Table 3 Fazekas rating performance.

From: Automated rating of Fazekas scale in fluid-attenuated inversion recovery MRI for ischemic stroke or transient ischemic attack using machine learning

Task

Agreement

Binary categorization

Q kappa

categories

AUROC†

AUPRC†

Internal validation set

 Regression

0.904

0 vs. 1, 2, 3

1.000 [1.000–1.000]

1.000 [1.000–1.000]

0, 1 vs. 2, 3

0.969 [0.956–0.981]

0.970 [0.955–0.985]

0, 1, 2 vs. 3

0.940 [0.918–0.961]

0.840 [0.786–0.894]

 Classification

0.917

0 vs. 1, 2, 3

0.997 [0.993–1.000]

0.999 [0.999–0.999]

0, 1 vs. 2, 3

0.975 [0.963–0.987]

0.977 [0.964–0.990]

0, 1, 2 vs. 3

0.957 [0.937–0.978]

0.900 [0.862–0.937]

Internal test set

 Regression

0.951

0 vs. 1, 2, 3

1.000 [1.000–1.000]

1.000 [1.000–1.000]

0, 1 vs. 2, 3

0.987 [0.975–0.999]

0.988 [0.975–0.998]

0, 1, 2 vs. 3

0.987 [0.973–1.000]

0.970 [0.954–0.984]

 Classification

0.956

0 vs. 1, 2, 3

1.000 [0.999–1.000]

1.000 [1.000–1.000]

0, 1 vs. 2, 3

0.991 [0.981–1.000]

0.992 [0.984–0.998]

0, 1, 2 vs. 3

0.991 [0.982–1.000]

0.977 [0.962–0.991]

External validation set

 Regression

0.898

0 vs. 1, 2, 3

0.957 [0.895–1.000]

0.997 [0.994–1.000]

0, 1 vs. 2, 3

0.982 [0.957–1.000]

0.984 [0.961–1.000]

0, 1, 2 vs. 3

1.000 [1.000–1.000]

1.000 [1.000–1.000]

 Classification

0.956

0 vs. 1, 2, 3

0.972 [0.922–1.000]

0.998 [0.995–1.000]

0, 1 vs. 2, 3

0.992 [0.978–1.000]

0.993 [0.983–1.000]

0, 1, 2 vs. 3

1.000 [1.000–1.000]

1.000 [1.000–1.000]

  1. †Values are expressed as mean [95% CI].
  2. AUPRC, area under the precision-recall curve; AUROC, area under the receiver operating characteristic curve; Q kappa, quadratic weighted kappa.
Back to article page