Construction of a Diagnostic Model for Lymph Node Metastasis of the Papillary Thyroid Carcinoma Using Preoperative Ultrasound Features and Imaging Omics

Table 4

The predictive performance of lymph node metastasis by BLS feature learning.

Models	Cut-off	Sensitivity (95% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)	AUC (95% CI)	Accuracy (95% CI)
Model 4^a	0.500	0.823 (0.777–0.868)	0.966 (0.948–0.983)	0.941 (0.911–0.971)	0.892 (0.863–0.921)	0.984 (0.977–0.990)	0.909 (0.887–0.931)
Model 4^b	0.500	0.667 (0.584–0.749)	0.910 (0.866–0.953)	0.848 (0.778–0.919)	0.782 (0.724–0.841)	0.857 (0.811–0.902)	0.805 (0.759–0.850)
Model 5^a	0.465	0.959 (0.936–0.983)	0.973 (0.958–0.989)	0.959 (0.936–0.983)	0.973 (0.958–0.989)	0.995 (0.992–0.998)	0.968 (0.954–0.981)
Model 5^b	0.465	0.778 (0.705–0.850)	0.843 (0.788–0.899)	0.790 (0.719–0.862)	0.833 (0.777–0.890)	0.853 (0.806–0.901)	0.815 (0.771–0.860)

^aUsing the training set; ^busing the testing set. PPV: positive predictive value; NPV: predictive value; AUC: area under the curve; CI: confidence internal.