BUNDLE BLOCK DETECTION USING DIFFERENTIAL EVOLUTION AND LEVENBERG MARQUARDT NEURAL NETWORK

Abstract

Globally heart diseases are the most prevalent cause for human mortality. Every year, 9.4 million
deaths are attributed to heart diseases (cardiac arrhythmia) including 51% of deaths due to strokes and
45% deaths due to coronary heart diseases. Hence identification of these heart diseases in the early stages
becomes important for the prevention of cardiac re- lated deaths. Although the existing conventional
ECG analysis methods like, RR interval, Wavelet transform (WT) with classification techniques, such as,
Support Vector machine (SVM), K-Nearest Neighbor (KNN) and Levenberg Marquardt Neural Network
(LMNN) are used for detection of cardiac arrhythmia, the fea- ture extraction using these methods
generally yield a large number of features, of which many might be insignificant. In this paper
Differential Evolution (DE) can be efficiently used to detect the changes in theECG using optimized features
from the ECG beats. For the detection of normal and BBB beats, these DE feature values are given as
the input for the LMNN classifier. The data was collected from MIT-BIH arrhythmia database