Identifying Mild Cognitive Impairment with Random Forest by Integrating Multiple MRI Morphological Metrics

Area of Science:

• Neuroimaging
• Neurology
• Biomedical Engineering

Background:

• Mild cognitive impairment (MCI) is a precursor to Alzheimer's disease (AD), necessitating early and accurate diagnostic methods.
• Identifying individuals with MCI is critical for timely intervention and potential disease modification.

Purpose of the Study:

• To develop and validate a robust method for early identification of MCI using neuroimaging biomarkers.
• To compare the diagnostic performance of combined morphological metrics against individual metrics.

Main Methods:

• Utilized voxel-based, deformation-based, and surface-based morphometry to extract multiple brain morphological metrics.
• Employed a random forest classifier trained on a discovery dataset (ADNI) and validated on independent datasets.
• Evaluated performance using 10-fold cross-validation and external datasets.

Main Results:

• The combined morphological metrics achieved approximately 80% accuracy in identifying MCI across all datasets.
• This performance significantly surpassed models using single metrics (45%-76% accuracy).
• Cortical thickness and gyrification index emerged as key biomarkers for MCI detection.

Conclusions:

• A multi-metric approach using random forest classification offers a reliable and accurate method for early MCI diagnosis.
• Surface-based morphometry metrics, specifically cortical thickness and gyrification index, are promising biomarkers for MCI.
• This method shows potential for assisting clinicians in the early diagnosis of MCI.