Enhancing Hypertension Risk Diagnosis Using a Hybrid Machine Learning Framework: Leveraging Body Composition Data.

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Hypertension is a major global health risk, leading to severe cardiovascular, renal, and neurological issues.
• Current prediction methods often lack interpretability and reliability, hindering early detection and personalized interventions.

Purpose of the Study:

• To develop and validate a dual-scenario hybrid machine learning framework for hypertension risk prediction.
• To enhance both the interpretability and predictive reliability of hypertension risk assessment using noninvasive body composition features.

Main Methods:

• Scenario 1: Unsupervised clustering (K-Means) on hypertensive individuals to identify physiological subgroups, validated with clustering indices.
• Scenario 2: Binary classification (ExtraTrees, KNN, SVM, GNB, Decision Tree) comparing standard and cluster-augmented datasets.

Main Results:

• Scenario 1 identified five distinct subgroups with significant intercluster variability (p < 0.001) in body composition and age, achieving 99.08% accuracy with SVM.
• Scenario 2, using a cluster-augmented dataset, showed superior performance with the ExtraTrees classifier (98.23% accuracy, 99.87% AUC).
• Both clustering and feature selection improved model generalization, especially for ensemble methods.

Conclusions:

• Integrating unsupervised clustering with supervised classification provides a robust and explainable framework for personalized hypertension risk prediction.
• The proposed hybrid model contributes to early detection and precision healthcare by offering enhanced accuracy and interpretability.