Prediction of Amyloid β-Positivity with both MRI Parameters and Cognitive Function Using Machine Learning

Area of Science:

• Neurology
• Radiology
• Biomedical Engineering

Background:

• Amyloid-beta (Aβ) positivity is a hallmark of Alzheimer's disease (AD) and mild cognitive impairment (MCI).
• Predicting Aβ status non-invasively is crucial for early diagnosis and intervention.
• Magnetic Resonance Imaging (MRI) offers potential biomarkers for Aβ detection.

Purpose of the Study:

• To identify MRI markers for predicting Aβ-positivity in MCI and AD patients.
• To compare MRI marker differences between Aβ-positive (Aβ [+]) and Aβ-negative groups.
• To evaluate the efficacy of machine learning (ML) in predicting Aβ-positivity.

Main Methods:

• 139 MCI and AD patients underwent amyloid PET-CT and brain MRI.
• Visual assessment included Fazekas scale for white matter hyperintensity (WMH) and cerebral microbleeds (CMB).
• Quantitative MRI analysis measured WMH and regional brain volumes; ML models (SVM, logistic regression) were employed.

Main Results:

• Aβ (+) patients showed higher WMH and CMB scores (p=0.02, p=0.04).
• Reduced volumes in hippocampus, entorhinal cortex, and precuneus were observed in Aβ (+) individuals (p<0.05).
• Increased third ventricle volume was noted in Aβ (+) patients (p=0.002); ML achieved 81.1% accuracy.

Conclusions:

• ML models integrating mini-mental state examination (MMSE) and regional brain volumes effectively predict Aβ-positivity.
• Third ventricle and hippocampal volumes are significant MRI predictors.
• This approach shows promise for non-invasive Aβ status prediction in clinical settings.