A preliminary study of an interpretable ensemble learning framework for atherosclerosis detection: algorithm...

Area of Science:

• Biomedical Informatics
• Machine Learning in Healthcare
• Cardiovascular Disease Research

Background:

• Accurate and interpretable models are crucial for atherosclerotic disease detection.
• Existing methods may struggle with noisy clinical data and lack biological interpretability.

Purpose of the Study:

• To develop a robust and interpretable ensemble learning framework for atherosclerotic disease detection.
• To introduce the Progressive Saturated Exponential Loss Adaptive Boosting (ProSEL-Boost) algorithm.
• To generate biologically meaningful features and identify novel biomarkers.

Main Methods:

• Developed the ProSEL-Boost algorithm with a dynamic loss function for data robustness.
• Implemented a medical knowledge-guided framework for feature engineering.
• Employed multi-strategy ensemble methods for feature selection.
• Validated the framework on public and proprietary atherosclerosis datasets.

Main Results:

• ProSEL-Boost demonstrated superior performance (accuracy = 0.9508, AUC = 0.8807).
• Identified the glucose-lipid interaction index (TG_FBG_Index) as a significant biomarker (Cohen's d = 1.97, AUC = 0.86).
• Highlighted a synergistic metabolic risk factor for cardiovascular disease.

Conclusions:

• The ProSEL-Boost framework offers a robust and interpretable approach to atherosclerotic disease detection.
• The identified glucose-lipid interaction index warrants further investigation as a clinical biomarker.
• This methodological validation provides a foundation for larger prospective studies in early disease detection and risk stratification.