Evaluation of Energy Power Spectral Distribution of QRS Complex for Detection of Cardiac Arrhythmia

Abstract

The proposed approach involves detection of QRS complex and energy power spectral distribution analysis of the segmented QRS complex to establish the presence of arrhythmic beats in Electrocardiogram (ECG). The methods consist of three steps: (i) the baseline drift and high-frequency artifacts could seriously affect the detection performance, so Moving Average Filtering (MAF) and Stationary Wavelet Transform (SWT) are implemented at preprocessing stage. (ii) Localization of R-peaks by implementing FFT-based windowing and thresholding techniques. Then Q and S points are detected using search interval method based on the medical definition. (iii) The segmented QRS complex is analyzed with period-gram and Continuous Wavelet Transform using FFT (CWTFT) to obtain time–frequency domain power and energy of the complex. (iv) Statistical analysis has been proposed using one-way ANOVA to differentiate the healthy and arrhythmic QRS complex. The proposed QRS detection and analysis methodologies are evaluated with MIT-BIH Arrhythmia Database (MITDB) and FANTASIA database. The detection performance, i.e., Sensitivity ${\text{S}}_{\text{e}} \left( \% \right)$ and the Specificity ${\text{S}}_{\text{p}} \left( \% \right)$ for FANTASIA 100% each respectively, where as ${\text{S}}_{\text{e}} = 100\%$ and ${\text{S}}_{\text{p}} = 98.18\%$ for MITD. The failed detection percentage, ${\text{F}}_{\text{d}} \left( \% \right) = 0$ for FANTASIA and ${\text{F}}_{\text{d}} \left( \% \right) = 1.85\%$ for MITDB. The energy power distributed parameters obtained from PSD and CWTFT are statistically analyzed with one-way ANOVA and the p-value are found to be  <0.05 (i.e., CI = 95%) for healthy and arrhythmia QRS complex which certainly signifies that the energy power features of the arrhythmic QRS complex are different than the normal QRS complex.