Fig. 1

Generation of meta-analytical task-fMRI maps for assessing cognitive-domain-specific tau-PET (A), surface rendering of regions of interests (ROIs) applied to tau-PET (B), and analysis flowchart (C). First, we compared the predictive accuracy of global tau-PET, temporal tau-PET, and cognitive-domain-specific tau-PET for predicting future cognitive decline using bootstrapped linear regressions. Note that separate regression models were run for each tau-PET ROI (CI). Second, we ran 1000 bootstrapped linear regression models in ADNI Aβ+ and extracted beta values for each variable of the regression model (CII), which were later used for assessing patient-specific cognitive composites in both ADNI Aβ+ and A05 Aβ+ (CIII). Specifically, we applied the 1000 bootstrapped linear model equations to subject level data to determine predicted cognitive change rates for each Aβ+ subject of the ADNI and A05 sample. Based on the rank of the predicted cognitive change rates, we then determined a tau-PET-informed personalized cognitive composite. Abbreviations: DV, dependent variable; IV, independent variable; CV, control variable; CR, change rate