Table 2.

Synthesis of the 5-fold cross-validation classifiers for focal lesional epilepsy vs. non-epilepsy classification, with AUC, accuracy, sensitivity and specificity assessment metrics. In a validation vs. training performance check, we report the difference between AUC scores in training and in validation.

#	Model	AUC	Accuracy	Sensitivity	Specificity	Val vs. Train
1	SVM (linear) and ANOVA	0.736 ± 0.044	0.689 ± 0.052	0.856 ± 0.031	0.531 ± 0.102	0.0426
2	SVM (gaussian) and PCA	0.730 ± 0.030	0.705 ± 0.020	0.692 ± 0.084	0.717 ± 0.061	0.0255
3	SVM (linear) and Corr-ANOVA	0.735 ± 0.037	0.689 ± 0.034	0.672 ± 0.162	0.704 ± 0.118	0.0345
4	MLP and ANOVA	0.732 ± 0.047	0.717 ± 0.051	0.749 ± 0.116	0.688 ± 0.117	0.0082
5	MLP and PCA	0.734 ± 0.026	0.698 ± 0.018	0.674 ± 0.141	0.724 ± 0.141	0.0576
6	RFC	0.745 ± 0.022	0.714 ± 0.012	0.730 ± 0.069	0.699 ± 0.048	0.254
7	RFC and ANOVA	0.718 ± 0.047	0.686 ± 0.050	0.767 ± 0.080	0.609 ± 0.161	0.282
8	RFC and PCA	0.706 ± 0.022	0.698 ± 0.023	0.655 ± 0.117	0.742 ± 0.140	0.272
9	LogReg and ANOVA	0.734 ± 0.048	0.698 ± 0.057	0.775 ± 0.209	0.629 ± 0.205	0.0344
10	LogReg and PCA	0.729 ± 0.033	0.698 ± 0.023	0.704 ± 0.107	0.693 ± 0.131	0.0028
11	LogReg and Corr-ANOVA	0.752 ± 0.038	0.717 ± 0.037	0.634 ± 0.189	0.795 ± 0.153	0.0173