Predicting conversion of brain β-amyloid positivity in amyloid-negative individuals

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Alzheimer's disease (AD) is characterized by cortical β-amyloid (Aβ) plaques.
• Predicting conversion from Aβ-negative to Aβ-positive status is critical for early AD intervention.
• Current diagnostic approaches focus on established Aβ positivity, necessitating predictive tools for preclinical stages.

Purpose of the Study:

• To develop an artificial intelligence-based classifier for predicting Aβ conversion in individuals.
• To identify key predictors of Aβ positivity conversion using baseline demographic, genetic, and imaging data.
• To validate the predictive model's performance in an independent cohort.

Main Methods:

• Utilized data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort for initially Aβ-negative individuals.
• Developed an artificial neural network classifier incorporating age, gender, APOE ε4 genotype, and PET-derived SUVRs.
• Employed 10-fold cross-validation and external validation at Samsung Medical Center (SMC) for model assessment.

Main Results:

• The classifier incorporating global and regional SUVRs achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.841 on the ADNI dataset.
• External validation using the SMC cohort demonstrated a high predictive performance with an AUROC of 0.900.
• Models incorporating positron emission tomography (PET) imaging data significantly outperformed those without.

Conclusions:

• Successfully developed and validated AI-driven prediction models for Aβ positivity conversion.
• These models can aid in identifying individuals at high risk for AD progression.
• The findings support the potential for early screening and intervention in AD.