Early Diagnosis of Alzheimer's: Machine Learning Analysis Leveraging Structural MRI

Area of Science:

• Neuroimaging
• Biomarkers
• Machine Learning

Background:

• Alzheimer's disease (AD) is associated with significant brain atrophy, often detected by structural MRI.
• Subcortical degeneration and its progression from Mild Cognitive Impairment (MCI) to AD are underexplored.
• Early detection of AD is crucial for timely intervention and management.

Purpose of the Study:

• To identify subcortical regions exhibiting progressive atrophy in individuals transitioning from MCI to AD.
• To evaluate the potential of surface-based morphometry (SBM) metrics for early AD diagnosis.
• To assess the efficacy of machine learning models in classifying and predicting AD progression using SBM data.

Main Methods:

• Longitudinal analysis of the ADNI dataset, including MCI patients who converted to AD and healthy controls (HC).
• Utilized surface-based morphometry (SBM) to quantify cortical thickness (CTh), gyrification index (GI), and sulcal depth (SD) in 68 subcortical regions.
• Trained and tested machine learning (ML) models to differentiate MCI-to-AD converters (MCI-AD) from HC across multiple time points (6 months to 3 years).

Main Results:

• Significant atrophy was detected in specific subcortical regions in MCI-AD patients compared to HC.
• Machine learning models accurately classified MCI-AD from HC, with performance improving at later time points.
• Cortical thickness (CTh) demonstrated the most significant decline, followed by sulcal depth (SD) and gyrification index (GI).

Conclusions:

• Subcortical SBM metrics, particularly CTh, serve as sensitive biomarkers for early AD-related atrophy and disease progression.
• ML models trained on SBM features provide a scalable and accurate framework for early AD detection and prediction.
• The findings highlight the utility of SBM and ML in identifying individuals at risk for AD, facilitating early intervention strategies.