Skip to main content

Table 2 Beta-amyloid positivity classifier performances. Three different models were developed with different feature combinations

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

Dataset Model Featuresa AUROC (95% CI) AUPRC (95% CI) Sensitivity (95% CI) Specificity (95% CI) PPV (95% CI) NPV (95% CI)
A. ADNI (development set) Model 1 Age, gender, APOE ε4 carriers (3) 0.674 (0.666–0.683) 0.374 (0.364–0.384) 0.606 (0.595–0.616) 0.692 (0.668–0.715) 0.373 (0.355–0.392) 0.853 (0.849–0.858)
Model 2 Age, gender, APOE ε4 carriers, global SUVR (4) 0.814 (0.806–0.821) 0.549 (0.534–0.564) 0.744 (0.707–0.780) 0.727 (0.690–0.764) 0.454 (0.430–0.478) 0.905 (0.896–0.913)
Model 3 Age, gender, APOE ε4 carriers, global SUVR, regional SUVR (44) 0.841 (0.832–0.849) 0.627 (0.610–0.645) 0.600 (0.581–0.619) 0.869 (0.862–0.875) 0.579 (0.562–0.597) 0.878 (0.873–0.884)
B. SMC (external validation set) Model 2 Age, gender, APOE ε4 carriers, global SUVR (4) 0.900 0.625 1.000 0.700 0.526 1.000
  1. Measurements were described as averages and the 95% confidence interval of 10 times repeated 10-fold cross-validation with the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset. The Samsung Medical Center (SMC) dataset was used for external validation to test the Model 2.
  2. AUROC area under the receiver operating characteristic, AUPRC area under the precision-recall curve, CI confidence interval, PPV positive predictive value, NPV negative predictive value, ADNI Alzheimer’s Disease Neuroimaging Initiative, SMC Samsung Medical Center, APOE ε4 apolipoprotein E ε4, SUVR standardized uptake value ratio
  3. aNumbers in parentheses indicate the total number of features used in each model