Table 3.

p-values of McNemar's test of linear-SVM models based on rs-fMRI data and LBP-TOP features using non-linear registration.

Parcellations	Radius	Non-BF and LBP-TOP	BF and LBP-TOP	BF and non-BF
AAL	R1	0.0056	0.0102	0.8467
	R2	0.0633	0.1052
	R3	0.0043	0.0106
CC200	R1	0.1586	0.0454	0.3619
	R2	0.0093	0.0018
	R3	0.0013	0.0002

The p-values without under lines denote the accuracies of LBP-TOP features are bigger than rs-fMRI data, and the accuracies of rs-fMRI data filtered by a bandpass filter are bigger than rs-fMRI data not filtered. The underlined p-values show the inverse relationship. Bold p-values correspond to significant results (p-value < 0.05). Abbreviations are as follows: AAL, automated anatomical labeling template; CC200, spatially constrained parcellation based on rs-fMRI; BF, rs-fMRI data filtered by a bandpass filter (0.009 Hz–0.08 Hz); and non-BF, rs-fMRI data not filtered by a bandpass filter.