Subtyping of mild cognitive impairment using a deep learning model based on brain atrophy patterns

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Neurology

Background:

• Mild cognitive impairment (MCI) exhibits significant individual variability in cognitive decline.
• Current MCI subtyping relies on cognitive measures, leading to unclear boundaries and limited validity.
• Heterogeneity in MCI necessitates more robust subtyping methods.

Purpose of the Study:

• To develop and validate a novel MCI subtyping method based exclusively on brain atrophy patterns.
• To leverage deep learning for classifying MCI subjects into distinct, biologically meaningful subgroups.

Main Methods:

• A deep learning model was trained to distinguish between Alzheimer's disease (AD) and cognitively normal (CN) subjects using whole-brain MRI data.
• The trained model classified MCI subjects into AD-like or CN-like atrophy patterns.
• Subtypes were validated using cognitive, clinical, fluid biomarker, and molecular imaging data.

Main Results:

• Brain atrophy patterns in MCI are sufficiently heterogeneous to enable distinct subtyping.
• The proposed method successfully classified MCI subjects into meaningful subgroups.
• Validation confirmed the biological and clinical relevance of the identified MCI subtypes.

Conclusions:

• Brain atrophy patterns offer a valid basis for subtyping individuals with mild cognitive impairment.
• This deep learning-based approach provides a novel, data-driven method for MCI classification.
• The identified MCI subtypes hold potential for improved diagnosis and personalized treatment strategies.