Detection of cardiac abnormality using entropy measure of ECG attractor

Abstract

Chaotic analysis and entropy measurement has been shown to be useful in a variety of medical applications, particularly in cardiology. Chaotic parameters like Poincare plot indexes have shown potential in the identification of diseases, especially in the analysis of biomedical signals like electrocardiogram (ECG). In this work, entropy measurement, Poincare plot indexes and time domain parameters in ECG signals have been analyzed. First, the ECG signal is processed through a series of steps to extract the QRS complex. From this extracted feature, bit-to-bit interval (BBI) and instantaneous heart rate (IHR) have been calculated. We quantified several time domain HRV parameters: mean IHR, variance, standard deviation of normal IHR data (SDNN) and the square root of the mean squared difference of the successive IHR data (RMSSD). and nonlinear techniques like approximate entropy, sample entropy, and Poincare plot indexes have been determined from the IHR. Standard database of MIT-BIH is used as the reference data where each ECG record contains 650000 samples. Approximate entropy (ApEn) and sample entropy (SampEn) are calculated for IHR for each ECG record of the database. A much higher value of ApEn and SampEn for IHR is observed for eight patients with abnormal beats like T, AFIB, VT. On the contrary, the ApEn and SampEn for IHR of eight patients with normal rhythm shows lower value. The IHR time series and their corresponding Poincare plots taken from the HRV patterns of patients with normal heart beat and abnormal heart beat are presented in Figure. Poincare plot indexes for IHR of eight normal rhythm records show lower values with a SDI of 5.6087 and SD2 of 7.I364.0n the other hand Poincare plot indexes for IHR of eight abnormal rhythm records show higher values with a SDI of 23.2093 and SD2 of 22.6107. Time domain parameter are found lower for eight normal rhythm records with variance of 52.6808, SD of 6.5875, SDSD of 7.9260 and SMSSD of 77.5388 .. Time domain parameter are found lower for eight abnormal rhythm records with variance of 562.3727, SD of 23.2972, SDSD of 32.8103 and SMSSD of 86.6614. These results indicate that ECG can be classified based on this chaotic modelling which works on the nonlinear dynamics of the system.