Coronary heart disease classification using deep learning approach with feature selection for improved accuracy.

Area of Science:

• Cardiology
• Artificial Intelligence
• Biomedical Informatics

Background:

• Coronary heart disease (CHD) is a leading cause of mortality, necessitating robust risk prediction models.
• Evaluating numerous risk factors for CHD is challenging for human experts, highlighting the need for effective feature selection.
• Deep learning (DL) models show promise for early CHD detection and cardiovascular risk assessment.

Purpose of the Study:

• To develop a deep Convolutional Neural Network (CNN) model for CHD classification.
• To employ the LASSO (least absolute shrinkage and selection operator) technique for efficient feature selection.
• To compare the proposed model's performance against existing studies using accuracy metrics.

Main Methods:

• Utilized the National Health and Nutritional Examination Survey (NHANES) CHD dataset with 49 features.
• Applied the LASSO technique for feature selection.
• Developed an improved CNN model with a feature extractor (two 1D convolution layers) and a classifier (two fully connected layers with SoftMax).
• Evaluated the model using accuracy, recall, specificity, and ROC metrics.

Main Results:

• The LASSO model effectively selected relevant features from the CHD dataset.
• The proposed CNN model achieved a high accuracy of 99.36%.
• This accuracy significantly surpasses the performance of previous models, which ranged from 80% to 92%.

Conclusions:

• The combined CNN and LASSO model offers an effective approach for CHD classification.
• This method can accelerate CHD diagnosis and improve cardiovascular disease prediction based on risk features.
• The model demonstrates significant potential for clinical application in identifying individuals at risk for cardiovascular events.