Predicting future amyloid conversion: the role of baseline amyloid distribution, genetic and cognitive resilience

Alzheimer's disease (AD) is characterized by early accumulation of amyloid beta (Aβ) prior to cognitive decline. Although Aβ plays a normal role in brain function, excessive buildup is a key early indicator of AD. The transition from Aβ-negative to Aβ-positive reflects disease progression, but patterns predicting this shift remain unclear. Using longitudinal Aβ PET imaging data from the Alzheimer's Disease Neuroimaging Initiative (ADNI), we developed a survival-based machine learning model to estimate the timing of Aβ conversion. The model showed strong predictive accuracy (concordance index = 0.845). SHAP analysis identified higher baseline Aβ burden in frontoparietal and striatal regions, particularly the rostral and caudal middle frontal cortex, as predictors of faster conversion. APOE4 homozygosity and lower baseline cognitive performance were also associated with earlier Aβ positivity. Kaplan-Meier analysis estimated a median transition time of 4.06 years. These findings support machine learning for early AD risk prediction and timely intervention planning.