Detection of Alzheimer's Disease using Explainable Machine Learning and Mathematical Models

Area of Science:

• Neuroimaging
• Machine Learning
• Mathematical Modeling

Background:

• Alzheimer's disease (AD) diagnosis relies on accurate staging.
• Magnetic Resonance Imaging (MRI) provides detailed brain structure information.
• Current classification methods can be improved for accuracy and efficiency.

Purpose of the Study:

• To develop a novel approach for classifying four Alzheimer's disease (AD) stages using MRI scans.
• To integrate mathematical modeling with machine learning (ML) for enhanced diagnostic capabilities.
• To explore the utility of physical concepts like moment of inertia tensors in neuroimaging analysis.

Main Methods:

• MRI pixel value matrices were mapped to 2x2 matrices using moment of inertia (MI) tensor principles.
• Eigenvalues of the inertia tensor were utilized in conjunction with ML algorithms.
• Various ML models were evaluated for their performance in classifying AD stages.

Main Results:

• The Gaussian Naïve Bayes classifier demonstrated the highest accuracy at 95.45%.
• The integrated mathematical and ML approach proved effective in distinguishing AD stages.
• Performance comparisons were made across multiple ML models.

Conclusions:

• The proposed method achieves high accuracy in Alzheimer's disease (AD) staging from MRI.
• The approach offers computational efficiency through dimensionality reduction.
• Inertia tensor analysis provides novel physical insights into AD progression.