An approach to predict Sudden Cardiac Death (SCD) using time domain and bispectrum features from HRV signal

Abstract

In this paper we present a method to predict Sudden Cardiac Arrest (SCA) with higher order spectral (HOS) and linear (Time) features extracted from heart rate variability (HRV) signal. Predicting the occurrence of SCA is important in order to avoid the probability of Sudden Cardiac Death (SCD). This work is a challenge to predict five minutes before SCA onset. The method consists of four steps: pre-processing, feature extraction, feature reduction, and classification. In the first step, the QRS complexes are detected from the electrocardiogram (ECG) signal and then the HRV signal is extracted. In second step, bispectrum features of HRV signal and time-domain features are obtained. Six features are extracted from bispectrum and two features from time-domain. In the next step, these features are reduced to one feature by the linear discriminant analysis (LDA) technique. Finally, KNN and support vector machine-based classifiers are used to classify the HRV signals. We used two database named, MIT/BIH Sudden Cardiac Death (SCD) Database and Physiobank Normal Sinus Rhythm (NSR). In this work we achieved prediction of SCD occurrence for six minutes before the SCA with the accuracy over 91%.

ECG-based Detection and Prediction Models of Sudden Cardiac Death: Current Performances and New Perspectives on Signal Processing Techniques

“…Shen et al [28] applied 1Hz to 50Hz IIR filter to remove baseline wander in their hardware while Raka et al [22] chose different bandwidth of 0.5Hz to 45Hz for the filter. Houshyarifar et al [9] and Ebrahimzahed et al [12] used two stages moving average filter while Devi et al [6] and Nayan et al [29] have used Fast Fourier Transform to identify and remove the noise. Meanwhile, Acharya et al [10] have used wavelet technique for denoising processes.…”

“…Various sample length is used to extract the ECG features in the available SCD detection and prediction models. Normally one-minute sample starting right before the event of VFib to allow minute-by-minute prediction for up to 3 hours recording [9,10,12,22]. However, Vanitha and Sheela et al [7,8] have used overlapping 10-minute sample window to analyze half an hour SDDB recordings while Devi et al [6] used four minutes sample duration for the same purpose.…”

“…Acharya et al [10] have used 18 non-linear features from wavelet coefficients of recurrence plot, fractal dimensions, Hurst's exponent, approximate entropy, sample entropy, DFA as well as correlation dimension (CD) for four-minute prediction of SCD [10,11]. Houshyarifar et al [9] extended the study to six-minute prediction using two linear and six non-linear bipsectrum features and produced slightly better prediction accuracy compared to Acharya at fourth minute prediction.…”

“…For SCD prediction, researchers have greatly achieved a promising accuracy, more than 90%, as early as six minutes before the event [9][10][11][12][13]. Various linear and nonlinear methods have been used to achieve such accuracy, utilizing the ECG's features in time, frequency and time-frequency domains.…”

Detection and diagnosis of Oscillation in Shell and Tube Heat Exchanger Process

“…Where X (f) is the Fourier transform of x (t) and * represents conjugate complex. As seen from (2), bispectrum contains the information about the relation of phase between the frequency components at f1, f2 and f1 + f2 [10]. Bispectrum can be estimated through several methods such as biased, parametric, direct and indirect methods.…”