Table 3.

Performance of prediction models for predicting subtypes in NSCLC

Training set	AUC (95% CI)	Sen (%)	Spe (%)	Acc (%)
Clin-Lab Model	0.887 (0.843–0.931)	78.57	88.75	80.91
PET-Rad Model	0.835 (0.780–0.890)	90.00	62.50	78.64
CT-Rad Model	0.784 (0.733–0.855)	69.29	81.25	75.00
Combined Model	0.932 (0.900–0.964)	96.25	95.00	84.09
Validation set	AUC (95% CI)	Sen (%)	Spe (%)	Acc (%)
Clin-Lab Model	0.860 (0.789–0.931)	80.65	76.56	72.63
PET-Rad Model	0.740 (0.639–0.840)	83.87	75.00	66.32
CT-Rad Model	0.710 (0.606–0.815)	70.97	60.94	68.42
Combined Model	0.901 (0.840–0.957)	93.55	81.25	85.95

Clin-Lab Clinical-Laboratory, PET-Rad positron emission tomography-radiomics, CT-Rad computed tomography-radiomics, AUC area under the receiver operating curve, CI confidence interval, Sen sensitivity, Spe specificity, Acc accuracy