Noninvasive Prediction of TERT Promoter Mutations in High-Grade Glioma by Radiomics Analysis Based on Multiparameter MRI

Table 4

Performance of 4 models for TERT promoter mutations prediction.

	Training cohort							Validation cohort
	AUC (95% CI)	SEN	SPE	ACC	PPV	NPV	Cutoff	AUC (95% CI)	SEN	SPE	ACC	PPV	NPV	Cutoff
Model A	0.955 (0.899-0.979)	0.947	0.840	0.886	0.818	0.955	-0.652	0.889 (0.746-0.959)	0.750	0.909	0.842	0.857	0.833	-0.652
Model B	0.917 (0.840-0.959)	0.973	0.745	0.841	0.735	0.974	-1.14	0.868 (0.668-0.923)	0.813	0.818	0.816	0.765	0.857	-1.14
Model C	0.841 (0.802-0.913)	0.919	0.725	0.807	0.708	0.925	-0.798	0.747 (0.586-0.891)	0.750	0.727	0.737	0.667	0.800	-0.798
Model D	0.832 (0.798-0.919)	0.914	0.717	0.795	0.681	0.927	24.899	0.714 (0.608-0.829)	0.867	0.652	0.737	0.619	0.882	24.899

Abbreviations: Model A: Age+Lac+Cho/Cr+Radscore +CNV; Model B: Age+Lac+Cho/Cr+Radscore; Model C: Radscore; Model D: CNV; AUC: area under the curve; SEN: sensitivity; SPE: specificity; ACC: accuracy; PPV: positive predictive value; NPV: negative predictive value; CI: confidence intervals. The bootstrap resampling method was adopted for 95% CI and the significance test of AUC (). The cutoff value was determined based on the output value of the radiomics nomogram in the training cohort.