Artificial intelligent based arrhythmia identification via single lead ECG recording
Loading...
Date
2017-06
Authors
Lim, Guo Jin
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Electrocardiogram (ECG) represents the electrical activities of our heart. It provides various
information about our heart status such as cardiac disorder or arrhythmia. ECG has become the
most common diagnostic tool in heart analysis as well as in monitoring for cardiac problem. In the
past century, arrhythmia has become the most common heart disease, showing the least symptoms
while having the greatest effect toward the victims. Despite the plenty of studies that have been
done in Arrhythmia detection, it problematic as Arrhythmia may only happen periodically. The
main goal of this study is to develop an artificial neural network based algorithm which is able to
classify the ECG rhythm. At the first stage, the ECG signal is classified into noisy ECG and clean
ECG. Only clean ECG signal will be fetched into the second stage to be classified into Arrhythmia
or Normal Sinus rhythm. Different features have been used in both stages and been fetched into
trained MLP neural network for classification purpose. At first stage classification, 6 features have
been selected as input and 15 number of neurons in hidden layer have been used. Meanwhile at
the second stage, 4 features have been selected as input and 40 numbers of hidden layer’s neuron
has been used. Final accuracy of 83.3% has been achieved during the training stage by using 300
training data. Final score of 0.7076 (Perfect score = 1) has been achieved when the 8528 data has
been fetched into the developed neural network. In conclusion, suitable features have been
identified which are average and standard deviation of heart rate and R-peak amplitude. Finally, a
high accuracy neural network has been developed in this study.