Predictive modeling for early risk detection of cardiovascular disease using deep convolutional neural networks

Abstract

Cardiovascular disease continues to be one of the world’s major causes of morbidity and mortality underscoring the vital need for precise and timely diagnostic instruments. By enabling automated feature extraction from complicated and high-dimensional data, recent developments in deep learning have shown great promise for medical decision support systems. Predictive modeling for early cardiovascular risk identification using deep convolutional neural networks is presented in this research through categorization of subtle correlations and nonlinear patterns in clinical and diagnostic data, the model’s usage of conventional layers increased prediction accuracy when compared to traditional machine learning techniques, the results of the experiment show that the model achieved a superior training accuracy for high and low-risks. Table 6 shows that it correctly classifies 9 out of 12 as high and misclassifies 3, yielding an apparent error of 26.73%. It also correctly classifies 12 out of 14 as low and misclassifies 2, yielding an apparent error of 14.29%, 96.4% accuracy, 94.1% sensitivity and 95.7% specificity for training. Table 7 shows that the deep learning model for testing correctly classifies 3 out of 4 as high and misclassifies 1, yielding an apparent error of 25%, also correctly classifies 6 out of 8 as low and misclassifies 2, yielding an apparent error of 25%, 91.2% accuracy, 89.3% sensitivity and 92.0% specificity.  Patients were categorized by the model into high-risk and low-risk groups according to their attributes.