Skip to main content

Table 2 Classification performance for Aβ PET positivity on ADNI, KBASE, and Severance hospital datasets

From: Deep learning-based amyloid PET positivity classification model in the Alzheimer’s disease continuum by using 2-[18F]FDG PET

 

Mean

(95% CI)

AUC value

Accuracy

Sensitivity

Specificity

F1-score

Internal validation (ADNI and KBASE)

All

0.811

(0.803, 0.819)

0.733

(0.726, 0.740)

0.678

(0.664, 0.691)

0.785

(0.771, 0.799)

0.709

(0.701, 0.717)

CU

0.717

(0.705, 0.729)

0.748

(0.737, 0.758)

0.381

(0.352, 0.410)

0.870

(0.857, 0.883)

0.422

(0.398, 0.446)

MCI

0.757

(0.746, 0.768)

0.682

(0.672, 0.692)

0.651

(0.636, 0.666)

0.718

(0.698, 0.738)

0.690

(0.680, 0.701)

AD

0.816

(0.789, 0.843)

0.862

(0.848, 0.875)

0.950

(0.939, 0.960)

0.336

(0.283, 0.390)

0.921

(0.912, 0.929)

External validation (Severance hospital)

All

0.798

(0.789, 0.807)

0.690

(0.681, 0.699)

0.768

(0.748, 0.789)

0.612

(0.586, 0.639)

0.712

(0.703, 0.721)

MCI

0.769

(0.762, 0.776)

0.698

(0.691, 0.705)

0.702

(0.688, 0.715)

0.694

(0.675, 0.714)

0.718

(0.711, 0.725)

Demented participants

0.806

(0.803, 0.809)

0.599

(0.573, 0.624)

0.868

(0.850, 0.886)

0.462

(0.414, 0.509)

0.598

(0.587, 0.609)

  1. Aβ = β-amyloid; CI = Confidence interval; AUC = Area under curve