Paper Title
Classification of Heart Disease Using ECG Signals With A Hybrid Deep Maxout Network Enhanced By Convolutional Block Attention
Abstract
In the entire world, cardiovascular disease (CVD) is the main cause of death. CVD has been diagnosed using electrocardiograms (ECGs), which document the electrical activity of the heart. The difficulty of training data when a dataset contains a large amount of data is one limitation that has been discovered from various surveys of existing works. To train machine learning techniques effectively, a lot of labelled data is required. It might take more time and memory when using ML techniques. Working with a large volume of data and efficient signal processing are required when analyzing heartbeats for the prediction of heart disease. To overcome the above issues, a better recognition technique for predicting the CVD is introduced in this research. Initially, data are collected from MIT-BIH Arrhythmia Database which are then pre-processed using Modified FIR filtering model (M-FIR) to filter the ECG signals, enhancing the signal quality and remove unwanted noises. The feature extraction is done using the technique of Radial Hilbert function transform network (RHFTN) and the feature selection is done using improved bald eagle search algorithm (IBEA) to minimize the irrelevant features. To detect and classify the heart disease from ECG signals using convolutional block attention assisted hybrid deep Maxout network model (CB-HDM). The performance outcomes are obtained as 98.81% in accuracy, 96.18% in precision, 96.87% in recall, and 94.39% in f1 score. The proposed model exhibits better efficiency when associated to other classifiers.
Keywords - Cardiovascular Disease, Electrocardiogram, FIR Filtering, Radial Hilbert Function, Bald Eagle Search Algorithm, Convolutional Block Attention, Hybrid Deep Maxout Network.