Sudden cardiac death prediction based on a nonlinear estimation

Abstract

Sudden Cardiac Death (SCD) associated with ventricular tachyarrhythmia, is one of the main causes of death worldwide. The Left Ventricle Ejection Fraction (LVEF) is the predictive parameter used in clinical practice for the stratification of risk of SCD. However, this has a poor predictive value. Several authors have proposed methods seeking to estimate characteristics that can be used as predictors of SCD from an analysis of the nonlinear dynamics of the signal of heart rate variability (HRV). These authors have shown the great potential of fractal analysis for this purpose. In this article, we worked under the hypothesis that if there is an underlying non-linear dynamic to the HRV signal, this dynamic should be best described by the multifractal analysis, which by fractal indices. Therefore, a comparison of fractal and multifractal features for SCD prediction was made and implemented a classifier that will combine this type of characteristics. The results show that hfluctuacion multifractal index shows a 94.44% sensitivity and 87.50% of specificity compared to the exponents of scale with Detrend Fluctuation Analysis (DFA) an 88.89% and 87.50% respectively. Combining these two features as input for a Support Vector Machine (SVM) is achieved an accuracy of the 97.06%, 100% sensitivity and specificity 94.44%, surpassing the results that are reported in the literature so far.