Ultra-high field MRI: Advancing systems neuroscience towards mesoscopic human brain function

A critical assessment of data quality and venous effects in ultra-high-resolution fMRI

“…26 27 In this paper, we perform ultra-high-resolution fMRI during a simple visual experiment whose general 28 design is representative of the kinds of experiments neuroscientists might conduct. For acquisition, we 29 use a gradient-echo echo planar imaging pulse sequence, motivated by the fact that gradient-echo 30 delivers the high levels of contrast-to-noise ratio that neuroscience experiments require, dominates in 31 neuroscience applications, and is widely available (for a consideration of spin-echo techniques, see 32 Discussion). We use isotropic voxels (0.8-mm) to ensure unbiased sampling of the convoluted cerebral 33 cortex, and we use multiband slice acceleration (Moeller et al, 2010) to achieve large coverage-such 34 coverage is important because sensory, cognitive, and motor function often reflect coordinated activity of 35 a large number of interacting brain regions.…”

“…For comparison of high-resolution results to 27 that obtained using more standard protocols, we conducted the fLoc experiment at 3T using a low-28 resolution fMRI protocol. This involved gradient-echo EPI at 2.4-mm isotropic resolution with partial brain 29 coverage (30 slices, TR 1.1 s, TE 30 ms, flip angle 62°, no partial Fourier, no in-plane acceleration, 30 multiband slice acceleration factor 2), along with gradient-echo fieldmaps. 31 32 2.5.…”

“…The entire surface-based pre-processing ultimately reduces to a single temporal resampling (to deal with 28 slice acquisition times and change of sampling rate) and a single spatial resampling (to deal with EPI 29 distortion, head motion, and registration to anatomy) ( Figure 1A). Minimizing unnecessary interpolations 30 preserves resolution in the functional data and is similar to the approach taken in the Human Connectome 31 Project (Glasser et al, 2013).…”

“…Line profiles 27 28 To generate the 'line profiles' visualization shown in Figure 11, we manually defined a line on cortex and 29 determined the sequence of vertices that most closely approximate the line. Cortical distance was 30 calculated as the Euclidean distance between successive pairs of vertices, where distance is measured 31 with respect to the white surface.…”

“…The T 1 volumes were processed using FreeSurfer to generate cortical surface 25 representations. To ensure that the surfaces have sufficiently high resolution to support the functional 26 measurements and to allow results to be examined as a function of cortical depth, we included two 27 additional steps beyond standard FreeSurfer processing: we increased the density of surface vertices by 28 a factor of two, and we created 6 surfaces spaced equally between 10% and 90% of the distance 29 between the pial surface and the boundary between gray and white matter. An example of the resulting 30 surfaces is shown in Figure 2.…”