A data-driven regional amyloid PET score predicts cognitive decline beyond Centiloid

A fixed nine-region Regional Amyloid PET Score (RAPS) predicts faster CDR-SB decline in amyloid-positive Alzheimer’s disease, with better rapid-decliner discrimination than Centiloid (AUC 0.813 vs 0.708) in ADNI.

• Why it matters: Centiloid summarizes amyloid PET globally, but regional deposition patterns may carry additional prognostic signal for the rate of cognitive decline in Alzheimer’s disease.
• What they did: RAPS was trained on 433 amyloid-positive ADNI participants using machine learning with bootstrap stability selection across 82 FreeSurfer regions to model CDR-SB slope, then applied without retraining to ADNI [18F]florbetaben (N = 71) and externally validated in NACC SCAN (N = 1531) and OASIS-3 (N = 428).
• The result: RAPS stayed associated with progression after adjusting for Centiloid in both external cohorts, and a cross-cohort meta-analysis favored RAPS over Centiloid alone (pooled ΔAUC +0.069; I2 = 1.6%).

AI summary of the abstract · may contain errors · not medical advice · read the paper.