A Deep Belief Network Model for Automatic Atrial Fibrillation Detection

Abstract

Atrial fibrillation is an abnormal heart rhythm categorized by the rapid and irregular pulse of the atrial chambers in the heart. The Electrocardiogram is normally used for the analysis of Atrial Fibrillation which is used to check a patient’s rhythm and electrical activity. The diagnosing method involved in this process is used to classify the electrocardiogram signal as normal or atrial fibrillation. In this analysis, Adaptive Dual Threshold Filter and Discrete Wavelet Transform methods are utilized to remove noises present in ECG signals representing Electromyogram, Power line interference, Baseline wandering, White Gaussian Noise, electrode motion artifact, and muscle contraction noises. In this process, a combination of heart rate variability and atrial activity is used to do the feature extraction. This work proposes a Chimp Optimization algorithm-based parameter-optimized Deep Belief classifier for effective Atrial Fibrillation detection. To validate the results comparison analysis is done with the three methods, Random Forest Classifier, Logistic Regression, and Decision Tree Classifier, to the check the outcomes of the proposed model. The simulation outcomes demonstrate that the accuracy of the proposed method achieves 10.81%, 10.81%, and 1.86% better than the existing methods.