Table 3.

Performance metrics of the best model, SVM with features selected by RFE

	Sensitivity (Actual)	Specificity	Balanced Accuracy
Whole Sample (AUC=0.962)
Sensitivity of 0.95	0.950	0.874	0.912
Sensitivity of 0.90	0.900	0.906	0.903
Sensitivity of 0.85	0.850	0.938	0.894
Sensitivity of 0.80	0.800	0.951	0.876
Sensitivity of 0.75	0.750	0.957	0.853
Best Balanced Accuracy	0.956	0.871	0.913
Only MCI (AUC=0.914)
Sensitivity of 0.95	0.951	0.734	0.843
Sensitivity of 0.90	0.901	0.822	0.862
Sensitivity of 0.85	0.851	0.877	0.864
Sensitivity of 0.80	0.801	0.882	0.842
Sensitivity of 0.75	0.751	0.883	0.817
Best Balanced Accuracy	0.880	0.867	0.874
Only PreMCI (AUC=0.994)
Sensitivity of 0.95	0.955	0.960	0.958
Sensitivity of 0.90	0.904	0.966	0.935
Sensitivity of 0.85	0.853	0.966	0.910
Sensitivity of 0.80	0.802	0.966	0.884
Sensitivity of 0.75	0.751	0.966	0.859
Best Balanced Accuracy	1.000	0.960	0.980

AUC, area under the receiving operating curve; MCI, mild cognitive impairment; PreMCI, premild cognitive impairment.