Prognostic classification of mild cognitive impairment and Alzheimer's disease: MRI independent component analysis

Area of Science:

• Neuroimaging
• Biomarkers
• Cognitive Neuroscience

Background:

• Early and accurate diagnosis of mild cognitive impairment (MCI) and Alzheimer's disease (AD) is essential for effective treatment and clinical trial enrollment.
• Identifying reliable predictors can improve diagnostic accuracy and patient stratification.

Purpose of the Study:

• To evaluate the diagnostic accuracy of independent component analysis (ICA)-derived magnetic resonance imaging (MRI) patterns in predicting cognitive status.
• To assess the contribution of MRI biomarkers, cerebrospinal fluid (CSF) biomarkers, and genetic factors in differentiating cognitive groups.

Main Methods:

• Utilized structural MRI data from 320 participants (cognitively normal, MCI, AD) in the Alzheimer's Disease Neuroimaging Initiative database.
• Applied ICA to derive 30 gray matter covariance patterns (ICs) and employed discriminant classifier analyses for prediction.
• Cross-validated models using a "leave-10-out" procedure.

Main Results:

• Achieved high diagnostic accuracies: 84.7% for CN vs. MCI, 94.8% for CN vs. AD, 83.1% for MCI vs. AD, and 80.3% for stable MCI vs. MCI progression.
• ICA-derived MRI biomarkers demonstrated excellent diagnostic performance, with minimal improvement from CSF biomarkers (p-tau(181p)) and ApoE ε4 status.

Conclusions:

• ICA-derived MRI biomarkers are powerful predictors for diagnosing MCI and AD and tracking MCI progression.
• These imaging biomarkers offer a robust, non-invasive approach to patient stratification for clinical trials and diagnosis.