Velocity-nulling (VN) gradient in GE-EPI. (a) The diagram of pulse sequence. The VN gradient is highlighted in light blue. (b) The simulated signal attenuation against b values.

Empirical results of finger tapping task with different strength of draining-vein suppression from a single participant. (a) The paradigm of finger tapping task. (b) The illustration of the slab acquisition covering M1. (c) The slab image with 0.9-mm isotropic resolution in three orthogonal views. The 20 layers of M1 are color-coded as shown in the bottom-right panel. (d-g) The depth-dependent profiles of BOLD activation at M1 associated with b=0, 15, 30, 48. The statistical maps were corrected (uncorrected p < 0.01; corrected p < 0.05) and color-coded as indicated by the color bar.

Sub-mm fMRI with brain-wide coverage in one participant. (a-d) The reconstructed images with different combination of SMS factor and b value. (e-h) The activation maps and depth-dependent profiles associated with different scanning protocols. The number of volumes acquired during the motor task was maintained at 282 across all protocols.

Illustration of TE impact on activation maps and depth-dependent profiles. A total of four finger-tapping experimental sessions were conducted on a single participant with varying TEs: (a) 33 ms, (b) 38 ms, (c) 43 ms, and (d) 43 ms with the VN gradient. No VN gradient was applied in (a)-(c). The red rectangle in the left-most column highlighted an enlarged view centered on the M1 region. Within M1, a total of 20 layers are represented and color-coded in the bottom-left panel. The statistical maps were corrected (uncorrected p < 0.05; corrected p < 0.05) and color-coded as indicated by the color bar on the far right. The images are presented in radiology view. The bottom row presents the enlarged views of the depth-dependent profiles, while the whiskers indicate the standard deviation across voxels within each individual layer.

Influence of TE and phase regression on BOLD activation maps in a finger-tapping task. The rows represent the maps associated with TEs of 33 ms, 38 ms, and 43 ms, from top to bottom, respectively. Columns from left to right display the non-phase-regressed, phase-regressed, and difference maps, respectively. The right-most column depicts the absolute difference between the non-phase-regressed and phase-regressed maps. The activation maps were corrected statistically (uncorrected p < 0.05; corrected p < 0.05). The z values in each column are color-coded as indicated by the color bars.

The visual task paradigm and the corresponding results from a single participant. (a) The checkerboard flashed alternately on the left and right sides for 16 seconds, followed by a 16-second resting period. (b) The BOLD activation maps correspond to the visual stimuli in the left or right visual hemifield. (c) The BOLD activation maps in the LGN, presented with and without a VN gradient. The z-statistics are overlaid on the individual T1 image and color-coded as indicated by the color bar. (d) The HDRs in the LGN, compared without and with a VN gradient (b value of the VN gradient is 30). The shaded area highlights the period of visual stimulation. The blue and red traces denote the mean HDRs of BOLD signal change obtained using the protocols without and with VN gradient, respectively. The whiskers denote the standard deviation across trials.

Activation maps and depth-dependent profiles of BOLD responses across participants. (a) BOLD activation maps in the absence of a VN gradient, with the solid green contour delineating the primary motor cortex (M1). Dashed traces delineate superficial, middle, and deep cortical layers. (b) Activation maps with a VN gradient applied. (c) Depth-dependent profiles of BOLD response in the absence of a VN. The dark blue traces denote the group-averaged data. (d) Depth-dependent profiles with VN applied. The dark red traces denote the group-averaged profiles with the VN gradient applied. In both (c) and (d), the traces with light colors indicate individual profiles.

Directed functional connectivity across functional networks (N = 6). (a) Fisher’s z transformed FC matrices. The FC matrix on the left represents the Fisher’s z values without a VN gradient, while the matrix on the right corresponds to the Fisher’s z values with a VN gradient. (b) Statistical results of FC within the visual, sensorimotor, frontoparietal, and default-mode networks. The matrix’s lower triangular parts show uncorrected t-values, while the upper triangular parts show t-values that survived FDR correction (corrected p < 0.001). (c) Depth-dependent FC matrices for a representative ROI pair in each network. T-values are color-coded as indicated by the color bar. Red circles highlight directed functional connectivities of particular interest. The upper row displays the results without a VN gradient, and the lower row presents the results with a VN gradient. Abbreviations: z’ – Fisher’s z; M1 – primary motor cortex; S1 – primary sensory cortex; L-IFC – left inferior frontal cortex; R-IFC – right inferior frontal cortex; vPCC – ventral posterior cingulate cortex; PCu – precuneus; Vis – visual network; SM – sensorimotor network; dAtt – dorsal attention network; vAtt – ventral attention network; Lim – limbic network; FP – frontoparietal network; DMN – default-mode network.

Schematic Overview of Directed Functional Connectivity Calculation Procedure. (a) The surface-based Shen268 functional parcellation. (b) The depth-dependent functional connectivity matrix. (c) The depth-dependent layers were separated into superficial, middle, and deep. The functional connectivity (FC) matrix for each ROI pair was reduced to a simplified 3-by-3 matrix for multiple comparison. Abbreviations: Vis – visual network; SM – sensorimotor network; dAtt – dorsal attention network; vAtt – ventral attention network; Lim – limbic network; FP – frontoparietal network; DMN – default-mode network.