Dynamic and Static Structure-Function Coupling With Machine Learning for the Early Detection of Alzheimer's Disease

Area of Science:

• Neuroscience
• Medical Imaging
• Computational Biology

Background:

• Alzheimer's disease (AD) progression involves intricate brain structure-function coupling (SFC) changes.
• Both static (overall) and dynamic (transient) SFC are crucial for understanding AD.
• Early detection of AD is vital for effective intervention.

Purpose of the Study:

• To evaluate the combined potential of static and dynamic SFC with machine learning (ML) for early AD detection.
• To quantify differences in static and dynamic SFC across AD, mild cognitive impairment (MCI), and healthy control (HC) groups.
• To assess the correlation between SFC features and early AD physiological biomarkers.

Main Methods:

• Analysis of discovery and validation cohorts including AD, MCI, and HC groups.
• Quantification of static and dynamic SFC, with feature selection using ElasticNet.
• Classification of AD stages using a Gaussian naive Bayes (GNB) classifier.

Main Results:

• Static SFC increased with AD progression, while dynamic SFC exhibited decreased stability.
• The GNB classifier achieved high AUCs: 91.1% for HC vs. MCI and 89.03% for MCI vs. AD.
• Significant correlations were observed between SFC features and physiological biomarkers.

Conclusions:

• Combined static and dynamic SFC features, selected via ElasticNet and classified by GNB, show high accuracy in differentiating AD stages.
• This integrated approach holds significant potential for the early detection and accurate classification of Alzheimer's disease.
• The study offers a novel perspective on AD mechanisms and contributes to improved early diagnostic strategies.