Validity of a multiphase CT-based radiomics model in predicting the Leibovich risk groups for localized clear cell renal cell carcinoma: an exploratory study

Insights into Imaging

Table 2 Results of four radiomics signatures’ predictive ability for predicting the Leibovich risk groups in three cohorts

Model	Cohort	AUC (95% CI)	Sensitivity	Specificity	Accuracy	Cutoff
U radiomics signature	Training	0.857 (0.803–0.911)	0.726	0.892	0.846	− 0.640
	Validation	0.846 (0.777–0.915)	0.821	0.733	0.763	− 0.910
	External testing	0.801 (0.664–0.939)	0.722	0.769	0.750	− 0.746
A radiomics signature	Training	0.849 (0.795–0.904)	0.726	0.866	0.828	− 0.730
	Validation	0.832 (0.758–0.907)	0.821	0.773	0.789	− 0.870
	External testing	0.803 (0.661–0.946)	0.722	0.846	0.795	− 0.717
V radiomics signature	Training	0.856 (0.804–0.909)	0.753	0.845	0.820	− 0.889
	Validation	0.832 (0.758–0.905)	0.795	0.773	0.781	− 1.004
	External testing	0.823 (0.697–0.948)	0.778	0.769	0.773	− 1.215
T radiomics signature	Training	0.862 (0.809–0.914)	0.753	0.866	0.835	− 0.794
	Validation	0.853 (0.785–0.921)	0.872	0.733	0.781	− 0.999
	External testing	0.837 (0.714–0.959)	0.765	0.852	0.818	− 0.589

AUC Area under the receiver operating characteristic curve, CI Confidence interval, U Unenhanced phase, A Arterial phase, V Portal-venous phase, T triphasic