Characterisation of laminar and vascular spatiotemporal dynamics of CBV and BOLD signals using VASO and ME-GRE at 7T in humans

Figure 1:. Functional data acquisition and stimulation. A Left: Sequence diagram of symmetric VASO acquisition. Right: Illustration if k-space trajectories for nulled and BOLD volumes. B Left: Stimulation onsets with respect to acquisition scheme and resulting in sub-second sampling of the stimulus. Right: Illustration of acquired data points of an arbitrary stimulus response per jitter. C Stimulus- and corresponding rest durations. Note that in a few sessions (5/20) of individual participants, shorter ITIs were used. For a full description of the scanning sessions, see Supplementary Table S1. D Coverage of functional slab overlaid on UNI of MP2RAGE (sub-05).

Figure 2:. Quality assessment and activation. A Mean nulled and BOLD images of all runs in the first session (n=6) of a representative participant (sub-06). B BOLD and VASO tSNR across all participants extracted from ROI (see Figure 3A and outline on a single slice in Figure 2C). C VASO (left) and BOLD (right) tSNR maps of one participant (sub-06) averaged across all runs of the first session (n=6). Green outline shows the ROI on the slice. Note that ROIs extend across multiple slices (see Figure 3A). D Position of the slice shown in all sub-panels illustrated on a whole-brain MP2RAGE UNI image. E VASO (upper) and BOLD (lower) z-maps of one participant (sub-06) resulting from GLM (contrast: stimulus duration rest) for all stimulus durations separately, averaged across all sessions. Red arrows indicate likely position of large pial vessels.

Figure 3:. Segmentation and event-related averages across cortical depth. A Left: Cortical gray matter segmentation (red & green) overlaid on a single slice of the MP2RAGE UNI image of one participant (sub-06). The ROI in the calcarine sulcus (V1) is indicated by green color. Middle: Zoomed in section of the segmentation in ROI with layering. Right: 3D-representation of the segmentation. Red translucent shape is general gray matter while green refers to the ROI. Note that here, only 3 layers are shown. For the layer profiles in Figure 4, we estimated 11 layers. B Group level (n=4) VASO (upper) and BOLD (lower) signal changes [%] in response to all stimulus durations separately for superficial, middle and deep layer compartments. Note the subtle differences of the response time, magnitude, and HRF shape across layers, stimulus durations and contrasts. Stimulation period is indicated by gray shaded areas and error bands indicate 95% confidence interval across participants.

Figure 4:. Depth-dependent profiles show clear peaks in middle cortical layers and modulation of microvascular weighting for VASO. A Left: Group level (n=4) VASO layer profiles showing z-scores across cortical depth for all stimulus durations separately. It can be seen that the VASO response decreases less strongly in superficial layer groups for shorter stimulus durations. Error bands indicate 95% confidence intervals across participants. Right: Same data as on the left, but min-max normalized to show differences in profile shape irrespective of activation amplitude. It can be seen that the VASO response has a relatively stronger weighting towards superficial layer groups for short stimuli (more concave). Error bands were omitted for better visibility because of the large overlap. B Same as A but for BOLD data.

Figure 5:. Time-to-peak analysis. VASO (upper) and BOLD (lower) time to peak (TTP) for all stimulus durations (indicated by white dashed line) and layers independently. For short stimuli, the TTP is after stimulus cessation. For stimuli with a duration of 12 seconds, the TTP is around stimulus cessation. For stimuli with a duration of 24 seconds, the TTP is after stimulus cessation. Furthermore, it can be seen that the layer-specific distributions of TTPs change across stimulus durations. For short stimuli, the superficial layers peak later, whereas for longer stimuli the middle layers tend to peak later. This could be explained by different relative weighings of various vascular compartments (see Figure 6). For BOLD, we could not see any trends of layer-dependent TTPs across any stimulus durations. TTP was defined as the highest signal change between the initial deflection above 0 and falling off below 0. Error bars indicate 95% confidence intervals across participants.

Figure 6:. Signal in vessel dominated voxels shows additional signal dynamics. A Acquisition of ME-GRE data at 0.35 mm isotropic resolution. Acquired slab overlaid on the MP2RAGE UNI image (sub-05) B Quantitative T2* map and segmentation of vessels and ROI in a representative participant (sub-06, top) and a 3D representation of the same data (bottom) C Group Level (n=4) VASO (top) and BOLD (bottom) signal changes in percent for gray matter and vessel dominated voxels separately for each stimulus duration. It can be seen that for short stimuli VASO shows slightly larger responses in large vessels compared to gray matter. For longer stimuli, parenchyma responses are stronger than vessel responses. BOLD shows strong vessel responses, markedly exceeding gray matter across all stimuli. Stimulation period is indicated by gray shaded areas and error bands indicate 95% confidence interval across participants.

Figure 7:. Signal in vessel dominated voxels shows additional signal dynamics. VASO (left) and BOLD (right) peak signal change of vessel dominated voxels divided by the peak signal change in gray matter for each stimulus duration independently. For VASO, the ratio depends on the stimulus duration, whereas for BOLD, this influence is negligible. Overall, the vessel bias is smaller for VASO than for BOLD across stimulus durations. It can be seen that for VASO, long stimuli result in responses that are more confined to gray matter.