden Cardiac Death (SCD) is sudden and unexpected death due to a cardiovascular cause occurring within one hour of symptoms onset. In this research, two combined schemes are proposed to predict SCD by analyzing 16 minutes (at eight successive intervals of 2 minutes) of Heart Rate Variability (HRV) signals from a normal population and subjects at risk of SCD. In the first procedure, the Ensemble Empirical Mode Decomposition (EEMD) approach is implemented on extracted HRV signals for decomposing them into different Intrinsic Mode Functions (IMFs). Subsequently, four entropy parameters, including Distribution Entropy, Sample Entropy, Fuzzy Entropy, and Rényi Entropy, are computed from the first three IMFs obtained. Then, a searching selection approach is applied to identify the best features for classification in feature space. Afterward, the classification is applied separately to each feature; then, the best distinguishing feature is selected based on the value of classification accuracy it returns. The most discriminating feature is thus combined with the other individual features to construct the best group combination. Eventually, these selected features are fed into various classifiers, such as Multilayer Perceptron (MLP), Support Vector Machine (SVM), and K-Nearest Neighbors (kNN), for the classification process. In the second procedure, the Empirical Mode Decomposition (EMD) method is utilized. Compared with well-performing methods, the two proposed schemes can predict subjects at risk of SCD up to 16 minutes earlier. The EEMD-based scheme demonstrates remarkable performance with an accuracy of 94.7%, 90.8%, 92.1%, 96.1%, 85.5%, 89.5%,85.5% and 86.8% for the 1st 2-min to 8th 2-min before SCD onset, respectively. The results illustrate the significant capacity of the proposed method for predicting sudden cardiac death compared with the previous studies and reducing the study's complexity.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.