Predicting cognitive dysfunction and regional hubs using Braak staging amyloid-beta biomarkers and machine learning

Area of Science:

• Neuroscience
• Biomarkers
• Artificial Intelligence

Background:

• Mild cognitive impairment (MCI) represents a critical stage between normal aging and Alzheimer's disease (AD).
• Extracellular amyloid-beta (Aβ) accumulation in specific brain regions correlates with cognitive decline in MCI and AD.
• Understanding the link between regional Aβ and cognitive function is vital for early AD detection and prevention.

Purpose of the Study:

• To develop and validate machine learning models for predicting cognitive dysfunction based on regional Aβ biomarkers.
• To identify the dominant brain regions associated with cognitive impairment using Aβ data.
• To explore the multivariate predictive relationships between Aβ biomarkers and cognitive function in MCI/AD.

Main Methods:

• Support Vector Regression (SVR) and Artificial Neural Network (ANN) models were trained using individual Aβ biomarkers and cognitive measurements.
• The Local Interpretable Model-Agnostic Explanations (LIME) technique was integrated with ANN to identify key brain regions.
• Model performance was evaluated on a test set to predict cognitive performance solely from Aβ biomarkers.

Main Results:

• Elevated Aβ levels were observed in MCI patients compared to controls.
• A significant correlation between Aβ biomarkers and cognitive function was found, particularly in Braak stages III-IV and V-VII.
• ANN models demonstrated strong predictive power for cognitive impairment using regional Aβ biomarkers.
• LIME analysis identified the parahippocampal gyrus, inferior temporal gyrus, and hippocampus as critical regions for cognitive decline prediction.

Conclusions:

• Machine learning, particularly ANN integrated with LIME, provides a robust framework for estimating cognitive impairment from Aβ biomarkers.
• The study highlights the parahippocampal gyrus, inferior temporal gyrus, and hippocampus as key brain regions in AD pathophysiology related to Aβ.
• This analytical approach aids in understanding Aβ-related cognitive dysfunction and supports early AD detection strategies.