Table 2. Diagnostic performance of each machine-learning model in the training cohort for the IL and IWA models.

Model	Method	AUC (95% CI)	Accuracy	Sensitivity	Specificity	PPV	NPV
IL	SVC	0.852 (0.786–0.909)	0.700	1.000	0.022	0.698	1.000
	DT	0.847 (0.779–0.899)	0.773	0.788	0.739	0.872	0.607
	RF	0.863 (0.805–0.918)	0.773	0.798	0.717	0.865	0.611
	KNN	0.849 (0.785–0.910)	0.813	0.933	0.543	0.822	0.781
	LR	0.853 (0.786–0.908)	0.753	0.769	0.717	0.860	0.579
	GBM	0.823 (0.714–0.918)	0.933	0.981	0.826	0.927	0.950
	MLP	0.865 (0.804–0.916)	0.807	0.904	0.587	0.832	0.730
IWA	SVC	0.885 (0.819–0.938)	0.800	0.990	0.370	0.780	0.944
	DT	0.875 (0.814–0.925)	0.820	0.856	0.739	0.881	0.694
	RF	0.896 (0.840–0.948)	0.833	0.827	0.848	0.925	0.684
	KNN	0.844 (0.777–0.906)	0.807	0.952	0.478	0.805	0.815
	LR	0.886 (0.830–0.941)	0.800	0.827	0.739	0.878	0.654
	GBM	0.826 (0.738–0.899)	0.880	0.875	0.891	0.948	0.759
	MLP	0.927 (0.875–0.966)	0.847	0.923	0.674	0.865	0.795

AUC, area under the curve; CI, confidence interval; DT, decision tree; GBM, gradient boosting machines; IL, infarct lesion; IWA, combination of infarct lesion, affected hemisphere, and whole brain; KNN, k-nearest neighbor; LR, logistic regression; MLP, multilayer perceptron; PPV, positive predictive value; NPV, negative predictive value; RF, random forest; SVC, support vector classification.