AdSOCNet: A Deep Learning Framework for Intelligent Cardiovascular Disease Detection using Adaptive Stellar Optimization

Abstract

Cardiovascular Disease (CVD) is a disorder that affects the function of the heart and is considered as the leading cause of death across the world. Timely diagnosis of CVD can increase the survival rate and patient outcomes, but manual detection based on clinical data is time-consuming and complex. Many existing Deep Learning (DL) models introduced to diagnose CVD have exhibited promising performance. However, these schemes experienced overfitting and acquired low accuracy rates because of the high-dimensional data and the existence of redundant features. Thus, a novel Adaptive Steller Optimized Stacked Convolutional Neural Network (AdSOCNet) is designed to identify CVD from heart disease data. At first, the heart disease data is normalized exploiting the Squash-Norm-Sigmoid (SNS) normalization. After that, the feature selection is carried out based on variance thresholding. Subsequently, CVD is identified utilizing the AdSOCNet, which is established by integrating Stacked Convolutional Neural Network (SCNN), where SCNN is tuned by applying the Adaptive Steller Oscillation (AdSO) algorithm. Moreover, the Adaptive concept and Steller Oscillation Optimizer (SOO) are merged to devise AdSO. Furthermore, the newly formulated AdSOCNet achieved the maximum accuracy, Geometric-Mean (G-Mean), and Matthews Correlation Coefficient (MCC) of 98.099%, 96.666%, and 95.545%, respectively.