Fig 1.
Ten dynamic lFCD maps and 4 dynamic ALFF maps were computed for each subject, session, and phase encoding direction using 5 different pipelines (see text). A total of 1600 lFCD and 640 ALFF maps covering the whole brain (white matter and cerebrospinal fluid regions were not masked out to assess the strength of the lFCD in these regions) with 2-mm isotropic resolution and 91×109×91 voxels were computed using 160 HCP datasets with “minimal preprocessing” [36] from the Q1 release. Smoothing was not used to preserve the high spatial resolution of the resting-state functional datasets.
Fig 2.
Temporal variability at the group level.
Average distributions of strength (lFCD; A) and standard deviation (SD; B) of the lFCD as well as their ratio (C) across subjects showing brain areas where these metrics had higher values than twice (A and B) their whole brain averages, superimposed on axial (right), sagittal (middle) and coronal (left) views of the cortical and subcortical gray matter template developed using the HCP structural scans (pipeline 4). lFCD, SD and SD/lFCD ratio maps were computed for each subject and averaged. Note that voxels in white matter and cerebrospinal fluid were not excluded and that the imaging threshold (twice the whole brain average) was the only criterion used to display the patterns. The scatter plot (D) demonstrates the lack of association between the relative temporal dynamics (SD/lFCD) and the strength of the lFCD hubs. Image voxels were sorted by the strength of the rescaled lFCD and averaged into bins of lFCD = 0.1, independently for lFCD, SD and for the SD-to-lFCD ratio.
Fig 3.
Temporal variability at the individual level.
(A) Exemplary series of dynamic lFCD maps (in voxels; i.e. without grand mean global scaling) from a typical resting state HCP dataset superimposed on an axial view of the T1 weighted brain structure (top) and lFCD time courses (colored lines) corresponding to four different voxels from gray matter regions (colored arrows). The standard deviation (SD; in voxels) maps in B and C quantify the temporal dynamics of the lFCD metric in the brain for a single individual. Pipeline 4.
Fig 4.
Average lFCD (A) and SD (C) maps across subjects with (G) and without (NG) GSN, with 0.08Hz or 0.15Hz low-pass filtering, and their statistical differences (two-sided t-score; B and D) superimposed on axial (right), sagittal (middle) and coronal (left) views of the cortical and subcortical gray matter template developed using the HCP structural scans.
Fig 5.
Dynamic association between lFCD and ALFF.
Typical single subject data (A-D) showing (A) exemplary time courses from a voxel in primary visual cortex (black arrow in C) for lFCD (red) and ALFF (blue) and their correlation as a function of time (B), as well as the spatial distribution of their temporal correlation coefficients in the brain (C and D), which after Fisher’s transformation (histograms) had normal distribution (red Gaussian curve fits). (E) Distribution of the average ALFF-lFCD correlation coefficients across subjects superimposed on three orthogonal views of the gray matter template.
Fig 6.
Effects of head motion on FC dynamics.
(A) Mean Fisher’s z-score values (top row) and their statistical significance across subjects superimposed on three orthogonal views of a gray matter template (t-test; bottom row), demonstrating the linear correlation between framewise displacements (FD) and the FC metrics (lFCD and ALFF) in the brain as a function of time. (B) Three orthogonal views showing the distribution in the brain of the group mean Fisher’s z-scores from partial correlation analyses (“Motion removed”; top row) and from standard Pearson correlation analyses (“With motion”; bottom row).
Fig 7.
lFCD dynamics: Gray matter specificity and test-retest reliability.
(A) Average sensitivity index for SD in subcortical and cortical gray matter and white matter, and average reproducibility and specificity indices across subjects for each of the processing pipelines in Fig 1. (B) Two-way mixed single measures intraclass correlation ICC(3,1) maps at 2-mm isotropic resolution depicting regional variability in test-retest reliability for SD (pipeline 4). (C) Scatter plot showing the linear association of the mean between-subject SD-differences (MBSD) across voxels in overlapping (No GM) and non-overlapping (GM) gray matter for all potential pairs of subjects (pipeline 4). Error bars are standard deviations.
Fig 8.
Statistical maps from Kolmogorov-Smirnov test (see methods) averaged across subjects, highlighting brain regions where the cumulative distribution function of the time-varying lFCD differed from that of a uniform random variable (A). Average entropy across subjects superimposed on B) lateral and medial views of the brain surface and C) three orthogonal views of the brain. The color bar displays the average entropy, H, across subjects relative to the maximal entropy, Hmax = log2(92), effectively attained with the concatenation of 4 time series (4 × 23 temporal windows) from each subject. D) Exponential saturation of H with lFCD in typical subject data.