Table 7.

Results of the proposed UFA-Net with different atlases for brain ROI parcellation. Here, the UFA-Net method uses the AAL atlas and the UFA-Net-D uses the Dosenbach atlas. The UFA-Net-D-LB is a degenerated version of UFA-Net-D, without feature alignment.

Model	ACC (%)	AUC (%)	SEN (%)	SPE (%)	PRE (%)
UFA-Net-D-LB	50.27±2.85	50.23±3.94	49.19±8.44	51.35±8.96	50.35±3.48
UFA-Net-D	57.16±2.08	58.23±1.46	62.70±15.09	51.62±11.82	56.41 ± 1.21
UFA-Net	59.73±1.63	62.50±2.50	69.46±6.43	50.00±9.63	58.49±2.58