Statistical Classification of Ventricular Tachycardia and Ventricular Fibrillation based on Histogram and Average Absolute Deviation

Shijie Zhou and Jason Gu

Keywords

Arrhythmias Classification, Histogram Feature Extraction, Average Absolute Deviation, Coarse-graining Processing Analysis

Abstract

In this paper, a novel, and computationally fast method was proposed to classify ventricular tachycardia (VT) and ventricular fibrillation (VF) by using histogram and average absolute deviation. To begin with, a time-domain method based histogram, was addressed to refine the general features that illustrate the density of amplitude distribution of a window length long ECG signal in successive numerical intervals of equal size. Besides, segments of monomorphic VT randomly collected from the MIT-BIH Malignant Arrhythmia subset are calculated by the histogram function for obtaining the reference histogram with different length. Histogram differences in case of monomorphic VT or VF calculated as features to generate a threshold for arrhythmias distinction. Then, an average absolute deviation algorithm is applied to obtain a deviation value that is compared to the threshold for monomorphic VT or VF classification. The novelty of this method is that ECG signal statistics, morphological analysis, the histogram of signal (density estimation) and average absolute deviation altogether have been used to achieve a higher classification rate. The test shows that the detection accuracy for monomorphic VT, and VF is approximate 100% for a test set of MIT-BIH database (98 monomorphic VT, and 128 VF). Compared with LZ (Lempel-Ziv) complexity measure based on time-domain analysis, the proposed method has high performance, low computational complexity, and will be able to well implement on ECG Tele-monitoring analysis systems.

Important Links:



Go Back