Radiomics-based analysis of CT imaging for the preoperative prediction of invasiveness in pure ground-glass nodule lung adenocarcinomas

Insights into Imaging

Table 3 Diagnostic performance of three models for the prediction of pGGN invasiveness

Cohort	Model	AUC	95% CI	Sensitivity	Specificity	Accuracy
Training cohort	Clinical-radiographic model	0.729	0.661–0.798	0.817	0.576	0.638
	Radiomics model	0.854	0.804–0.905	0.883	0.663	0.720
	Combined model	0.856	0.804–0.908	0.850	0.727	0.759
Testing cohort	Clinical-radiographic model	0.652	0.523–0.782	0.680	0.452	0.510
	Radiomics model	0.846	0.766–0.927	0.840	0.678	0.729
	Combined model	0.859	0.781–0.937	0.880	0.685	0.735
Validation cohort	Clinical-radiographic model	0.692	0.542–0.843	0.682	0.600	0.635
	Radiomics model	0.814	0.699–0.928	0.955	0.668	0.769
	Combined model	0.765	0.638–0.893	0.773	0.656	0.762