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

Table 3

Experiment 3: mean correlations of subject-specific FC spatial maps with (1) the task-related GLM map for the same subject (“Task-Related” column) and (2) the group IC of interest (“Group IC” column). Results of a paired 𝑡-test comparison of the mean correlations are shown in the rightmost columns. Higher correlations with the task-related maps than with the group IC of interest are an indication that the results reflect the individual-subject data without being overly biased by spatial features of the group IC. The group ICs were generated with PICA or Infomax. These results were based on group ICAs that produced 14 components.

Method of Generating Maps	Group ICA Source	Mean Correlation With Subject-Specific Map (95% Confidence Interval)		Same = Seed Null Hypothesis Test
		Task-Related	Group IC	𝑡	df	𝑃
Single-voxel seed	PICA	0.84 (0.77–0.89)	0.52 (0.48–0.55)	7.7	13	<.001
9-voxel ROI seed	PICA	0.85 (0.79–0.89)	0.53 (0.49–0.57)	8.2	13	<.001
40-voxel ROI seed	PICA	0.86 (0.80–0.90)	0.54 (0.49–0.58)	8.9	13	<.001
Dual regression, 1 spatial map	PICA	0.88 (0.84–0.91)	0.55 (0.52–0.58)	10.0	13	<.001
Dual regression, all spatial maps	PICA	0.53 (0.47–0.59)	0.59 (0.56–0.61)	−2.3	13	.04
Back-reconstruction	Infomax	0.30 (0.22–0.38)	0.62 (0.58–0.66)	−10.4	13	<.001
Dual regression, 1 spatial map	Infomax	0.81 (0.73–0.87)	0.43 (0.40–0.46)	6.6	13	<.001