Empirical mode decomposition of heart rate variability. A methodological study

Abstract

Aim of this study is to investigate advantages and disadvantages of empirical mode decomposition (EMD) approaches for the investigation of heart rate variability (HRV). Signal-adaptive approaches like EMD can be used to separate components of HRV which are associated with cardiovascular regulatory mechanisms. Two EMD approaches, standard EMD and complete empirical mode decomposition (CEMD) are used to decompose the HRV of children during temporal lobe epilepsy (TLE; 10 min recordings of 18 children). As nonlinear properties are preserved by EMD, analysis of nonlinear predictability of HRV components is applied resulting in a nonlinear, time-variant, frequency-selective examination of HRV. Especially mode mixing problems are investigated. Complementary analysis steps are suggested to detect their occurrence. CEMD is able to better separate defined HRV components and to reduce, but not completely solve, mode mixing. Nonlinear analysis of CEMD based HRV components results in more distinct differences between specific seizure-related states.