Cardiovascular disease detection: A hybrid machine learning-AI framework for personalized diagnosis and risk...

Area of Science:

• Computational biology
• Medical informatics
• Artificial intelligence in healthcare

Background:

• Cardiovascular disease (CVD) remains the leading global cause of mortality, necessitating improved diagnostic and therapeutic tools.
• Traditional screening methods often lack personalization, leading to suboptimal patient outcomes.
• The complexity of CVD diagnosis requires advanced computational approaches for accurate risk assessment and early intervention.

Purpose of the Study:

• To develop a hybrid computational framework integrating Support Vector Machine (SVM), Particle Swarm Optimization (PSO), and SHapley Additive exPlanations (SHAP) for enhanced cardiovascular disease diagnosis.
• To leverage diverse patient data, including electronic health records, medical images, and genomic data, for comprehensive patient modeling.
• To improve clinical decision-making through accurate disease prognosis, identification of high-risk individuals, and facilitation of personalized treatment strategies.

Main Methods:

• Development of a mathematical model to address the diagnostic complexity of cardiovascular disease.
• Integration of a Support Vector Machine (SVM) classifier for disease prediction.
• Application of Particle Swarm Optimization (PSO) for hyperparameter tuning of the SVM model.
• Utilization of SHapley Additive exPlanations (SHAP) for AI-based interpretation and understanding of diagnostic predictions.
• Training and validation of the framework using diverse patient data from electronic health records, medical images, and genomic data.

Main Results:

• The proposed hybrid ML-AI framework achieved superior performance on the MIMIC-III clinical database (v1.4).
• Key performance metrics included high accuracy (98.4%), precision (97.5%), recall (96.4%), F1 score (96.9%), and AUC-ROC (97.35%).
• The framework demonstrated strong diagnostic power with high sensitivity (96.4%), specificity (98.7%), and a low negative likelihood ratio (0.036), effectively identifying high- and low-risk patients.

Conclusions:

• The hybrid ML-AI framework offers a significant advancement for the early detection and risk stratification of cardiovascular disease.
• This approach facilitates personalized treatment strategies, potentially reducing healthcare costs associated with ineffective therapies.
• The framework enhances healthcare delivery by improving predictive capabilities, ultimately aiming to improve patient outcomes and quality of life.