Cognitive Profiling Related to Cerebral Amyloid Beta Burden Using Machine Learning Approaches

Area of Science:

• Neuroscience
• Biomarkers
• Machine Learning

Background:

• Cerebral amyloid beta (Aβ) is a key Alzheimer's disease (AD) biomarker.
• Current detection methods (imaging, CSF) are costly and invasive.
• Need for accessible, cost-effective Aβ detection methods.

Purpose of the Study:

• Identify cost-efficient markers for Aβ positivity in non-demented individuals.
• Utilize machine learning (ML) to develop a predictive model.
• Establish a potential surrogate for non-invasive Aβ detection.

Main Methods:

• Analyzed 762 participants from ADNI-2 cohort.
• Used demographic and neuropsychological data as predictors.
• Applied adaptive LASSO ML algorithm to identify significant predictors of Aβ positivity.

Main Results:

• Visuospatial ability and episodic memory scores were significant predictors of Aβ positivity.
• The ML model successfully distinguished individuals with abnormal Aβ levels.
• Achieved high accuracy in predicting Aβ status across different severity groups (AUCs 0.754-0.864).

Conclusions:

• A cost-efficient neuropsychological and demographic model can predict Aβ positivity.
• This offers a potential non-invasive screening tool for AD biomarker evidence.
• Facilitates recruitment for secondary prevention trials in non-demented populations.