Biomarkers

Area of Science:

• Neuroscience
• Medical Imaging
• Population Health

Background:

• Alzheimer's Disease (AD) affects millions, with significant demographic disparities.
• Understanding ethnic variations in AD biomarkers is critical for equitable diagnostics.
• Current neuroimaging research needs to address population-specific differences.

Purpose of the Study:

• To investigate ethnic variations in neuroimaging biomarkers for Alzheimer's Disease.
• To evaluate the performance of machine learning models in distinguishing cognitive impairment across ethnic groups.
• To identify key neuroimaging features predictive of cognitive status in diverse populations.

Main Methods:

• Analysis of MRI, DTI, and PET scan data from 2,950 participants across African American, Hispanic, and Non-Hispanic White groups.
• Utilized Support Vector Machine (SVM) with SHapley Additive exPlanations (SHAP) for feature importance evaluation.
• Distinguished between cognitively impaired (CI) and cognitively unimpaired (CU) individuals.

Main Results:

• Machine learning model performance varied significantly by ethnicity: Hispanic (99.04% AUC), NHW (82.19%), and AA (75.44%).
• PET-Tau Posterior Cingulate SUVR was a consistent predictor across all groups.
• Ethnic-specific secondary biomarkers identified include R bankssts thickness (AA), L isthmuscingulate thickness (Hispanic), and AB FBB frontal SUVR (NHW).

Conclusions:

• Distinct neuroimaging biomarker patterns exist across ethnic groups in Alzheimer's Disease.
• Tau-PET measurements are consistently important, but biomarker sensitivity varies by population.
• Development of ethnically sensitive diagnostic tools and further research into population-specific AD manifestations are essential.