Hybrid ICA-Seed-Based Methods for fMRI Functional Connectivity Assessment: A Feasibility Study

Table 4

Experiment 3: mean percentages of thresholded task-related GLM maps that are covered by the thresholded subject-specific FC maps generated with each method below. The right-most column gives percentages after adjusting the z-score used to threshold the FC maps so that number of false positives equals number of false negatives. For the BR-I method, adjusting the threshold helped to compensate for what might have been an overly conservative thresholding, but the results for this method were still significantly lower those than for the other six methods used.

Method of Generating Maps	Group ICA Source	Percent of Task-Related Map Covered (95% Confidence Interval)
		Before Threshold Adjustment	After Threshold Adjustment
Single-voxel seed (SV)	PICA	57 (46–69)	63 (54–73)
9-voxel ROI seed (FV)	PICA	61 (51–71)	64 (55–74)
40-voxel ROI seed (MV)	PICA	65 (55–75)	66 (56–75)
Dual regression, 1 spatial map (DRS)	PICA	78 (70–85)	71 (64–78)
Dual regression, all spatial maps (DRA)	PICA	28 (20–36)	50 (43–57)
Back-reconstruction (BR-I)	Infomax	9 (6–12)	29 (21–36)
Dual regression, 1 spatial map (DRS-I)	Infomax	59 (47–72)	56 (45–68)