Optimizing Alzheimer's disease prediction through ensemble learning and feature interpretability with SHAP-based...

Area of Science:

• Artificial Intelligence
• Machine Learning
• Neuroscience

Background:

• Alzheimer's disease (AD) is a leading cause of dementia, necessitating early diagnosis for effective management.
• Developing accurate predictive models for AD is crucial for clinical intervention.
• Interpretable machine learning (ML) offers a pathway to understand disease risk factors.

Purpose of the Study:

• To develop and validate an interpretable ML model for early prediction of Alzheimer's disease (AD).
• To identify key predictors of AD risk using explainable AI techniques.
• To assess the potential clinical applicability of the developed model.

Main Methods:

• Utilized an open-access clinical dataset of 2149 adults aged 60-90 years.
• Trained a stacking ensemble model combining Gradient Boosting and XGBoost, with Logistic Regression as the meta-learner.
• Employed SHapley Additive exPlanations (SHAP) for model interpretability and identification of significant predictors.

Main Results:

• The ensemble model achieved high performance: 97% accuracy, 0.97 AUC, 0.97 precision, 0.94 recall, and 0.96 F1 score for AD prediction.
• SHAP analysis identified memory complaints, Mini-Mental State Examination (MMSE) scores, functional assessments, behavioral symptoms, cholesterol levels, and lifestyle factors (activity, diet, sleep) as primary predictors.
• The model demonstrated strong predictive power and provided insights into the contributing factors for AD risk.

Conclusions:

• The developed ensemble ML model offers accurate and interpretable predictions for early Alzheimer's disease risk.
• SHAP analysis enhances understanding of AD predictors, supporting clinical decision-making.
• The model shows promise for integration into future clinical decision support systems for AD management.