Application of empirical mode decomposition (EMD) for automated identification of congestive heart failure using heart rate signals

摘要: Electrocardiogram is widely used to diagnose the congestive heart failure (CHF). It primary noninvasive diagnostic tool that can guide in management and follow-up of patients with CHF. Heart rate variability (HRV) signals which are nonlinear nature possess hidden signatures various cardiac diseases. Therefore, this paper proposes a methodology, empirical mode decomposition (EMD), for an automated identification classification normal CHF using HRV signals. In work, subjected EMD obtain intrinsic functions (IMFs). From these IMFs, thirteen features such as approximate entropy $$ (E_{\text{ap}}^{x} ) $$(Eapx), sample (E_{\text{s}}^{x} $$(Esx), Tsallis (E_{\text{ts}}^{x} $$(Etsx), fuzzy (E_{\text{f}}^{x} $$(Efx), Kolmogorov Sinai (E_{\text{ks}}^{x} $$(Eksx), modified multiscale (E_{{{\text{mms}}_{y} }}^{x} $$(Emmsyx), permutation (E_{\text{p}}^{x} $$(Epx), Renyi (E_{\text{r}}^{x} $$(Erx), Shannon (E_{\text{sh}}^{x} $$(Eshx), wavelet (E_{\text{w}}^{x} $$(Ewx), signal activity (S_{\text{a}}^{x} $$(Sax), Hjorth mobility (H_{\text{m}}^{x} $$(Hmx), complexity (H_{\text{c}}^{x} $$(Hcx) extracted. Then, different ranking methods rank extracted features, later, probabilistic neural network support vector machine differentiating highly ranked into classes. We have obtained accuracy, sensitivity, specificity 97.64, 97.01, 98.24 %, respectively, identifying The proposed technique able identify person having alarming (alerting) clinicians respond quickly proper treatment action. Thus, method may act valuable increasing survival many patients.