Prediction of lymph node metastasis in operable cervical cancer using clinical parameters and deep learning with MRI data: a multicentre study

Insights into Imaging

Table 3 Diagnostic efficacy of clinical model, DL model, and hybrid model

Model	Training cohort (n = 225)				Internal validation cohort (n = 98)				External validation cohort (n = 69)
Model	Sensitivity	Specificity	AUC	p	Sensitivity	Specificity	AUC	p	Sensitivity	Specificity	AUC	p
M1	0.580	0.840	0.736 (0.736–0.838)		0.727	0.646	0.690 (0.609–0.820)		0.600	0.870	0.732 (0.567–0.839)
M2	0.692	0.818	0.757 (0.669–0.802)		0.693	0.821	0.714 (0.576–0.803)		0.619	0.816	0.765 (0.591–0.873)
M3	0.747	0.813	0.838 (0.783–0.893)		0.750	0.800	0.764 (0.665–0.862)		0.85	0.730	0.835 (0.721–0.948)
M1 vs. M2				0.650				0.750				0.650
M2 vs. M3				< 0.001				0.280				0.040
M1 vs. M3				< 0.001								0.016