Machine learning-based classification of valvular heart disease using cardiovascular risk factors

Area of Science:

• Biomedical Informatics
• Cardiovascular Medicine
• Machine Learning in Healthcare

Background:

• Valvular Heart Disease (VHD) is a major global cause of mortality, especially in aging populations.
• Uncertainties persist regarding significant cardiovascular disease risk factors contributing to VHD.
• Advancements in interventions have not fully resolved VHD risk factor complexities.

Purpose of the Study:

• To investigate uncertainties in VHD risk factors.
• To explore machine learning for categorizing VHD based on cardiovascular risk factors.
• To enhance diagnostic capabilities for Valvular Heart Disease.

Main Methods:

• A two-part study involving feature extraction (wrapping approach, binary logistic regression) and classification.
• Utilized classifiers: Artificial Neural Network (ANN), XGBoost, Random Forest (RF), Naïve Bayes, Support Vector Machine (SVM).
• Advanced methods included SVM with Principal Component Analysis (PCA) and a majority-voting ensemble (MV5).

Main Results:

• SVM combined with PCA demonstrated the highest overall performance.
• The MV5 ensemble method achieved high accuracy with balanced sensitivity and specificity.
• Statistical measures like ROC curve, F-measure, accuracy, MCC, and Kappa were used for assessment.

Conclusions:

• The study highlights the significance of specific risk factors in VHD prevalence.
• SVM+PCA and MV5 show exceptional performance in diagnosing VHD.
• These findings contribute to advancing biomedicine and managing the VHD burden.