Laminar perfusion imaging with zoomed arterial spin labeling at 7 Tesla

Abstract

Laminar fMRI based on BOLD and CBV contrast at ultrahigh magnetic fields has been applied for studying the dynamics of mesoscopic brain networks. However, the quantitative interpretations of BOLD/CBV fMRI results are confounded by different baseline physiology across cortical layers. Here we introduce a novel 3D zoomed pseudo-continuous arterial spin labeling technique at 7T that offers the unique capability for quantitative measurements of laminar cerebral blood flow (CBF) both at rest and during task activation with high spatial specificity and sensitivity. We found arterial transit time in superficial layers is ~100 msec shorter than in middle/deep layers revealing the dynamics of labeled blood flowing from pial arteries to downstream microvasculature. Resting state CBF peaked in the middle layers which is highly consistent with microvascular density measured from human cortex specimens. Finger tapping induced a robust two-peak laminar profile of CBF increases in the superficial (somatosensory and premotor input) and deep (spinal output) layers of M1, while finger brushing task induced a weaker CBF increase in superficial layers (somatosensory input). We further demonstrated that top-down attention induced a predominant CBF increase in deep layers and a smaller CBF increase on top of the lower baseline CBF in superficial layers of V1 (feedback cortical input), while bottom-up stimulus driven activity peaked in the middle layers (feedforward thalamic input). These quantitative laminar profiles of perfusion activity suggest an important role of M1 superficial layers for the computation of finger movements, and that visual attention may amplify deep layer output to the subcortex.

Feasibility of 3T layer-dependent fMRI with GE-BOLD using NORDIC and phase regression

“…However, its weighting towards macrovasculature is well established (Menon, 2012; Turner, 2002; Uludağ et al, 2009; Yacoub et al, 2003) and needs to be dealt with for laminar purposes where spatial specificity is of the essence. The superficial bias that follows from large vein sensitivity was clearly present in the NORDICmagn profiles which indeed possessed the characteristic positive gradient from WM to CSF (Figure 4A and Figure 5B) consistently found with GE-BOLD (Aitken et al, 2020; de Hollander et al, 2021; Han et al, 2021; Huber et al, 2015; Kok et al, 2016; Shao et al, 2021; Stanley et al, 2020) (see also Figure S3). Also, activation maps revealed higher percent signal changes towards the surface and in CSF which adds to the evidence for large vein contamination (Figure 4B).…”

“…Although we do believe that the lack of two distinct peaks is partly explained by macrovascular influence, it may also be explained by the fact that we extract signal from relatively large ROIs spanning several slices (Han et al, 2021; Pais-Roldán et al, 2020). Moreover, not all subjects have double peak profiles even with non-GE-BOLD sequences (Beckett et al, 2020; Shao et al, 2021), thus we cannot rule out that more distinct peaks would emerge in the group profile with different subjects. Third, NORDICmicro activation maps still contain CSF voxels with strong percent signal changes (Figure 4B).…”

“…That being said, several lines of evidence indicate that non-negligible contributions from macrovascular sources were still in play. First, even though the group layer-profile for NORDICmicro tends to flatten towards CSF as compared to the NORDICmagn profile which keeps rising, the peak is located at the GM/CSF border, whereas sequences proven to have higher spatial specificity peaks within the GM of M1 during finger tapping (VASO/VAPER: (Beckett et al, 2020; Chai et al, 2020; Huber et al, 2017; Persichetti et al, 2020), SE/GRASE: (Beckett et al, 2020; Han et al, 2021), ASL: (Shao et al, 2021)). A peak towards CSF after phase regression was also observed for the visual cortex in Stanley et al, (2020), whereas neural activity is expected to peak in middle layers during visual stimulation (Liu et al, 2020).…”

“…ROIs were manually defined in the hand knob area of Brodmann area 4a (BA4a) similarly to previous layer-fMRI reports focusing on M1 and finger tapping (Beckett et al, 2020; Chai et al, 2020; Han et al, 2021; Huber et al, 2017; Persichetti et al, 2020; Shao et al, 2021). First, the hand knob was visually identified from its characteristic omega-shape (Yousry et al, 1997) and voxel selection was then limited to the upper and lateral part of the knob corresponding to BA4a (Huber et al, 2017).…”

“…This selection was restricted by the following criteria: 1) ROIs were defined as columns (no holes) going roughly perpendicular onto the surface to make it somewhat anatomical meaningful; 2) ROIs had to consist of minimum 50 voxels to increase the ability to compare the same cortical portion across sessions despite e.g. imaging slab positioning not being fully identical, to minimize influence of inaccuracies in e.g., segmentation or across session alignment, and to increase representability of measured M1 activation; 3) ROIs should not extend into more than four slices; 4) the peak signal in depths 0.3-0.4 should not exceed that of depths 0.6-0.8 in order to identify the same functional representation as previous M1 layer-fMRI studies, which generally found stronger activity in superficial compared to deep layers (Beckett et al, 2020; Chai et al, 2020; Han et al, 2021; Huber et al, 2017; Shao et al, 2021); 5) Empty axial slices were not allowed in the final union ROI (see below), that is, it had to be consecutive. These initial ROIs defined in native space of each individual session were then transformed into the space of each subject’s first session where the union of all ROIs was computed and transformed back to native space (Figure 1).…”