Development of a combined radiomics and CT feature-based model for differentiating malignant from benign subcentimeter solid pulmonary nodules

European Radiology Experimental

Table 3 Diagnostic performance of the three prediction models in training and testing set

Group	Model	AUC (95% CI)	Accuracy	Sensitivity	Specificity	PPV	NPV
Training set	Clinical model	0.920 (0.885–0.956)	0.850	0.800	0.897	0.880	0.825
	Radiomics model	0.907 (0.875–0.938)	0.824	0.861	0.789	0.795	0.856
	Combined model	0.942 (0.918–0.966)	0.870	0.880	0.861	0.858	0.883
Testing set	Clinical model	0.835 (0.758–0.912)	0.745	0.604	0.880	0.829	0.698
	Radiomics model	0.900 (0.867–0.932)	0.815	0.848	0.783	0.788	0.844
	Combined model	0.930 (0.902–0.957)	0.858	0.867	0.849	0.846	0.870

AUC Area under the receiver operating characteristic curve, CI Confidence interval, NPV Negative predictive value, PPV Positive predictive value