Contents lists available at ScienceDirect

NeuroImage

journal homepage: www.elsevier.com/locate/neuroimage

High resolution data analysis strategies for mesoscale human functionalMRI at 7 and 9.4 T

Valentin G. Kempera, Federico De Martinoa,b, Thomas C. Emmerlinga, Essa Yacoubb,Rainer Goebela,c,⁎

a Department of Cognitive Neuroscience, Faculty of Psychology and Neuroscience, Maastricht University, Oxfordlaan 55, 6229 EV Maastricht, TheNetherlandsb Center for Magnetic Resonance Research, University of Minnesota Medical School, 2021 Sixth Street SE, Minneapolis, MN 55455, United Statesc Department of Neuroimaging and Neuromodeling, Netherlands Institute for Neuroscience, Royal Netherlands Academy of Arts and Sciences (KNAW), 1105BA Amsterdam, The Netherlands

A R T I C L E I N F O

Keywords:High resolutionSubmillimeter functional magnetic resonanceimaging7 Tesla9.4 TeslaCortical thicknessEquidistantEqui-volumeCortical depth samplingMyelinOcular dominanceVisual cortex

A B S T R A C T

The advent of ultra-high field functional magnetic resonance imaging (fMRI) has greatly facilitatedsubmillimeter resolution acquisitions (voxel volume below (1 mm³)), allowing the investigation of corticalcolumns and cortical depth dependent (i.e. laminar) structures in the human brain. Advanced data analysistechniques are essential to exploit the information in high resolution functional measures. In this article, we userecent, exemplary 9.4 T human functional and anatomical data to review the advantages and disadvantages of(1) pooling high resolution data across regions of interest for cortical depth profile analysis, (2) pooling acrosscortical depths for mapping patches of cortex while discarding depth-dependent (i.e. columnar) effects, and (3)isotropic sampling without pooling to assess individual voxel’s responses. A set of cortical depth meshes may bea solution to sampling information tangentially while keeping correspondence across depths. For quantitativeanalysis of the spatial organization in fine-grained structures, a cortical grid approach is advantageous. Wefurther extend this general framework by combining it with a previously introduced cortical layer volume-preserving (equi-volume) approach. This framework can readily accommodate the research questions whichallow for spatial smoothing within or across layers. We demonstrate and discuss that equi-volume samplingyields a slight advantage over equidistant sampling given the current limitations of fMRI voxel size, participantmotion, coregistration and segmentation. Our 9.4 T human anatomical and functional data indicate theadvantage over lower fields including 7 T and demonstrate the practical applicability of T2

* and T2-weightedfMRI acquisitions.

Introduction

Functional magnetic resonance imaging (fMRI) has rapidly devel-oped since its introduction and has become a key neuroscientificresearch tool. More and more detailed aspects of the functionalarchitecture of the human brain are unraveled by means of steadilyincreased data quality (signal-to-noise ratio (SNR), spatial and tem-poral resolution). As the spatial resolution is pushed from tens of cubicmillimeters to fractions of one cubic millimeter (submillimeter), morefine-grained structures such as cortical laminae and columns becomeaccessible in-vivo in humans, closing the gap between invasive (animal)electrophysiological or optical research methods and human fMRI.While high and ultra-high field strengths have played an important rolein making higher spatial resolutions accessible, advanced data analysis

strategies paralleled this development to exploit the new physicalpossibilities.

Using exemplary data, we review in this article how ultra-high fieldfMRI allows investigating the mesoscopic functional organization ofthe cortex. In this context it is useful to differentiate applications bytheir purposes leading to different requirements for high-resolutionmeasurements and analysis strategies. A major distinction can be madebetween applications that allow pooling of measurements along at leastone cortical dimension from those applications that consider eachvoxel’s response individually. Submillimeter resolution fMRI has been,for example, used to separate (coarsely) cortical laminae reportingdifferential pooled functional (or anatomical) responses along largerstretches of cortex (Koopmans et al., 2010; Koopmans et al., 2011;Ress et al., 2007). Other studies have separated neighboring responses

http://dx.doi.org/10.1016/j.neuroimage.2017.03.058Accepted 28 March 2017

⁎ Corresponding author.E-mail address: r.goebel@maastrichtuniversity.nl (R. Goebel).

NeuroImage 164 (2018) 48–58

Available online 14 April 20171053-8119/ © 2017 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).

MARK

along the cortex while integrating signals across cortical depth, e.g. tomap columnar feature representations with non-isotropic voxels (e.g.Cheng et al., 2001; Goodyear and Menon, 2001; Yacoub et al., 2008,2007). In recent years, studies have started to separate functionalresponses along both cortical distance as well as cortical depth. Thesestudies are particularly challenging, as they do not allow pooling alongcortex either tangentially or radially and thus require high isotropicspatial resolution with sufficient contrast-to-noise ratio (CNR) at eachsubmillimeter voxel. While challenging, this approach is suited forresearch questions that aim to map feature representations along thecortex while studying the consistency (“columnarity”) of these mapsacross cortical depth levels simultaneously. In order to properly relatefunctional responses across depth levels, correspondence betweenlocations in different layers needs to be established. This requiresappropriate cortex modeling and analysis strategies that not onlyseparate layers but also link positions across depths. Initial results ofthis approach include axis-of-motion maps in area hMT or soundfrequency maps in area A1 for different cortical layers (De Martinoet al., 2015; Zimmermann et al., 2011).

In the sections below, we review the above-mentioned applicationscenarios and analysis strategies in more detail by the means of recenthigh resolution 9.4 T data, and illustrate the benefits of ultra-high fieldimaging (beyond 7 T). Furthermore, we describe in greater detail theregular grid depth sampling approach that allows us to accuratelyanalyze, process, and visualize responses along layers and columnswith high spatial resolution in restricted cortical regions. Besidesvarying cortical thickness and curvature supporting equidistant (i.e.cortical depth preserving) sampling, the analysis method now incorpo-rates an equi-volume (i.e. cortical layer volume preserving, Waehnertet al., 2014) layer separation model.

Standard cortical meshes

A relative strength of fMRI compared to optical imaging methods isthe accessibility of areas in cortical sulci. For analysis and visualizationpurposes, folded, inflated or flattened surface representations of thevolumetrically sampled data have proven useful (Dale et al., 1999;Goebel et al., 1998). These triangulated surface representations aretypically reconstructed along the white matter/gray matter (or mid-gray matter) boundary of a whole hemisphere from a segmentedanatomical whole brain acquisition, e.g. T1-weighted MPRAGE(Mugler and Brookeman, 1990) or more recent extensions of it (e.g.MP2RAGE, MEMPRAGE, Marques et al., 2010; van der Kouwe et al.,2008). The triangulation approach is the natural choice because iteasily allows smoothing (removing its original voxelated appearance)and flattening, allowing visualization of sampled functional data aswhole-hemisphere surface maps with minimal distortions. Spatiallytransforming a reconstructed hemisphere mesh into a sphere facilitatescortex based alignment in multi-subject analysis, usually providingbetter group-level results than volumetric template-based alignmentapproaches (e.g. in MNI or Talairach normalization, Frost and Goebel,2012).

For separating cortical layers with high resolution fMRI, the surfaceapproach has been extended by adding multiple cortical meshes atdifferent cortical depths (Fig. 1a). These can be generated by evolving awhite matter/gray matter (WM/GM) boundary surface towards theboundary with cerebrospinal fluid (CSF) along vertex normals, inconjunction with a cortical thickness estimate, which requires segmen-tation of both the WM/GM boundary as well as the GM/CSF boundary(Dale et al., 1999; Fischl and Dale, 2000; Polimeni et al., 2010). Thisprocedure automatically grants correspondence between the samevertices (mesh nodes) at different cortical depths, which would be lostif independent surfaces were created for different cortical depths.Constraints can be applied to enforce smooth curvature (second-ordersmoothness) and uniform vertex spacing, and to prevent self-intersec-tion (Fischl and Dale, 2000).

Construction of high resolution cortical depth grids

The surface mesh approach based on triangulation described aboveis very suitable for whole hemisphere representations because it worksrobustly in the presence of highly variable curvatures. However,distances and angles between neighboring nodes vary strongly andthereby impede the analysis of (quantitative) spatial relations. Thisissue results from reconstructing the voxelated boundary of segmentedanatomical data followed by mesh smoothing to obtain a model of thecortical sheet (Fig. 1a). A further issue of whole-hemisphere meshes aredistortions in local distances and angles introduced when smoothingmeshes or, more pronounced, when flattening meshes for visualization.

When only a small region of cortex is of interest, a Cartesian gridapproach (introduced in Zimmermann et al. (2011)) promises morepracticability because it creates smooth regular cortex grids directly ingrey matter at a specified spatial resolution instead of starting withmeshes reconstructed at white matter voxel boundaries (Fig. 1b). Suchregular 2D grids better reflect local distances and angles whenquantitatively analyzing and visualizing sampled (e.g. functional) datasince no mesh resampling is necessary. Note that in the standard meshapproach, the number of neighbors of a vertex depends on theoriginally reconstructed voxel boundaries (e.g. a vertex has lessneighbors if located in a straight wall than if located at the fundus ofa sulcus). After smoothing the mesh, such differences lead to differentangles between edges formed by the triangles, which connect a vertexwith its neighbors. These irregularities result in inhomogeneousdensity sampling of the targeted depth surfaces. Since regular gridsare created explicitly following iso-surfaces, each vertex is alwayssurrounded by 4 neighboring vertices at a constant distance forming90 degree angles between edges (distances stretch or shrink as in themesh approach when projecting mid-GM surfaces to other depth levelsin curved cortex). The explicitly created regular grids simplify accuratepost-processing of sampled functional data such as shortest distance,gradient and curvature calculation. Spatial mesh smoothing withinlayers, for example, can be implemented using standard Gaussiankernels for 2D Cartesian coordinates. While two-dimensional opera-tions such as within-layer (Gaussian) smoothing of sampled functionaldata is also possible with standard meshes (Ahveninen et al., 2016;Andrade et al., 2001; Jo et al., 2007; Kiebel et al., 2000), they are morestraightforward and precise when using regularly designed grids sincethey do not require resampling. How valuable these expected advan-tages are in specific cases needs to be shown by direct comparisons ofthe whole-mesh and cortical grid approach in future work.

The high resolution cortical grids that the authors have used inrecent studies (De Martino et al., 2015, 2013; Kemper et al., 2015b;Zimmermann et al., 2011) are based on the following procedure: Afterbrain segmentation of WM, GM, and CSF, the mid-GM surface is foundvia an assessment of cortical thickness using the Laplace method(Jones et al., 2000). The smooth intensity field within GM obtained bythe Laplace method allows calculation of a gradient of intensitiespointing in the direction of cortical depth at each GM voxel (Fig. 1b).Integrating along these gradient values produces “field lines” or“streamlines” (Jones et al., 2000) establishing one-to-one radialcorrespondence between the WM/GM and GM/CSF surface for sub-sequent modelling. A regularly spaced Cartesian grid is created arounda seed point in the mid-GM surface traversing in orthogonal directionwith respect to the streamlines, i.e. along the tangent plane.Importantly, the mid-GM surface is not directly taken from the middleequipotential isocontour, but generated at middle cortical thickness toyield an equidistant cortical depth grid. Within this surface, regularmetric distances can be guaranteed (Fig. 1b, blue grid). Then, grids atany deeper and more superficial relative depths are added by evolvingthe mid-GM regular grid towards the WM/GM or GM/CSF boundaryfollowing streamlines (Fig. 1b, green grid). Because of the uniquecorrespondence between grid points at different cortical depths, thismethod is particularly suited to study possible columnar organizations

V.G. Kemper et al. NeuroImage 164 (2018) 48–58

49

of functional responses. To establish correspondence across grid pointsin a piece of curved cortex, equal sampling distances of sampling pointsat cortical depths other than the starting one must be, however,violated (indicated by “stretched” green grid in Fig. 1b). Note thatthe grid resolution oversamples the voxel resolution at least twice inorder to ensure sufficiently dense voxel sampling in upper and lowergrids despite expansions and shrinkages of distances between corre-sponding points at different depth levels in folded cortex.

Cortical depth grids provide advantages for visualization of func-tional depth-sampled data. First, grids can easily be visualized in 2Dformat without distortions introduced by flattening standard meshes.Second, a stack of depth grids with associated sampled functional data

can also be converted in volume space with two dimensions represent-ing the extent of cortex and one dimension representing cortical depth.These grid volumes allow using standard volume tools for analysis andvisualization including volume rendering and slicing, which is helpfulfor inspecting consistency of responses across depth, i.e. columnarity.

An application of this modeling approach is demonstrated in Fig. 2.The results of an ocular dominance experiment at 9.4 T are overlaid ona high resolution cortical depth grid. The direct correspondence ofsampling points at different cortical depths allows for inspection andstraight-forward quantitative analysis (De Martino et al., 2015) of (dis-)similarity of results at different cortical depths, while the columnarpattern tangentially to the cortex can also be observed.

Fig. 1. High-resolution cortical depth analyses using Cartesian grids. (a) The standard mesh approach first reconstructs a triangulated mesh at the voxel boundaries of a WM/GM (ormid-grey matter) boundary of a segmented volume. In order to better fit the underlying smooth cortical sheet, the voxelated mesh is smoothed leading to unequal distances and anglesbetween vertices. In order to keep correspondence across depth, additional meshes at any relative depth level are created following vertex normals. (b) In the Cartesian grid approach,smooth meshes are directly created within grey matter based on a smooth vector field obtained by solving the Laplace equation (Jones et al., 2000). In order to keep correspondenceacross depth, additional grids at any relative depth level are created by following streamlines. (c) Besides placing grids at equi-distant relative depth levels, depth can be adjustedfollowing the equi-volume model. Using small segments of connected grid points forming depth frustums, grid points can be adjusted in depth in order to keep layer volumes constant infolded cortex; the schematic explanation uses 3 grids and, correspondingly, 2 layers. (d) Results from equidistant and equi-volume sampling using 4 grids (3 layers) in a cut of a three-dimensional section of a segmented brain; left panel shows partial view of created grids while right panel shows filled layers in corresponding slice.

V.G. Kemper et al. NeuroImage 164 (2018) 48–58

50

Equi-volume model

In our previous publications (e.g. De Martino et al., 2015;Zimmermann et al., 2011) grids were created at cortical depth levelsfollowing an equidistant approach keeping a specified relative corticaldepth level to the cortex boundaries, i.e. the absolute distance at a localregion for a specified relative depth level is obtained by multiplying thedepth level by the measured local cortical thickness. Here we describean extension of regular depth grid creation integrating the equi-volumeapproach (Waehnert et al., 2014), which has also been applied in fMRI(Huber et al., 2015; Kok et al., 2016). The equi-volume model adjuststhe thickness of layers in cortical segments to preserve their volumecompensating for cortical folding. Using the geometric properties of thegrid sampling approach, this principle can be accommodated simply bycalculating respective (sub-)volumes in a depth volume frustum (por-tions of solids that lie between two parallel planes, see Fig. 1c) sincevarying curvature is explicitly reflected in connections of corresponding

points across depth by expanding/shrinking distances to neighboringgrid points across layers. In flat cortex, distances to neighboring gridpoints will be constant and calculated volume fractions correspond tocorresponding relative distances, i.e. in flat cortex the results of theequidistant and equi-volume model are identical. In folded cortex,however, the height of layers (i.e. distance of intermediate grids) needto be adjusted in order to keep constant relative volumes which arecalculated using the volume equation for frustums. More specifically, acone frustum is calculated at each grid position by connecting thecorresponding points at the WM/CSF and GM/CSF boundary forminga vertical segment vector s and by calculating the area of a disk(indicated in Fig. 1c) at the WM and GM boundary (AWM, AGM)oriented orthogonally to the vertical segment vector. A cone frustum(instead of a frustum with a rectangular base) is used, in order toensure that the frustum volume equation is also applicable in caseswith substantially different curvatures in different directions (e.g.saddle points). The volume of the created cone frustum can be

Fig. 2. : Results of human 9.4 T visual fMRI experiment targeting ocular dominance. Selectivity index s represents degree of exclusiveness of response to left eye (negative values) orright eye (positive values) in monocular visual stimulation, defined as s = (βR – βL)/( βR + βL), where βR and βL represent BOLD signal response amplitudes to left and right eyestimulation, respectively. (a) Ocular dominance responses in early visual cortex (V1 and V2 boundary delineated based on pRF mapping experiment) on inflated cortex representation.View from posterior onto the occipital pole. (b) High resolution cortical grid volume representations of regions of interest in V1 (equidistant sampling along cortical depth). The gaps inthe rendered representations allow inspection of the consistency of ocular dominance across cortical depth. Results from T2-weighted BOLD fMRI using 3D-GRASE (0.7 mm isotropicnominal resolution) overlaid on anatomical reference generated from T1-weighted MPRAGE (see methods section for details).

V.G. Kemper et al. NeuroImage 164 (2018) 48–58

51

Fig. 3. : (a) Cortical depth profiles of high resolution T2*-weighted anatomical data sampled in V1 (green) and V2 (red) according to equi-volume (solid lines) and equidistant (dashed

lines) cortical models. Plot represents region averages within one hemisphere, each. Shaded areas represent std. errors. The gray band represents the approximate position of the Stria ofGennari as assessed from Nieuwenhuys (2013). (b) High resolution T1-weighted reference images (left column, 0.6 mm isotropic interpolated to 0.35 mm isotropic), and T2

*-weightedimages (right column, 0.35 mm isotropic native resolution), superimposed with GM segmentation (middle row) and V1/V2 regions of interest (bottom row). Note the clear appearance ofthe Stria of Gennari in the T2

*-weighted image (black arrows) and dark lines just underneath the gray matter in white matter (white arrows). See also supplementary materials S1 and S2for multi-slice T2

*-weighted images in two participants.

V.G. Kemper et al. NeuroImage 164 (2018) 48–58

52

calculated based on the height h of the vertical segment and the diskareas at the WM and GM boundary. In order to calculate the sub-volume of any layer, the calculated disk radii at the WM and GMboundaries are interpolated linearly to any intermediate depth level;the interpolated radii are used to calculate adjusted disk areas fromwhich the volume of any sub-frustum (layer) can be obtained (Fig. 1c).A comparison of equidistant and equi-volume grid sampling is shownin Fig. 1d.

Fig. 3(a) shows an application of cortical depth sampling usinghuman anatomical data acquired at 9.4 T. High resolution (0.35 mmisotropic) T2

*-weighted images were acquired and co-registered with aT1-weighted reference image, which was used for WM/GM/CSFsegmentation. Cortical depth profiles were assessed within retinotopi-cally defined primary (V1) and secondary (V2) visual areas (see below)according to the equidistant and equi-volume models. The highlymyelinated stria of Gennari, which is known to have short T2

* andhence low intensity in T2

*-weighted images (Bridge et al., 2005; Buddeet al., 2011; Duyn et al., 2007; Fukunaga et al., 2010; Sánchez‐Panchuelo et al., 2012) is expected to generate a dip in the depthdependent profiles of area V1 (but not V2) at middle cortical depths.Even at this very high resolution the results between the two methodsare very similar. Nonetheless, closer inspection unveils that equidistantsampling yields slightly smoother profiles (most noticeable in partici-pant 2, left hemisphere), as a consequence of sampling in convex andconcave areas. When sampling concave and convex areas separately(i.e. sulci and gyri, respectively), systematic differences between thetwo sampling methods emerge (see Fig. 4). In line with the ex-vivofindings of Waehnert et al. (2014), the trough is located at the samemiddle cortical depth in concave and convex areas of V1 with equi-volume sampling, while it is shifted using equidistant sampling.

Spatially pooled sampling alternatives

When considering the multiple alternatives of depth sampling, thechosen analysis strategy has to serve the specific research question. Forexample, obtaining regular grids over an entire hemisphere is notpossible. In such cases, the triangulation approach seems currentlymost suitable. For this reason, triangulated surfaces at different corticaldepths have been used in anatomical (e.g. De Martino et al., 2014b)and functional (mesoscopic) studies (Goncalves et al., 2015). The lattersampled only middle or deep and middle cortical depths and ignoredsuperficial layers because of the well-known detrimental effect ofmacrovasculature in these layers (Polimeni et al., 2010, see also in

ultra-high field imaging section below).To study feedforward-feedback relations, the cortical depth profile

of a region of interest may be sufficient. This requires an accuratedefinition of the brain area under investigation and assumes that in ahomogeneously tuned cortical area with similar receptive fields (oranatomical properties, De Martino et al., 2014b) in order to averagesignals (Huber et al., 2015; Kok et al., 2016; Scheeringa et al., 2016) orinformation (Muckli et al., 2015) across many voxels at the samecortical depth. Evidently, this approach significantly reduces thedemand for high SNR because of the averaging across voxels whileallowing much finer cortical depth resolution than the voxel size.

Limitations

A common limitation of the various cortical depth samplingapproaches described above is that they rely on the segmentation ofthe cortex derived from an anatomical reference image, most com-monly a T1-weighted image at submillimeter resolution. Two errorsources are therefore common to various cortical sampling algorithms:(1) Segmentation errors of the WM/GM boundary and the GM/CSFboundary lead to an erroneous definition of the cortical depth.Additional anatomical contrasts (e.g. additional T2

*-weighted acquisi-tion) may help to reduce this ambiguity (De Martino et al., 2014a). (2)Motion and distortion artefacts cause misregistration to the anatomicalreference across extended areas. Advanced coregistration techniquessuch as boundary-based registration may improve the results (Greveand Fischl, 2009). Independently of the cortical sampling technique,these two aspects will always limit the correctness of assigned corticaldepth and surface position. To avoid some of the issues present whenaligning to a standard anatomical data set, a reference image with thesame distortions as the functional data can be used (Renvall et al.,2016). This can be achieved either by acquiring additional images withstrong anatomical contrast (e.g. T1-weighed EPI references), or bysegmenting the functional data directly (e.g. in combination withcomplex phase information, Fracasso et al., 2016).

Another limitation common to all sampling methods is the intrin-sically limited effective spatial resolution of the functional data, whichis below the nominal resolution. Not only do interpolation andmisregistration impede the effective spatial resolution, but so do imageblurring and functional (BOLD) signal point-spread, which will bediscussed in the following section.

Our data (see Fig. 3) support the previously expressed notion(Fracasso et al., 2016; Guidi et al., 2016; Waehnert et al., 2014) thatalthough equi-volume depth sampling according to the Bok model(Bok, 1929; Waehnert et al., 2014) yields a small advantage, thedifferences with equidistant sampling are limited in the context ofcurrently feasible in-vivo functional experiments. This is also true forthe ocular dominance data (Fig. 2): If resampled according to the equi-volume model, only 0.2% of the labeled grid points switch oculardominance from left to right or vice versa, and 90.7 % of the labels(right/left/undefined) in the sampled grid stay identical. These resultssuggest that previous fMRI studies using equidistant sampling wouldyield only slightly different results when using equi-volume sampling.

Ultra-high field for high SNR at high isotropic resolution

As discussed above, high SNR is a prerequisite for approaches thatconsider each voxel’s response individually, and ultra-high fieldscanners can provide this. As in structural imaging, the baseline signalhas better SNR because of the improved physical spin statistics andmore favorable interaction between the RF field and the biologicaltissue (Ugurbil et al., 2003). In addition, an approximately linearincrease in dynamic blood oxygenation level dependent effect (BOLDeffect) size is theoretically expected (Uludag et al., 2009) and experi-mentally observed with stronger magnetic fields (Yacoub et al., 2003,2001). Both effects, better image SNR and larger BOLD effect size,

Fig. 4. : Cortical depth profiles of high resolution T2*-weighted anatomical data sampled

in V1 separately for concave areas (light green, sulci), convex areas (dark green, gyri),and combined (medium green), averaged across all hemispheres. Equi-volume (solidlines) and equidistant (dashed lines) sampling are drawn separately. A dotted line atmiddle cortical depth (50%) is drawn for visual guidance.

V.G. Kemper et al. NeuroImage 164 (2018) 48–58

53

Fig. 5. (a–d) Population receptive field mapping results from 9.4 T human fMRI experiments displayed on inflated cortical mesh representations. View from posterior onto the occipitalpole. Eccentricity (a), visual angle (b), pRF sizes (c), and resulting delineation of V1 and V2 (d) are shown for 4 hemispheres of two participants. e) Comparison of voxelwise correlationvalues between 9.4 T and similarly acquired 7 T data (red dashed line and blue solid line, respectively). Shaded areas indicate standard deviation across hemispheres (12 measurementsat 9.4 T, 22 hemispheres at 7 T, see Methods section for details).

V.G. Kemper et al. NeuroImage 164 (2018) 48–58

54

independently improve CNR and counterbalance the reduction of SNRdue to higher spatial resolution.

The aforementioned improvements in CNR are expected to con-tinue also beyond the field strength of 7 T, which is currently thehighest field strength accessible for most human fMRI. Improvementsin SNR have also been demonstrated experimentally (Budde et al.,2014; Pohmann et al., 2015). Fig. 5 showcases the CNR benefit inresults of a routine human 9.4 T experiment. We acquired populationreceptive field (pRF, Dumoulin and Wandell, 2008) data to map theretinotopic organization in human visual cortex (Goebel, 2015). Thefigure shows example pRF mapping results from 9.4 T and histogramsof the voxelwise correlation values, R, in comparison to data from aprevious 7 T experiment with similar imaging properties (see methodssection for details). R values close to 1 reflect good agreement betweenvoxel’s time courses and the modelled pRF responses taking intoaccount its horizontal and vertical position and (Gaussian-shaped)receptive field in the visual field. Median R in the 9.4 T participants issignificantly higher than at 7 T in V1 and V2 (0.68 ± 0.07 vs 0.52 ±0.10 at p < 0.02 (two-sided Wilcoxon rank-sum test considering all 7 Themispheres and median results for each 9.4 T hemisphere)). Thisdemonstrates the feasibility of routine experiments employing rela-tively high spatial resolution (1.05 mm isotropic) using GE-EPI at9.4 T.

While high isotropic resolution sampling of fMRI data yieldsadditional advantages over anisotropic sampling (e.g. pooling acrossdepth or pooling across a cortical area), the underlying fMRI data hasto support the spatial specificity to make this evident. To this end,gradient-echo (T2

*-weighted) BOLD fMRI data as typically acquired,comes short compared to more specific acquisition techniques such asT2-weighted BOLD imaging, cerebral blood volume (CBV) weighted orcerebral blood flow (CBF) based methods (Duong et al., 2001; Duonget al., 2003; Duong et al., 2002; Goense et al., 2016; Goense andLogothetis, 2006; Yacoub et al., 2007). A widespread use of these lattermethods is hindered by the significantly smaller sensitivity comparedto T2

*-weighted BOLD imaging. Nevertheless, our early 9.4 T results(Fig. 2) show that T2-weighted BOLD fMRI can be used to unveil fine-grained patches of differential ocular dominance. Differently fromearlier studies targeting ocular dominance columns, which usedanisotropic resolution with thick slices spanning the entire corticalthickness (Cheng et al., 2001; Goodyear and Menon, 2001; Yacoubet al., 2007), the consistency across cortical depth can be probeddirectly due to the high isotropic spatial resolution (0.7 mm isotropic).Moreover, recent modeling results suggest that in T2-weighted BOLDfMRI, the laminar location of measured signal is linked more closely tothe actual lamina, whereas T2

*-weighted BOLD signals drift towardssuperficial laminae as they go downstream in the draining veins(Markuerkiaga et al., 2016). This insight extends the previouslydemonstrated, well established finding that T2

*-weighted BOLD signalsincrease towards the surface (e.g. De Martino et al., 2013; Goense andLogothetis, 2006; Kok et al., 2016; Polimeni et al., 2010; Zhao et al.,2004), in that it identifies the increased superficial T2

*-weighted signalas the additive signal stemming from deep and superficial layers.Nonetheless, (purely) T2-weighted signal contains also unspecificmacrovascular contributions at ultra-high field (Uludag et al., 2009),and results with regard to the laminar origin of signals have to beinterpreted carefully.

Besides relatively low sensitivity, T2-weighted imaging has otherlimitations: (1) Pure T2-weighting without T2

* contamination isdifficult to achieve in echo-train imaging pulse sequences such as EPI(whereas non-EPI approaches are limited by their temporal ineffi-ciency) and (2) the RF energy deposition requires compromisesbetween spatial resolution, imaging volume, SNR, and temporalresolution. We address these challenges at 9.4 T with the applicationof the 3D-GRASE imaging pulse sequence and the use of a local, RFtransmit efficient surface array in combination with a high SNR receivearray. 3D-GRASE utilizes a slab selective excitation pulse and ortho-

gonal refocusing pulses (Feinberg et al., 2008; Feinberg and Oshio,1991). This creates an inner volume selection, reducing the necessaryphase-encoding steps, while also obviating the need for multipleexcitation pulses as in 2D multi-slice imaging. The local RF transmitcoil reduces RF power deposition in unrelated portions of the head (seesupplementary Figure S3). Parallel transmit techniques promise tofurther increase homogeneity in the imaging FoV (De Martino et al.,2012; Poser et al., 2014 and references therein; Tse et al., 2016; Wuet al., 2013).

Summary

The relatively new field of high field high resolution fMRI allowsinvestigation of cortical columns and cortical laminar structures inhumans. Considerations with regard to analysis strategies for highresolution functional data continue to play an important role in theliterature. In this article, we propose that, depending on the neuros-cientific application, high resolution data may be pooled across acortical patch to study cortical depth dependent effects (i.e. laminarprofiles). If the intention is to map the tangential layout of a corticalregion without investigation of depth-dependent (i.e. columnar) in-formation, data may be pooled across cortical depth (while potentiallyignoring signals in superficial laminae). If individual voxel’s responsesare of interest and may not be pooled across any dimension, a set ofcortical depth meshes may be a solution for sampling informationtangentially while keeping correspondence across depth. For quantita-tive analysis of the spatial organization in fine-grained structures, acortical grid approach is advantageous with respect to calculations andvisualizations. Combined with a cortical layer volume-preserving (equi-volume) model, the cortical grid approach can also be applied toresearch questions that allow for spatial smoothing within or acrosslayers. We demonstrated and discussed that the difference betweenequidistant and equi-volume sampling is minor given the currentlimitations of participant motion, coregistration, segmentation, andfMRI voxel size, however, equi-volume sampling yields an advantagewhen investigating subtle differences in cortical depth profiles. Our 9.4Tesla human anatomical and functional data indicate the advantageover lower fields including 7 T and demonstrate the practical applic-ability of T2

* and T2-weighted fMRI acquisitions.

Materials and methods

Experimental setup and participants

All experiments were performed at the research facilities ofMaastricht University and were in accordance with the local ethicalboard. After giving informed consent, two healthy female participants(mean age 25 ± 1) were scanned in two sessions on a 9.4 T MRIscanner (Siemens, Erlangen, Germany) equipped with a head-onlygradient coil (max. 80 mT/m at 400 mT/m/s slew rate) and an 8-channel transmit 24-channel receive surface coil (Life Services,Minneapolis, MN, USA) covering the posterior part of the participantshead. A static, non-subject-specific B1

+ phase shim was used through-out the measurements, after it had been validated that this approachgenerated sufficiently efficient and homogeneous RF fields in earlyvisual areas in various participants (Kemper et al., 2015a) usingDREAM B1 mapping (Nehrke and Börnert, 2012; Tse et al., 2014).Only the global scaling factor (reference voltage) was adjusted to theindividual participants after acquiring a pre-saturation approach basedflip angle map. SAR supervision was maintained by imposing themaximum local SAR level observed in simulations as global RF powerdeposition limits.

Second order shimming was applied to homogenize the B0 magneticfield in a manually drawn region of interest in early visual areas using acustom-made dual-echo field mapping GRE sequence and MATLAB(The MATHWORKS Inc., Natick, MA, USA) routines (Tse et al., 2016).

V.G. Kemper et al. NeuroImage 164 (2018) 48–58

55

Visual stimulation

Visual stimulation was provided using a projector and a frostedscreen near the participant’s head, seen via a tilted mirror. PRFmapping was performed using randomly moving bars (Senden et al.,2014) in the freely available stimulation software BrainStim (https://github.com/svengijsen/BrainStim). Horizontal, vertical, and obliquefull contrast flickering checkerboard bars (flickering rate 7.5 Hz) werepresented at 12 different positions spanning 0.63° visual angle eachwith maximum eccentricity of 7.5° (square-shaped). During anatomicalacquisitions, participants watched animated cartoon movies.

For binocular eye stimulation using PsychoPy (version 1.82, Peirce,2007), participants wore anaglyph glasses (Lee filters, Andover,Hampshire, UK; red #026, cyan #116) throughout the entire secondexperimental session. The stimulus colors were adjusted such that theeyes could be stimulated individually using the pass-band opticalwavelength of the filter plates. Participants were instructed to fixate acentral fixation dot, while full field concentric flickering checkerboardpatterns were presented. The fixation dot was moved 2.5° upwards,such that more of the lower visual field was stimulated. A block designswitching between right and left eye every 48 s was chosen (Yacoubet al., 2007). In the beginning and end 24 seconds of gray screen werepresented at the same luminance as the non-stimulated eyes perceived.Seven runs of 7:12 min duration each were performed.

Imaging pulse sequences

Anatomical reference images were acquired with a T1-weightedMPRAGE sequence (TE/TI/TR = 2.5/1200/3600 ms; nominal flipangle = 4°; matrix size = 384×384×256; GRAPPA acceleration factor3, partial Fourier factor 6/8 (partition direction); total duration8:54 min.). In addition, a proton density weighted MPRAGE withoutthe inversion module (identical imaging parameters except TR =1620 ms and total duration = 4:01 min.) was acquired to correct fortransmission and receive coil biases (Van de Moortele et al., 2009). TIand flip angle were adapted from 7 T protocols based on literaturevalues for T1 and proton density in WM and GM at 7 and 9.4 T and theequations in Deichmann et al. (2000) and Marques et al. (2010).

PRF mapping was performed using a gradient-echo EPI sequence(1.05 mm isotropic nominal resolution; TE/TR = 14.6/2000 ms; in-plane FoV 105×105 mm; 54 coronal slices; nominal flip angle 70°; echospacing 0.64 ms; GRAPPA 2, partial Fourier 7/8; 13:20 min. duration,3 repetitions).

Anatomical high resolution T2*-weighted images were acquired

using a 2D GRE sequence with limited FoV (0.35 mm isotropicnominal resolution (0.043 mm³); TE/TR = 20/1500 ms; in-planeFoV 100×100 mm; phase-encoding direction left-right; 48 slices;nominal flip angle; 70°; bandwidth 80 Hz/Pixel; no parallel imaging,no partial Fourier; 7:09 min. duration, 6 averages). The narrow FoV inphase-encoding direction causes aliasing of anterior lateral portions ofthe images, which are not of interested and discarded in the analysis(see below).

Ocular dominance imaging was performed using a high-resolutioninner-volume 3D-GRASE sequence (TE/TR =30.85/2000 ms; in-planeFoV 22.4×105 mm; 12 slices; nominal flip angles (90-107-60-56-53-54-54-58-70-107)°; echo spacing 0.75 ms; partial Fourier 6/8 (parti-tion direction); 8:00 min. duration, 7 repetitions). Images wereacquired at a nominal isotropic resolution of 0.7 mm, and were theninterpolated to 0.35 mm isotropic resolution using the complex,uncombined coil data. This was done in order to minimize resolutionlosses due to coregistration or motion-correction. Refocusing flipangles were adjusted based on extended phase graph theory simula-tions to yield reduced across-slice blurring of estimated 2.3 voxels(Kemper et al., 2016).

Data analysis

Functional data analysis was performed using BrainVoyager QX2.8.4 (Brain Innovation, Maastricht, the Netherlands) and MATLABroutines. Preprocessing of the pRF data included slice-time correction,3D-motion correction, and high-pass filtering. Additionally, data wasdistortion-corrected using acquisitions with reversed phase encodingdirection and the COPE plug-in (v0.5) in BrainVoyager (Anderssonet al., 2003; Fritz et al., 2014). The three runs were averaged andanalyzed to generate pRF estimates for each voxel. Model time coursesfor pRFs of receptive field positions between ± 7.5° eccentricity inhorizontal and vertical position and size (Gaussian full width at halfmaximum) between 0.2° and 10° were calculated, to which theexperimental data was compared to find the best match (highestcorrelation). Functional data were co-registered to the intra-sessionreference anatomical data using positional and edge information.Anatomical alignment across sessions was used to align all (functionaland anatomical) data to the anatomical reference image of the second(ocular dominance) session. The resulting maps for voxel-wise retino-topic eccentricity, polar angle, and pRF size were projected onto thesurface representations of the hemispheres (see below). Visual areas V1and V2 were delineated by the reversal of polar angle at the verticalmeridian. Delineation was possible down to an eccentricity of approxi-mately 2°. Below this limit, the polar angle did not yield sufficientinformation.

After delineation of visual areas V1 and V2, the voxelwise correla-tion values R were analyzed and compared to data acquired similarly in11 participants of a previous study at 7 T (Emmerling et al., 2016). Thespatial resolution of the 7 T datasets was 1.1 mm isotropic. The samenumber of volumes (first 300 out of the 400 available per run for the9.4 T data) was used in all pRF mapping runs, and preprocessing andanalysis were matched closely (see Emmerling et al. (2016) for furtherdetails).

Preprocessing of the ocular dominance functional data only in-cluded 3D-motion correction and linear trend removal. Individual runswere aligned with one another and co-registered with the T1-weightedanatomical acquisition using positional information. Fine-tuning of thealignment was performed manually using image edge information. Ageneral linear model was calculated using the right and left eyecondition predictors after convolution with a standard 2-gamma-function hemodynamic response function. Additionally, motion para-meters were used as confound predictors (three translational and threerotational per run). As a measure for eye preference, we calculatedselectivity, s, as the normalized relative contribution between theBOLD responses to the right and left eye (s = (βR−βL)/( βR + βL)).The following three criteria were used to identify voxels with reason-able characteristics to analyze eye preference on: (1) Voxels had to havea small eccentricity in the pRF measurement below 5° visual angle.This limit was chosen because participants reported ocular rivalry“creeping in” from peripheral visual field areas, such that, at times,participants would perceive the color-shaded gray background in thenon-stimulated eye more pronounced than the checkerboards in thestimulated eye. Such effects are more likely to occur at highereccentricity (Blake et al., 1992). (2) BOLD responses to right or lefteye individually had to be significant (p < 0.001) and both positive torule out areas of negative BOLD responses. (3) BOLD signal changesabove a threshold of 7 % were discarded as likely macrovascularcontributions. The group of voxels that fulfilled the criteria 1–3 (smalleccentricity, small BOLD % signal change, significant activation forright or left eye individually) was considered to have sufficient signalquality for ocular dominance to be observed if present.

Permuting the spatial distribution of the selectivity index s allowedus to assess the null distribution of the median absolute s (median(|s|))in V1 and V2 and thus the significance of the difference between V1 andV2. Spatial permutations were created by swapping at random N voxelsfrom V1 and V2 (where N was determined as 70% of the smaller

V.G. Kemper et al. NeuroImage 164 (2018) 48–58

56

number of voxels fulfilling criteria 1–3 in V1 and V2). This procedurewas repeated 1000 times to obtain the empirical null distribution ofselectivity values against which the selectivity of V1 and V2 werecompared (see Supplementary Fig. S4 for results).

The anatomical high-resolution T2*-weighted images were co-

registered to one another and then averaged. Prior to that, the imageswere cropped such that no aliasing in phase-encoding direction waspresent. The averaged dataset was divided by a strongly spatiallysmoothed version of it (Garcia, 2010). This step retained localanatomical contrast but eliminated low-spatial-frequency image inho-mogeneity stemming from the RF coil profiles. The brain was maskedfor this step based on intensity so that the local average was notreduced at the edges. The resulting image was co-registered with theMPRAGE anatomical reference.

The T1-weighted anatomical reference images were divided by theproton density weighted images. Further, low-spatial frequency inho-mogeneity correction was applied. Automatic image segmentation ofWM and GM was performed and refined manually. Finally, thehemispheres were separated and individual surface representationswere created from the mid-GM sheet.

Acknowledgements

We would like to thank Scott Schillak (Life Services, LLC), Dr.Benedikt Poser, Dr. Desmond Tse, and Dr. Christopher Wiggins(Scannexus) for technical support, and Faruk Gulban for help withthe analysis. This study was supported by European Research Council(ERC) Grant 269853, European FET Flagship project ‘Human BrainProject’ FP7-ICT-2013-FET-F/604102, and the National Institute ofBiomedical Imaging and Bioengineering (NIBIB) P41 EB015894.F.D.M. was funded by NWO VIDI (Grant 864-13-012).

Appendix A. Supplementary material

Supplementary data associated with this article can be found in theonline version at http://dx.doi.org/10.1016/j.neuroimage.2017.03.058.

References

Ahveninen, J., Chang, W.-T., Huang, S., Keil, B., Kopco, N., Rossi, S., Bonmassar, G.,Witzel, T., Polimeni, J.R., 2016. Intracortical depth analyses of frequency-sensitiveregions of human auditory cortex using 7 T fMRI. Neuroimage.

Andersson, J.L., Skare, S., Ashburner, J., 2003. How to correct susceptibility distortionsin spin-echo echo-planar images: application to diffusion tensor imaging.Neuroimage 20, 870–888.

Andrade, A., Kherif, F., Mangin, J.F., Worsley, K.J., Paradis, A.L., Simon, O., Dehaene,S., Le Bihan, D., Poline, J.B., 2001. Detection of fMRI activation using corticalsurface mapping. Hum. Brain Mapp. 12, 79–93.

Blake, R., O'Shea, R.P., Mueller, T., 1992. Spatial zones of binocular rivalry in central andperipheral vision. Vis. Neurosci. 8, 469–478.

Bok, S.T., 1929. Der Einfluß der in den Furchen und Windungen auftretendenKrümmungen der Großhirnrinde auf die Rindenarchitektur. Z. Die Gesamte Neurol.Psychiatr. 121, 682–750.

Bridge, H., Clare, S., Jenkinson, M., Jezzard, P., Parker, A.J., Matthews, P.M., 2005.Independent anatomical and functional measures of the V1/V2 boundary in humanvisual cortex. J. Vis. 5, 1.

Budde, J., Shajan, G., Hoffmann, J., Ugurbil, K., Pohmann, R., 2011. Human imaging at9.4 T using T(2) *-, phase-, and susceptibility-weighted contrast. Magn. Reson. Med.65, 544–550.

Budde, J., Shajan, G., Zaitsev, M., Scheffler, K., Pohmann, R., 2014. Functional MRI inhuman subjects with gradient‐echo and spin‐echo EPI at 9.4 T. Magn. Reson. Med.71, 209–218.

Cheng, K., Waggoner, R.A., Tanaka, K., 2001. Human ocular dominance columns asrevealed by high-field functional magnetic resonance imaging. Neuron 32, 359–374.

Dale, A.M., Fischl, B., Sereno, M.I., 1999. Cortical surface-based analysis: i.Segmentation and surface reconstruction. Neuroimage 9, 179–194.

De Martino, F., Moerel, M., Ugurbil, K., Formisano, E., Yacoub, E., 2014a. Less noise,more activation: multiband acquisition schemes for auditory functional MRI. Magn.Reson. Med..

De Martino, F., Moerel, M., Ugurbil, K., Goebel, R., Yacoub, E., Formisano, E., 2015.Frequency preference and attention effects across cortical depths in the humanprimary auditory cortex. Proc. Natl. Acad. Sci. USA 112, 16036–16041.

De Martino, F., Moerel, M., Xu, J., van de Moortele, P.F., Ugurbil, K., Goebel, R., Yacoub,E., Formisano, E., 2014b. High-resolution mapping of myeloarchitecture in vivo:localization of auditory areas in the human brain. Cereb. Cortex.

De Martino, F., Schmitter, S., Moerel, M., Tian, J., Ugurbil, K., Formisano, E., Yacoub, E.,de Moortele, P.F., 2012. Spin echo functional MRI in bilateral auditory cortices at7 T: an application of B(1) shimming. Neuroimage 63, 1313–1320.

De Martino, F., Zimmermann, J., Muckli, L., Ugurbil, K., Yacoub, E., Goebel, R., 2013.Cortical depth dependent functional responses in humans at 7 T: improvedspecificity with 3D GRASE. PLoS One 8, e60514.

Deichmann, R., Good, C.D., Josephs, O., Ashburner, J., Turner, R., 2000. Optimization of3-D MP-RAGE sequences for structural brain imaging. Neuroimage 12, 112–127.

Dumoulin, S.O., Wandell, B.A., 2008. Population receptive field estimates in humanvisual cortex. Neuroimage 39, 647–660.

Duong, T.Q., Kim, D.S., Ugurbil, K., Kim, S.G., 2001. Localized cerebral blood flowresponse at submillimeter columnar resolution. Proc. Natl. Acad. Sci. USA 98,10904–10909.

Duong, T.Q., Yacoub, E., Adriany, G., Hu, X., Ugurbil, K., Kim, S.G., 2003. MicrovascularBOLD contribution at 4 and 7 T in the human brain: gradient-echo and spin-echofMRI with suppression of blood effects. Magn. Reson. Med. 49, 1019–1027.

Duong, T.Q., Yacoub, E., Adriany, G., Hu, X., Ugurbil, K., Vaughan, J.T., Merkle, H., Kim,S.G., 2002. High‐resolution, spin‐echo BOLD, and CBF fMRI at 4 and 7 T. Magn.Reson. Med. 48, 589–593.

Duyn, J.H., van Gelderen, P., Li, T.-Q., de Zwart, J.A., Koretsky, A.P., Fukunaga, M.,2007. High-field MRI of brain cortical substructure based on signal phase. Proc.Natl. Acad. Sci. USA 104, 11796–11801.

Emmerling, T.C., Zimmermann, J., Sorger, B., Frost, M.A., Goebel, R., 2016. Decodingthe direction of imagined visual motion using 7 T ultra-high field fMRI. Neuroimage125, 61–73.

Feinberg, D.A., Harel, N., Ramanna, S., Ugurbil, K., Yacoub, E., 2008. Sub-millimeterSingle-shot 3D GRASE with Inner Volume Selection for T2 weighted fMRIapplications at 7 Tesla. In: Proceedings of the 16th Annual Meeting InternationalSociety for Magnetic Resonance in Medicine, 2373.

Feinberg, D.A., Oshio, K., 1991. GRASE (gradient- and spin-echo) MR imaging: a newfast clinical imaging technique. Radiology 181, 597–602.

Fischl, B., Dale, A.M., 2000. Measuring the thickness of the human cerebral cortex frommagnetic resonance images. Proc. Natl. Acad. Sci. USA 97, 11050–11055.

Fracasso, A., Petridou, N., Dumoulin, S.O., 2016. Systematic variation of populationreceptive field properties across cortical depth in human visual cortex. Neuroimage139, 427–438.

Fritz, L., Mulders, J., Bremann, H., Peters, J., Bastiani, M., Roebroeck, A., Andersson, J.L., Ashburner, J., Weiskopf, N., Goebel, R., 2014. Comparison of EPI distortioncorrection methods at 3T and 7T. OHBM Annual Meeting, Hamburg ,Germany.

Frost, M.A., Goebel, R., 2012. Measuring structural-functional correspondence: spatialvariability of specialised brain regions after macro-anatomical alignment.Neuroimage 59, 1369–1381.

Fukunaga, M., Li, T.-Q., van Gelderen, P., de Zwart, J.A., Shmueli, K., Yao, B., Lee, J.,Maric, D., Aronova, M.A., Zhang, G., Leapman, R.D., Schenck, J.F., Merkle, H.,Duyn, J.H., 2010. Layer-specific variation of iron content in cerebral cortex as asource of MRI contrast. Proc. Natl. Acad. Sci. USA 107, 3834–3839.

Garcia, D., 2010. Robust smoothing of gridded data in one and higher dimensions withmissing values. Comput. Stat. Data Anal. 54, 1167–1178.

Goebel, R., 2015. Functional organization of the primary visual cortex. In: Toga, A.W.(Ed.), Brain Mapping: An Encyclopedic Reference. Academic Press: Elsevier,Amsterdam, 287–291.

Goebel, R., Khorram-Sefat, D., Muckli, L., Hacker, H., Singer, W., 1998. The constructivenature of vision: direct evidence from functional magnetic resonance imaging studiesof apparent motion and motion imagery. Eur. J. Neurosci. 10, 1563–1573.

Goense, J., Bohraus, Y., Logothetis, N.K., 2016. fMRI at high spatial resolution:implications for BOLD-models. Front Comput. Neurosci., 10.

Goense, J.B., Logothetis, N.K., 2006. Laminar specificity in monkey V1 using high-resolution SE-fMRI. Magn. Reson. Imaging 24, 381–392.

Goncalves, N.R., Ban, H., Sanchez-Panchuelo, R.M., Francis, S.T., Schluppeck, D.,Welchman, A.E., 2015. 7 tesla FMRI reveals systematic functional organization forbinocular disparity in dorsal visual cortex. J. Neurosci. 35, 3056–3072.

Goodyear, B.G., Menon, R.S., 2001. Brief visual stimulation allows mapping of oculardominance in visual cortex using fMRI. Hum. Brain Mapp. 14, 210–217.

Greve, D.N., Fischl, B., 2009. Accurate and robust brain image alignment usingboundary-based registration. Neuroimage 48, 63–72.

Guidi, M., Huber, L., Lampe, L., Gauthier, C.J., Möller, H.E., 2016. Lamina-dependentcalibrated BOLD response in human primary motor cortex. Neuroimage 141,250–261.

Huber, L., Goense, J., Kennerley, A.J., Trampel, R., Guidi, M., Reimer, E., Ivanov, D.,Neef, N., Gauthier, C.J., Turner, R., Moller, H.E., 2015. Cortical lamina-dependentblood volume changes in human brain at 7 T. Neuroimage 107, 23–33.

Jo, H.J., Lee, J.-M., Kim, J.-H., Shin, Y.-W., Kim, I.-Y., Kwon, J.S., Kim, S.I., 2007.Spatial accuracy of fMRI activation influenced by volume-and surface-based spatialsmoothing techniques. Neuroimage 34, 550–564.

Jones, S.E., Buchbinder, B.R., Aharon, I., 2000. Three-dimensional mapping of corticalthickness using Laplace's equation. Hum. Brain Mapp. 11, 12–32.

Kemper, V.G., De Martino, F., Tse, D.H., Poser, B.A., Yacoub, E., Goebel, R., 2015a. T2weighted high-resolution fMRI in human visual cortex at 9.4 T using 3D-GRASE.23nd Annual Meeting of ISMRM 23, 1944.

Kemper, V.G., De Martino, F., Vu, A.T., Poser, B.A., Feinberg, D.A., Goebel, R., Yacoub,E., 2015b. Sub-millimeter T2 weighted fMRI at 7 T: comparison of 3D-GRASE and2D SE-EPI. Front. Neurosci., 9.

Kemper, V.G., De Martino, F., Yacoub, E., Goebel, R., 2016. Variable flip angle 3D-

V.G. Kemper et al. NeuroImage 164 (2018) 48–58

57

GRASE for high resolution fMRI at 7 tesla. Magn. Reson. Med. 76, 897–904.Kiebel, S.J., Goebel, R., Friston, K.J., 2000. Anatomically informed basis functions.

Neuroimage 11, 656–667.Kok, P., Bains, L.J., van Mourik, T., Norris, D.G., de Lange, F.P., 2016. Selective

activation of the deep layers of the human primary visual cortex by top-downfeedback. Curr. Biol..

Koopmans, P.J., Barth, M., Norris, D.G., 2010. Layer-specific BOLD activation in humanVI. Hum. Brain Mapp. 31, 1297–1304.

Koopmans, P.J., Barth, M., Orzada, S., Norris, D.G., 2011. Multi-echo fMRI of thecortical laminae in humans at 7 T. Neuroimage 56, 1276–1285.

Markuerkiaga, I., Barth, M., Norris, D.G., 2016. A cortical vascular model for examiningthe specificity of the laminar BOLD signal. Neuroimage 132, 491–498.

Marques, J.P., Kober, T., Krueger, G., van der Zwaag, W., Van de Moortele, P.F.,Gruetter, R., 2010. MP2RAGE, a self bias-field corrected sequence for improvedsegmentation and T1-mapping at high field. Neuroimage 49, 1271–1281.

Muckli, L., De Martino, F., Vizioli, L., Petro, L.S., Smith, F.W., Ugurbil, K., Goebel, R.,Yacoub, E., 2015. Contextual feedback to superficial layers of V1. Curr. Biol..

Mugler, J.P., Brookeman, J.R., 1990. Three‐dimensional magnetization‐prepared rapidgradient‐echo imaging (3D MP RAGE). Magn. Reson. Med. 15, 152–157.

Nehrke, K., Börnert, P., 2012. DREAM—a novel approach for robust, ultrafast, multisliceB1 mapping. Magn. Reson. Med. 68, 1517–1526.

Nieuwenhuys, R., 2013. The myeloarchitectonic studies on the human cerebral cortex ofthe Vogt-Vogt school, and their significance for the interpretation of functionalneuroimaging data. Microstructural Parcellation of the Human Cerebral Cortex.Springer, pp. 55–125.

Peirce, J.W., 2007. PsychoPy—psychophysics software in Python. J. Neurosci. Methods162, 8–13.

Pohmann, R., Speck, O., Scheffler, K., 2015. Signal-to-noise ratio and MR tissueparameters in human brain imaging at 3, 7, and 9.4 tesla using current receive coilarrays. Magn. Reson. Med..

Polimeni, J.R., Fischl, B., Greve, D.N., Wald, L.L., 2010. Laminar analysis of 7T BOLDusing an imposed spatial activation pattern in human V1. Neuroimage 52,1334–1346.

Poser, B.A., Anderson, R.J., Guérin, B., Setsompop, K., Deng, W., Mareyam, A., Serano,P., Wald, L.L., Stenger, V.A., 2014. Simultaneous multislice excitation by paralleltransmission. Magn. Reson. Med. 71, 1416–1427.

Renvall, V., Witzel, T., Wald, L.L., Polimeni, J.R., 2016. Automatic cortical surfacereconstruction of high-resolution T 1 echo planar imaging data. Neuroimage 134,338–354.

Ress, D., Glover, G.H., Liu, J., Wandell, B., 2007. Laminar profiles of functional activityin the human brain. Neuroimage 34, 74–84.

Sánchez‐Panchuelo, R.M., Francis, S.T., Schluppeck, D., Bowtell, R.W., 2012.Correspondence of human visual areas identified using functional and anatomicalMRI in vivo at 7 T. J. Magn. Reson. Imaging 35, 287–299.

Scheeringa, R., Koopmans, P.J., van Mourik, T., Jensen, O., Norris, D.G., 2016. Therelationship between oscillatory EEG activity and the laminar-specific BOLD signal.Proc. Natl. Acad. Sci., 201522577.

Senden, M., Reithler, J., Gijsen, S., Goebel, R., 2014. Evaluating population receptivefield estimation frameworks in terms of robustness and reproducibility. PLoS One 9,e114054.

Tse, D.H., Poole, M.S., Magill, A.W., Felder, J., Brenner, D., Jon Shah, N., 2014.Encoding methods for B1(+) mapping in parallel transmit systems at ultra high field.J. Magn. Reson. 245, 125–132.

Tse, D.H., Wiggins, C.J., Ivanov, D., Brenner, D., Hoffmann, J., Mirkes, C., Shajan, G.,Scheffler, K., Uludağ, K., Poser, B.A., 2016. Volumetric imaging with homogenisedexcitation and static field at 9.4 T. Magn. Reson. Mater. Phys. Biol. Med., 1–13.

Ugurbil, K., Adriany, G., Andersen, P., Chen, W., Garwood, M., Gruetter, R., Henry, P.G.,Kim, S.G., Lieu, H., Tkac, I., Vaughan, T., Van De Moortele, P.F., Yacoub, E., Zhu,X.H., 2003. Ultrahigh field magnetic resonance imaging and spectroscopy. Magn.Reson. Imaging 21, 1263–1281.

Uludag, K., Muller-Bierl, B., Ugurbil, K., 2009. An integrative model for neuronalactivity-induced signal changes for gradient and spin echo functional imaging.Neuroimage 48, 150–165.

Van de Moortele, P.F., Auerbach, E.J., Olman, C., Yacoub, E., Ugurbil, K., Moeller, S.,2009. T1 weighted brain images at 7 Tesla unbiased for Proton Density, T2* contrastand RF coil receive B1 sensitivity with simultaneous vessel visualization.Neuroimage 46, 432–446.

van der Kouwe, A.J., Benner, T., Salat, D.H., Fischl, B., 2008. Brain morphometry withmultiecho MPRAGE. Neuroimage 40, 559–569.

Waehnert, M.D., Dinse, J., Weiss, M., Streicher, M.N., Waehnert, P., Geyer, S., Turner,R., Bazin, P.L., 2014. Anatomically motivated modeling of cortical laminae.Neuroimage 93 (Pt 2), 210–220.

Wu, X., Schmitter, S., Auerbach, E.J., Moeller, S., Uğurbil, K., de Moortele, V., 2013.Simultaneous multislice multiband parallel radiofrequency excitation withindependent slice‐specific transmit B1 homogenization. Magn. Reson. Med. 70,630–638.

Yacoub, E., Duong, T.Q., Van De Moortele, P.F., Lindquist, M., Adriany, G., Kim, S.G.,Ugurbil, K., Hu, X., 2003. Spin-echo fMRI in humans using high spatial resolutionsand high magnetic fields. Magn. Reson. Med. 49, 655–664.

Yacoub, E., Harel, N., Ugurbil, K., 2008. High-field fMRI unveils orientation columns inhumans. Proc. Natl. Acad. Sci. USA 105, 10607–10612.

Yacoub, E., Shmuel, A., Logothetis, N., Ugurbil, K., 2007. Robust detection of oculardominance columns in humans using Hahn Spin Echo BOLD functional MRI at 7Tesla. Neuroimage 37, 1161–1177.

Yacoub, E., Shmuel, A., Pfeuffer, J., De Moortele, V., Adriany, G., Andersen, P., Vaughan,J.T., Merkle, H., Ugurbil, K., Hu, X., 2001. Imaging brain function in humans at 7Tesla. Magn. Reson. Med. 45, 588–594.

Zhao, F., Wang, P., Kim, S.G., 2004. Cortical depth‐dependent gradient‐echo and spin‐echo BOLD fMRI at 9.4 T. Magn. Reson. Med. 51, 518–524.

Zimmermann, J., Goebel, R., De Martino, F., van de Moortele, P.F., Feinberg, D.,Adriany, G., Chaimow, D., Shmuel, A., Ugurbil, K., Yacoub, E., 2011. Mapping theorganization of axis of motion selective features in human area MT using high-fieldfMRI. PLoS One 6, e28716.

V.G. Kemper et al. NeuroImage 164 (2018) 48–58

58