Table 5.

Prediction performance of three models in training and validation cohorts.

Model	AUC (95% CI)	Accuracy	Sensitivity	Specificity	PPV	NPV	Youden index
GLV
Training cohort	0.774 (0.669–0.878)	0.744	0.780	0.703	0.744	0.743	0.483
Validation cohort	0.736 (0.602–0.870)	0.698	0.734	0.690	0.715	0.729	0.424
PLV
Training cohort	0.729 (0.618–0.840)	0.679	0.688	0.692	0.733	0.711	0.380
Validation cohort	0.674 (0.528–0.820)	0.660	0.621	0.680	0.706	0.639	0.301
GPLV
Training cohort	0.868 (0.790–0.945)	0.808	0.854	0.757	0.795	0.824	0.611
Validation cohort	0.777 (0.652–0.903)	0.736	0.750	0.720	0.739	0.731	0.470

GLV, gross lesion volume; PLV, perilesional volume; GPLV, gross combined perilesional volume; PPV, positive predictive value; NPV, negative predictive value.