Conversion Discriminative Analysis on Mild Cognitive Impairment Using Multiple Cortical Features from MR Images.

Area of Science:

• Neuroimaging
• Neurology
• Biomedical Engineering

Background:

• Mild cognitive impairment (MCI) diagnosis and progression monitoring are critical for Alzheimer's disease (AD) prevention.
• Neuroimaging, particularly magnetic resonance imaging (MRI), offers valuable insights into brain structure and changes associated with MCI.
• Cortical features are key indicators of anatomical patterns that can distinguish between normal cognition and MCI states.

Purpose of the Study:

• To investigate the discriminative power of four cortical features (gray matter volume, cortical thickness, surface area, mean curvature) for classifying three groups: stable MCI (sMCI), converted MCI (cMCI), and normal controls (NC).
• To develop and validate a machine learning approach for improved sub-typing and prediction of MCI progression.

Main Methods:

• Utilized a dataset of 158 subjects (72 NC, 46 sMCI, 40 cMCI) from the Alzheimer's Disease Neuroimaging Initiative.
• Employed a sparse-constrained regression model (l2-1-norm) for feature dimensionality reduction and selection.
• Applied a support vector machine (SVM) classifier with an optimized feature addition strategy.

Main Results:

• Cortical thickness achieved high classification accuracies: 98.84% (sMCI-cMCI), 92.37% (cMCI-NC), and 93.75% (sMCI-NC).
• The SVM classifier using selected essential features outperformed models using all retained features by 5-40%.
• The proposed method demonstrated significant potential in recognizing anatomical patterns for MCI sub-typing.

Conclusions:

• Cortical features, especially cortical thickness, are highly effective for differentiating MCI subtypes and normal controls.
• The developed sparse regression and SVM approach enhances classification performance for MCI diagnosis.
• This method shows promise for improving clinical diagnosis of MCI and predicting conversion risk to Alzheimer's disease.