Classification of mild cognitive impairment developmental trajectories using multispatial scale structural brain...

Area of Science:

• Neuroscience
• Medical Imaging Analysis
• Computational Biology

Background:

• Mild cognitive impairment (MCI) is a transitional stage between normal aging and dementia, with a subset progressing to Alzheimer's disease (AD).
• Accurate classification of MCI converters (MCI-c) from stable MCI (MCI-s) is crucial for timely intervention and clinical trial stratification.
• Understanding brain connectivity at multiple spatial scales is vital for deciphering complex neurological disorders.

Purpose of the Study:

• To investigate brain region interactions across different spatial scales in MCI patients.
• To develop a classification model distinguishing MCI-c from MCI-s.
• To identify precise brain connectivity features indicative of Alzheimer's disease progression.

Main Methods:

• Structural brain networks were constructed at multiple anatomical resolutions (210, 40, and 12 regions).
• Intralayer and interlayer networks were created to analyze within- and between-scale brain region connections.
• A whole-brain network was formed by merging intralayer and interlayer networks, followed by feature selection and classifier training.

Main Results:

• The developed whole-brain network model achieved a classification accuracy of 92.41% for distinguishing MCI converters from stable MCI patients.
• Meaningful connectivity features crucial for precise classification were successfully identified.
• Frequently reported abnormal brain regions in MCI were localized to more specific anatomical areas.

Conclusions:

• The constructed whole-brain network model demonstrates significant potential for accurate MCI classification and early detection of Alzheimer's disease progression.
• Understanding hierarchical brain network relationships is key to comprehending brain structure-function interplay.
• Future work should involve validation in independent cohorts and integration with functional imaging modalities like fMRI and PET.