AIM: An Advanced Hybrid Inference Model Combining Clinical Rules and Lifelog-Based Learning for Health Risk...

Area of Science:

• Artificial Intelligence in Healthcare
• Clinical Decision Support Systems
• Metabolic Health

Background:

• Early identification of metabolic health risks is crucial for effective preventive interventions.
• Routine laboratory testing for metabolic health is often unavailable in general health management settings.
• Purely data-driven AI models may lack clinical interpretability.

Purpose of the Study:

• To introduce the Advanced Hybrid Inference Model (AIM), a clinically interpretable framework for metabolic risk screening.
• To combine biomarker estimation, AI-based risk prediction, and rule-based interpretation for enhanced clinical utility.

Main Methods:

• A three-stage Random Forest-centered pipeline was implemented for the AIM framework.
• Stage 1: Estimation of metabolic biomarkers from anthropometric and demographic data.
• Stage 2: Random Forest model for metabolic risk prediction using measured/estimated biomarkers and clinical variables.
• Stage 3: Rule-based interpretation to translate model outputs into clinically meaningful risk messages.

Main Results:

• Experimental validation was performed on clinically collected, class-imbalanced datasets.
• The AIM framework demonstrated potential in identifying high-risk metabolic patterns.
• Findings suggest AIM's utility as a screening-oriented approach.

Conclusions:

• AIM is an exploratory clinical screening support framework, not a diagnostic tool.
• The framework prioritizes interpretability, rule-based reasoning, and risk prioritization.
• AIM offers a novel approach to metabolic risk assessment in resource-limited settings.