Table 6 ACS classification performance on the external cohort. Reported is \(\mu \pm 2\sigma\) where the samples come from 5-fold stratified cross-validation on the training set.
Method | Prevalence | Coverage | Sensitivity | Specificity | PPV | NPV | AUROC | Accuracy |
|---|---|---|---|---|---|---|---|---|
GBDT | \(18 \pm 0\) | \(100 \pm 0\) | \(90 \pm 4\) | \(86 \pm 3\) | \(59 \pm 4\) | \(97 \pm 1\) | \(88 \pm 1\) | \(87 \pm 2\) |
GBDT+SC | \(16 \pm 2\) | \(75 \pm 5\) | \(94 \pm 2\) | \(96 \pm 4\) | \(81 \pm 10\) | \(99 \pm 0\) | \(95 \pm 2\) | \(95 \pm 3\) |
DeepEnsemble | \(18 \pm 0\) | \(100 \pm 0\) | \(87 \pm 6\) | \(84 \pm 6\) | \(55 \pm 8\) | \(97 \pm 1\) | \(85 \pm 2\) | \(84 \pm 4\) |
DeepEnsemble+SC | \(19 \pm 2\) | \(76 \pm 7\) | \(92 \pm 4\) | \(90 \pm 7\) | \(70 \pm 15\) | \(98 \pm 1\) | \(91 \pm 4\) | \(91 \pm 6\) |
MCDropout | \(18 \pm 0\) | \(100 \pm 0\) | \(75 \pm 26\) | \(90 \pm 11\) | \(65 \pm 16\) | \(95 \pm 5\) | \(83 \pm 9\) | \(87 \pm 6\) |
MCDropout+SC | \(14 \pm 7\) | \(79 \pm 5\) | \(82 \pm 24\) | \(96 \pm 8\) | \(79 \pm 21\) | \(98 \pm 2\) | \(89 \pm 10\) | \(94 \pm 6\) |
