Table 4.

Performance of the models for predicting PTC recurrence in the validation cohort

Model	Classifier	AUC	95% CI		ACC	SPE	SEN
			Low	High
Radiomics models	LR	0.706	0.601	0.812	0.739	0.908	0.371
	SVM	0.710	0.604	0.816	0.721	0.961	0.200
	KNN	0.617	0.499	0.735	0.676	0.842	0.314
	NN	0.567	0.454	0.679	0.685	1.000	0
Clinical models	LR	0.709	0.597	0.821	0.748	0.895	0.429
	SVM	0.669	0.556	0.782	0.694	0.934	0.171
	KNN	0.642	0.522	0.763	0.739	0.855	0.486
	NN	0.665	0.552	0.778	0.730	0.895	0.371
Combined models	LR	0.746	0.640	0.852	0.739	0.895	0.400
	SVM	0.754	0.649	0.859	0.766	0.921	0.429
	KNN	0.669	0.552	0.785	0.730	0.842	0.486
	NN	0.711	0.607	0.816	0.766	0.947	0.371

ACC: Accuracy; AUC: Area under the receiver operating characteristic curve; CI: Confidence interval; KNN: K-nearest neighbor; LR: Logistic regression; NN: Neural network; SEN: Sensitivity; SPE: Specificity; SVM: Support vector machine