Multiscale analysis of heart rate variability for the characterization of subjects with reflex syncope during head-up tilt test

Abstract

This study aims to characterize the heart rate variability of subjects with reflex syncope during head-up tilt test and to find out non-linear parameters that can differentiate between subjects who develop syncope during the test and others who do not. For this purpose, the Empirical Mode Decomposition was applied to RR time series of subjects underwent a head-up tilt test. It decomposes the time series into waveforms called Intrinsic Mode Functions (IMF). The IMFs and their cumulative sums were used to extract several parameters as Sample Entropy, Detrended Fluctuation Analysis and indicators from Poincare plot. The results show that the proposed parameters are significantly different between the two groups in several time scales (p-value<0.05), which indicate that the heart rate variability has different non-linear characteristics in patients with reflex syncope that can be explored with a multi-resolution tool. Furthermore, pertinent parameters were selected using feature selection algorithm, and employed in the classification test using K-nearest neighbor (KNN). The classification performance show that the selected parameters are efficient to separate between the two populations with a sensitivity of 71% during the first 5 minutes of the tilted position. These findings can help to early predict the outcomes of the head-up tilt test and reduce its duration.