Table 2.

Performance comparison of the PPC-SSAE classifier with other classifiers.

	Sensitivity		Specificity		Accuracy
	Mean, %	St. Dev., %	Mean, %	St. Dev., %	Mean, %	St. Dev., %
PPC-SSAE	93.1	5.0	99.5	1.8	96.8	6.4
Spectra-SSAE	90.1	21.0***	98.4	5.0**	95.5	10.0*
PPC-SVM	81.1	32.0**	97.1	14.0***	90.0*	16.0**
PPC-Random Forest	78.0*	24.6***	91.5**	13.2***	85.6***	11.6**

PPC, power-to-power coupling; SSAE, stacked sparse autoencoder; SVM, support vector machine. PPC-SSAE, algorithm based on PPC as a classifying feature and the stacked sparse autoencoder as a classifier; Spectra-SSAE, algorithm based on the EEG spectrum as a classifying feature and the SSAE classifier; PPC-SVM, algorithm based on PPC as a classifying feature and the SVM classifier; PPC-Random Forest, algorithm based on PPC as a classifying feature and the Random Forest classifier. The PPC-SSAE algorithm is compared to all three other algorithms with statistically significant differences indicated by asterisks: p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***) (t-test for the comparison of the mean values, Bonferroni-corrected; Bartlett’s test for the comparison of variances, Matlab function vartestn).