A clinical–radiomics model based on noncontrast computed tomography to predict hemorrhagic transformation after stroke by machine learning: a multicenter study

Insights into Imaging

Table 6 Performance of the three models

Model	Training cohort						Internal validation cohort						External validation cohort
Model	AUC	ACC	SEN	SPE	PPV	NPV	AUC	ACC	SEN	SPE	PPV	NPV	AUC	ACC	SEN	SPE	PPV	NPV
Clinical	0.996	0.969	0.981	0.959	0.952	0.984	0.898	0.844	0.854	0.837	0.814	0.872	0.911	0.833	0.788	0.872	0.839	0.829
Radiomics	0.999	0.972	0.994	0.954	0.947	0.995	0.922	0.878	0.963	0.837	0.826	0.932	0.883	0.847	0.727	0.949	0.923	0.804
Combined	0.995	0.980	0.994	0.9691	0.964	0.995	0.950	0.900	0.951	0.857	0.848	0.945	0.942	0.861	0.758	0.949	0.926	0.822