Table 3 Classification metrics of HTNet with ultrasound images and serologic markers as input.
From: Deep learning to diagnose Hashimoto’s thyroiditis from sonographic images
Classification metrics | Internal-testing set 1 (static-image, n = 945) | Internal-testing set 1 (static-image plus serologic markers, n = 945) | Internal-testing set 1 (video, n = 185) | Internal-testing set 1 (video plus serologic markers, n = 185) |
|---|---|---|---|---|
Accuracy (95% CI) | 0.861 (0.838–0.883) | 0.877 (0.855–0.897) | 0.832 (0.771–0.883) | 0.892 (0.838–0.933) |
Sensitivity (95% CI) | 0.765 (0.709–0.815) | 0.830 (0.779–0.873) | 0.846 (0.719–0.931) | 0.923 (0.815–0.979) |
Specificity (95% CI) | 0.899 (0.874–0.920) | 0.896 (0.870–0.918) | 0.827 (0.752–0.887) | 0.880 (0.812–0.930) |
