A Novel Hybrid Deep Learning System for Cardiovascular Detection and Salient Feature Extraction from ECG Data

Abstract

The Cardiovascular system is responsible for the circulation of blood throughout the body. Abnormalities in the cardiovascular system can lead to various diseases such as arrhythmia, heart failure, and myocardial infarction. The Electrocardiogram (ECG) is unique and commonly used diagnostic tools for detecting cardiovascular diseases. The conventional methods of ECG analysis require expert interpretation and are time-consuming. Automated ECG analysis systems cantered on machine learning and also deep learning techniques have been proposed to overcome the limitations of conventional methods. In this research, we propose a hybrid deep learning-based cardiovascular detection system that can accurately detect various cardiovascular diseases by extracting salient features from ECG data. The suggested approach combines feature extraction using a convolutional neural network with wavelet transform and principal component analysis. The fused signals obtained from the previous steps are optimized using Sequential Minimal Optimization (SMO) algorithm to improve classification accuracy. Therefore, the development of a reliable and automated ECG analysis system is highly desirable testing on a publicly available ECG dataset from MIT-BIH Arrythmia.