Predicting Amyloid-β Levels in Amnestic Mild Cognitive Impairment Using Machine Learning Techniques

Area of Science:

• Neuroimaging and Biomarkers
• Machine Learning in Medicine
• Alzheimer's Disease Research

Background:

• Amyloid-beta (Aβ) positivity via PET imaging is crucial for Alzheimer's disease (AD) clinical trial enrollment, especially for amyloid-targeted therapies.
• Predicting Aβ status pre-PET can reduce patient burden and trial costs.

Purpose of the Study:

• To evaluate a machine learning model's performance in predicting individual Aβ positivity risk.
• Utilizing PET imaging as the gold standard for Aβ status confirmation.

Main Methods:

• Data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) amnestic mild cognitive impairment (aMCI) cohort were used.
• Models incorporated demographics, ApoE4 status, neuropsychological tests (NP), MRI volumetrics, and cerebrospinal fluid (CSF) biomarkers.

Main Results:

• Models with NP and MRI measures achieved AUCs of 0.74 and 0.72, respectively.
• Combining NP and MRI did not enhance prediction.
• Models incorporating CSF biomarkers demonstrated superior performance with AUCs ranging from 0.89 to 0.92.

Conclusions:

• Predictive models are effective in identifying individuals with aMCI who are likely to be amyloid-positive.
• This approach aids in streamlining patient selection for AD clinical trials.