Identification of mild cognitive impairment subtypes predicting conversion to Alzheimer's disease using multimodal...

Area of Science:

• Neurology
• Biostatistics
• Medical Imaging

Background:

• Mild cognitive impairment (MCI) is a precursor to Alzheimer's disease (AD) dementia.
• MCI patients exhibit diverse pathologies and conversion rates to AD.
• Accurate subtyping and prediction are crucial for early intervention and understanding disease mechanisms.

Purpose of the Study:

• To develop a model for simultaneous subtyping of MCI subjects and prediction of conversion to AD dementia.
• To analyze the distinct biological characteristics of each identified MCI subtype.
• To improve the identification of individuals at high risk for AD progression.

Main Methods:

• Utilized a heterogeneous mixture learning (HML) method to construct a decision tree-based model.
• Integrated multimodal data: cerebrospinal fluid (CSF) biomarkers, structural MRI, APOE genotype, and age.
• Evaluated model performance using F1 score and compared it with random forest and CART methods.

Main Results:

• The HML model achieved an average F1 score of 0.721, outperforming CART and comparable to random forest.
• Identified five distinct subtypes of MCI based on the HML decision tree.
• Classified subtypes into low, moderate, and high conversion rate groups, with moderate groups further distinguished by CSF biomarkers or brain atrophy.

Conclusions:

• The developed HML model effectively subtypes MCI and predicts AD conversion.
• Distinct MCI subtypes possess unique biological profiles and varying rates of progression to AD.
• This subtyping approach aids in identifying at-risk populations and understanding MCI heterogeneity.