Analysis of structural brain MRI and multi-parameter classification for Alzheimer's disease

Area of Science:

• Neuroimaging
• Machine Learning
• Medical Diagnostics

Background:

• Structural Magnetic Resonance Images (sMRI) can reveal neuroimaging markers for Alzheimer's Disease (AD).
• Distinguishing AD patients from Healthy Controls (HC) is crucial for early diagnosis and treatment.
• Machine learning (ML) offers potential for analyzing complex neuroimaging data to identify disease-specific patterns.

Purpose of the Study:

• To investigate differences in brain atrophy between AD patients and HC.
• To apply ML methods for classifying AD patients and HC using sMRI-derived parameters.
• To evaluate the efficacy of single and multi-parameter combinations for AD classification.

Main Methods:

• Acquired T1-weighted sMRI scans from 158 AD patients and 145 age-matched HC from the ADNI database.
• Extracted five parameters: cortical thickness, surface area, gray matter volume, curvature, and sulcal depth.
• Employed Recursive Feature Elimination (RFE) with Support Vector Machines (SVM) and Leave-One-Out Cross-Validation (LOOCV) for feature selection and classification.

Main Results:

• Decreased cortical thickness and gray matter volume were significant indicators of atrophy in AD.
• Key differences were identified in the entorhinal cortex and medial orbitofrontal cortex.
• A multi-parameter combination (cortical thickness, gray matter volume, surface area) achieved 90.76% classification accuracy and an Area Under the Curve (AUC) of 0.94.

Conclusions:

• Multi-parameter combinations provide more effective classification features than single parameters for AD diagnosis.
• Cortical thickness and multi-parameter combinations, after feature selection, are optimal for separating AD from HC.
• ML-based analysis of neuroimaging data shows promise for clinical diagnostic applications in AD.