Optimized deep learning for cardiovascular disease diagnosis
TL;DR
An optimization-enabled deep learning model, Exponential weighted Serval Optimization Algorithm Shuffle-Attention Network (Ex-SOA_SA-Net) for the diagnosis of CVD is presented and the efficacy of Ex-SOA_SA-Net is validated by utilizing different performance indicators.
Abstract
The heart plays a key role in the human body, where it helps to deliver oxygen and nutrients by circulating blood all over the body. Cardiovascular Disease (CVD) is the leading cause of increasing mortality rates globally, specifically among the elderly and adults. In recent years, different approaches have been utilized to accurately and timely diagnose CVD to provide timely medical interventions and safeguard patients’ lives. However, these methods encounter difficulties in handling noisy data and extracting meaningful patterns from high-dimensional data. This paper presents an optimization-enabled deep learning model, Exponential weighted Serval Optimization Algorithm Shuffle-Attention Network (Ex-SOA_SA-Net) for the diagnosis of CVD. Initially, median normalization is used to normalize the accumulated input data. Then, the selection of relevant features is performed using Pearson’s Correlation Coefficient (PCC), Relief, and information gain from the normalized data. In addition, the efficacy of Ex-SOA_SA-Net is validated by utilizing different performance indicators. The experimental outcomes show that the Ex-SOA_SA-Net attained high results with an accuracy of 93.469%, True Positive Rate (TPR) of 95.237%, True Negative Rate (TNR) of 91.543%, precision of 97.098%, and an F1-score of 97.960%, respectively.