Efficient Explainable Models for Alzheimer's Disease Classification with Feature Selection and Data Balancing...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Informatics

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder and the leading cause of dementia.
• Early AD diagnosis is crucial but challenging due to disease complexity.
• Clinical data is vital for AD classification, yet faces hurdles like data imbalance and high dimensionality.

Purpose of the Study:

• To propose a computationally efficient, reliable, and transparent machine learning (ML) framework for AD patient classification.
• To enhance model interpretability for medical practitioners to understand complex patient patterns.
• To develop a system that aids in the early and accurate diagnosis of Alzheimer's disease.

Main Methods:

• Employed ensemble learning (boosting algorithms) for improved classification accuracy.
• Utilized random sampling for data balancing and feature selection/reduction for model optimization.
• Integrated Explainable AI (XAI) tools (SHAP, LIME, ALE, ELI5) for model transparency and feature importance insights.

Main Results:

• Achieved a robust, interpretable, and clinically relevant framework for AD diagnosis.
• The proposed ML approach demonstrated a high accuracy of 95% in classifying Alzheimer's disease patients.
• The framework effectively identified key clinical features influencing classification, enhancing diagnostic reliability.

Conclusions:

• Integrating ensemble learning with XAI and balanced, feature-selected data significantly improves AD classification accuracy and interpretability.
• This approach offers a promising tool for early and informed clinical decision-making in Alzheimer's disease.
• The developed framework provides transparency, enabling clinicians to trust and understand the predictive factors for AD.