Functional MRI-guided probabilistic tractography of cortico-cortical and cortico-subcortical language networks in children

NeuroImage 63 (2012) 1561–1570 Contents lists available at SciVerse ScienceDirect NeuroImage journal homepage: www.elsevier.com/locate/ynimg Functional MRI-guided probabilistic tractography of cortico-cortical and cortico-subcortical language networks in children Philip Julian Broser a, b,⁎, Samuel Groeschel a, b, Till-Karsten Hauser c, Karen Lidzba a, b, Marko Wilke a, b a b c Pediatric Neurology & Developmental Medicine, University of Tübingen, Germany Experimental Pediatric NeuroImaging, Children's Hospital, University of Tübingen, Germany Department of Neuroradiology, Radiological Clinic, University of Tübingen, Germany a r t i c l e i n f o Article history: Accepted 28 July 2012 Available online 3 August 2012 Keywords: Arcuate fasciculus Constrained spherical deconvolution Functional MRI Language Lateralization Thalamus Caudate nucleus Probabilistic tractography Diffusion MR imaging a b s t r a c t In this study, we analyzed the structural connectivity of cortico-cortical and cortico-subcortical language networks in healthy children, using probabilistic tractography based on high angular resolution diffusion imaging. In addition to anatomically defining seed and target regions for tractography, we used fMRI to target inferior frontal and superior temporal cortical language areas on an individual basis. Further, connectivity between these cortical and subcortical (thalamus, caudate nucleus) language regions was assessed. Overall, data from 15 children (8f) aged 8–17 years (mean age 12.1 ± 3 years) could be included. A slight but non-significant trend towards leftward lateralization was found in the arcuate fasciculus/superior longitudinal fasciculus (AF/SLF) using anatomically defined masks (p> .05, Wilcoxon rank test), while the functionallyguided tractography showed a significant lateralization to the left (p b .01). Connectivity of the thalamus with language regions was strong but not lateralized. Connectivity of the caudate nucleus with inferior-frontal language regions was also symmetrical, while connectivity with superior-temporal language regions was strongly lateralized to the left (pb .01). To conclude, we could show that tracking the arcuate fasciculus/superior longitudinal fasciculus is possible using both anatomically and functionally-defined seed and target regions. With the latter approach, we could confirm the presence of structurally-lateralized cortico-cortical language networks already in children, and finally, we could demonstrate a strongly asymmetrical connectivity of the caudate nucleus with superior temporal language regions. Further research is necessary in order to assess the usability of such an approach to assess language dominance in children unable to participate in an active fMRI study. © 2012 Elsevier Inc. All rights reserved. Introduction The ability of humans to use language to communicate is unique. While the most obvious and dramatic changes in language occur in infancy and early childhood (Benedict, 1979), most studies investigating language processes were done in adults (Price, 2010). Two main cortical areas have long-since been identified to play a major part in the processing of language — an inferior frontal brain area traditionally known as Broca's area and a superior temporal brain area known as Wernicke's area. Especially frontal language regions are usually strongly lateralized Abbreviations: AF/SLF, arcuate fasciculus/superior longitudinal fasciculus; COM, center of mass; CSD, constrained spherical deconvolution; DTI, diffusion tensor imaging; DW, diffusion-weighted; fMRI, functional magnetic resonance imaging; FOD, fiber orientation distribution; EHI, Edinburgh handedness inventory; HARDI, high angular resolution diffusion imaging; LI, lateralization index; MAD, median absolute deviation; MNI, Montreal neurological institute; Ncl., nucleus; PDF, probability distribution function. ⁎ Corresponding author at: Pediatric Neurology & Developmental Medicine, University Children's Hospital, Hoppe-Seyler-Str. 1, D‐72076 Tübingen, Germany. Fax: +49 7071 29 5473. E-mail address: philip.broser@med.uni-tuebingen.de (P.J. Broser). 1053-8119/$ – see front matter © 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.neuroimage.2012.07.060 to the left hemisphere (Price, 2010). The inferior frontal and the superior temporal brain areas are strongly interconnected via at least two different white matter pathways, a dorsal pathway and a ventral pathway (Frey et al., 2008). The dorsal pathway – known as the arcuate fasciculus (AF) – is thought to play a key role in the complex interaction between the two major cortical language regions (Marslen-Wilson and Tyler, 2007). For example, it was found to underlie the processing of syntactically complex sentences (Bornkessel et al., 2005). The arcuate fasciculus is considered to be one of the four parts of the Superior Longitudinal Bundle. Burdach (1822) was the first to describe the Superior Longitudinal Bundle in humans. He also described a fiber tract that is originating from the caudal part of the superior temporal gyrus, bends around the Sylvian fissure and propagates into the frontal lobe. He called this bundle the “Fasciculus Arcuatus” and considered it part of the Superior Longitudinal Bundle. In non-human primates, four parts of the Superior Longitudinal Fasciculus (SLF I–III + AF) could be delineated; however, a separation of the arcuate fasciculus and the SLF II is not possible in humans (Makris et al., 2005), we will therefore use the recently-used convention of arcuate fasciculus/superior longitudinal fasciculus (AF/SLF) (Brauer et al., 2011) in this manuscript. 1562 P.J. Broser et al. / NeuroImage 63 (2012) 1561–1570 The extent and location of the AF/SLF can be measured noninvasively by probabilistic tractography (Frey et al., 2008). Probabilistic tractography is a technique that uses diffusion weighted magnetic resonance imaging (dMRI) data to estimate the likelihood of connection between two points in the brain (Behrens et al., 2003; Tournier et al., 2011). This technique essentially samples the distribution of the greatest diffusion directions of water in each voxel of the brain. The technique then assumes that the diffusivity is greatest along axonal fiber bundles and makes a probabilistic statement about the fiber orientation (Mori and Zhang, 2006) in each voxel. By defining seed and target masks, probabilistic tractography can estimate the likelihood of connectivity between the two brain areas. It was previously shown that the left AF/SLF is more prominent than the right in adults (Catani et al., 2005) and, very recently, also in children (Tiwari et al., 2011), mirroring the observable functional lateralization. As language networks involve white matter regions with multiple fiber populations (Behrens et al., 2007), we used high angular resolution diffusion imaging and constrained spherical deconvolution, allowing to resolve crossing fiber tracts (Tournier et al., 2008). Aside from cortical language areas, subcortical structures have been implicated in language processing in the past. For example, Riecker et al. (2000) could show a functional lateralization of the thalamus using a syllable repetition task, and Crinion et al. (2006) showed an asymmetric activation of the left caudate nucleus during language discrimination tasks in bilingual adults. Using connectivity analyses, these structures were also detected to be functionally connected with the major language areas (Wilke et al., 2009). Given that axonal connections between the head of the caudate nucleus and the superior temporal and inferior parietal cortex have previously been described in monkeys (Yeterian and Pandya, 1993) using anatomical tracers, the hypothesis is that such a pathway also exists in humans, and is already present in children. Therefore, the question arises whether connections of cortical language areas with subcortical brain structures are lateralized to a similar degree as the connection between the major language areas. More precisely, we aimed at assessing the white matter connections from the two cortical language areas to the thalamus and the caudate nucleus. As noted above, most of the data published on the arcuate fasciculus so far has been acquired in the adult population. The technique of reconstructing this fiber bundle usually relies on the generation of anatomical masks for the two language areas, derived from one or more adult brains. There are two problems with this approach in children: one, the pediatric brain continues to change substantially on the structural level, with significant age-related changes occurring well into adolescence (Castellanos et al., 2002; Giedd et al., 1999; Groeschel et al., 2010; Wilke et al., 2007). Two, there are also substantial changes occurring regarding the functional layout of language in the developing brain (Lidzba et al., 2011; Szaflarski et al., 2006). For example, while lateralized activation in language cortex was already seen in babies (Hertz-Pannier et al., 2002), lateralization continues to increase during childhood (Holland et al., 2001). Therefore, it seems important to develop an approach that takes these differences into account in order to avoid processing bias. To this effect, using functional localizers (Saxe et al., 2006) seems to be a promising alternative. The objective of this study therefore was twofold: one, following reproduction of previous results using anatomically-defined masks, we aimed at deriving functionally-defined masks for tracing the AF/ SLF. Two, we wanted to assess whether connections between the major cortical and subcortical language areas are already lateralized in healthy children. Subjects and methods Participants A total of 43 healthy children were examined during the study; however, the full extensive (structural, functional, and diffusion) dataset could only be acquired in 15 children (8 female) with sufficient quality. The children were aged between 8 and 17 years (mean age 12.1 years; SD = 3 years). As described in Ebner et al. (2011), participants were recruited via public announcements and newspaper articles. Interested parents underwent telephone screening with the study coordinator. General MR contraindications applied; study specific exclusion criteria were the presence of neurological or psychiatric disorders, hearing deficits, cognitive impairment, or prematurity (birth before 37 weeks of gestational age). All children were native German speakers and had normal or corrected-to-normal vision. Handedness was assessed using the Edinburgh Handedness Inventory (EHI; Oldfield, 1971). There were 14 right- and 1 left-handed subjects. Neuropsychological testing revealed cognitive abilities in the normal range in all subjects, as assessed using the HAWIK IV, a German version of the Wechsler Intelligence Scale for children (Petermann and Petermann, 2007). All procedures were approved by the local institutional review board. All parents gave written informed consent and all children gave assent. Children were compensated for their participation. Anatomical images were found to be free of structural abnormalities as determined by an experienced pediatric neuroradiologist. Data acquisition Data was acquired on a 1.5-T whole body MR scanner (Avanto, Siemens Medizintechnik, Erlangen, Germany), using a 12-channel head coil. A T2-weightened echo-planar imaging (EPI) sequence was used to acquire functional images with the following parameters: TR = 3000 ms, TE = 40 ms, matrix = 64 × 64, 40 slices, covering the whole brain, yielding a voxel size of 3 × 3 × 3 mm 3. For each task, 110 images were acquired. Additionally, a gradient-echo B0 fieldmap was acquired with TR=546 ms, TE=5.19/9.95 ms, with the same slice prescription as the functional series. An anatomical T1-weightened 3D-data set with TR=1300 ms, TE=2.92 ms was also acquired, yielding 176 contiguous slices with an in-plane matrix of 256×256, resulting in a voxel size of 1×1×1 mm3. Diffusion-weighted datasets were acquired using a high angular resolution (twice-refocused) EPI sequence with TR=6900 ms, TE=109 ms, using an isotropic set of 60 non-collinear directions, a diffusion-weighting factor of b=2000 s mm−2 (one image with b=0), with 45 contiguous axial slices with matrix=80×80, resulting in a voxel size of 2.5×2.5×2.5 mm3. Care was taken to ensure comfortable placement of the children and a foam cushion was used to minimize head movement. Visual stimuli were presented on an MRcompatible screen while acoustic stimulation was achieved via MRcompatible headphones (MR-Confon, Magdeburg, Germany). Responses were recorded using MR-compatible pushbuttons (Current Design, Philadelphia, PA, USA). fMRI tasks The children performed two previously published functional MRI tasks, the vowel identification task (VIT; Wilke et al., 2006) and the beep stories task (BST; Wilke et al., 2005). In order to assess productive, frontal language areas, the VIT was used. In this block-design paradigm, the active conditions aimed at inducing phonological processing: here, the picture of a concrete object was presented and the children were required to decide whether the vowel “i” (always pronounced /i:/ in German) was present in the name of the object. If so, a button press was required. The control condition was a visuospatial task of equivalent complexity with no phonological processing; this data was not used here. Task adherence is monitored as described by Ebner et al. (2011), and subjects performing below chance level were excluded. The VIT shows a strongly lateralized inferior frontal activation pattern (involving Broca's area; Wilke et al., 2006) in the dominant hemisphere. In order to assess perceptive, posterior-temporal language areas, the BST was used (Wilke et al., 2005). This is a modified story-listening task P.J. Broser et al. / NeuroImage 63 (2012) 1561–1570 (Karunanayaka et al., 2007), consisting of short stories (30 s) that alternate with sinus tones of different durations in the frequency range of human language (30 s). In order to increase attention, 6–8 key words were removed from each story, which induces an increase in frontal activation (Wilke et al., 2005). Task performance was monitored by quizzing the child after the scan with a short question for each of the 5 stories, and data from subjects with correct answers below chance level was discarded. The BST shows a more bilateral, but slightly left-dominant superior-temporal and inferior-parietal activation pattern (involving Wernicke's area; Lidzba et al., 2011; Wilke et al., 2005). Data processing Processing of fMRI data As described by Ebner et al. (2011), fMRI data was processed using SPM8 software (Wellcome Trust Centre for Neuroimaging, London, UK), running in Matlab (Mathworks, Natick, MA, USA). The first 10 scans of each functional series (corresponding to the first block of the control condition) were rejected to allow for the stabilization of longitudinal magnetization, leaving 100 scans per series (5 blocks each of the active and the control condition). Functional images were realigned and unwarped using the individually-acquired B0 fieldmap, correcting for both EPI-distortions and motion ∗ B0 interactions (Andersson et al., 2001). Any functional series with motion exceeding one voxel size (3 mm) in any direction were rejected. Global signal trends were removed (Macey et al., 2004) and the functional images were smoothed with a 6 mm full width at half maximum (FWHM) Gaussian filter. First level analyses were done using the General Linear Model (Friston et al., 1995), contrasting the active condition with the control condition. Only voxels with a t-value greater than a given threshold were kept for further analysis. The threshold was calculated such that the FDR was controlled to 5% for each subject. Data pre-processing for probabilistic diffusion tractography The diffusion-weighted (DW) data was processed using constrained spherical deconvolution (CSD) (Tournier et al., 2007) as implemented in mrtrix (J-D Tournier, Brain Research Institute, Melbourne, Australia, http://www.brain.org.au/software/). The fiber orientation distribution (FOD) is estimated by spherical deconvolution of the DW signal assuming that the DW signal measured from any fiber bundle is adequately described by a single response function (Tournier et al., 2004). This method has shown to provide FOD estimates that are robust to noise while preserving angular resolution and allowing tracking in regions of crossing fibers (Tournier et al., 2007, 2008). CSD was performed with the maximum harmonic order set to 8. T1-weighted MRI data processing The T1-weightened 3D-data was segmented by probabilistic tissue type segmentation. Only voxels with an estimated percentage of gray matter of more than 35% were included as targets (Smith, 2002). The T1-weighted data was skull-stripped (Smith, 2002). The diffusionweighted data was registered to the T1-weighted data by affine registration (Jenkinson and Smith, 2001) to derive the transformation matrices. The T1-weighted data was then registered, also by affine registration (Jenkinson and Smith, 2001) to the Montreal Neurological Institute (MNI) single brain space in order to calculate group averages and to give MNI coordinates. The high-resolution, single-subject MNI brain was used in order to optimize the later denormalization of standard brain regions (Tzourio-Mazoyer et al., 2002). In previous works it could be shown that the scaling from the affine only registration does not correlate significantly with age in this age range (Wilke et al., 2002). Thalamus and caudate nucleus segmentation For each subject, the left and right thalamus and the two caudate nuclei were segmented from the T1-weighted 3D data by using the FIRST tool from the FMRIB Software Library (Smith et al., 2004). The 1563 segmentation was visually validated and compared against the Harvard-Subcortical-Structural Atlas (http://www.cma.mgh.harvard. edu/) for each subject. The FIRST tool segmented both structures accurately so that no further manual editing was necessary. Masks for tractography Tractography requires seed and target masks in order to estimate the connectivity between two brain regions. In this study, two different types of masks were used: masks derived from anatomical templates and masks created individually for each subject, based on the fMRIguided locations of the two major cortical language brain areas. Anatomical masks Anatomical masks were defined on the reference brain by selecting Brodmann areas (BA) of interest. We chose BA45 and 44 for the inferior frontal language regions and BA22 for the superior temporal language area. These masks were transferred to each individual subject's brain by using the inverse of the affine registration matrix (Jenkinson and Smith, 2001). Functional language region localizers To create tractography masks based on the individual functional activation pattern, the following approach was used: first, two broad search masks were defined on the reference brain, restricting the search volume to anterior (frontal lobe without insular cortex) and posterior (temporal plus posterior parietal lobe) brain regions. These search masks were transferred to each individual space using affine transformation (Jenkinson and Smith, 2001) by applying the calculated transformation matrices. Then, the weighted center of mass was calculated from all significant voxels within the mask, according to: COM ¼  → → 1  → ∑ exp Tmap x x: x∈Tmap ∑ exp Tmap x This was done using the VIT t-map for the anterior mask and the BST t-map for the posterior mask, resulting in individual, functionallydefined activation foci for each subject for both the major anterior and posterior language regions in the dominant hemisphere. Mirroring the localization to the non‐dominant hemisphere Both fMRI tasks only show reliable activation on the dominant hemisphere. In order to compare corresponding brain regions of both hemispheres, the locations for the two language regions from the dominant hemisphere were mirrored onto the non-dominant (right) side by defining a reflection plane on the reference brain which was projected into each individual's native space. This plane is characterized by the origin — the anterior commisure (MNI coordinates 0,0,0) and two direction vectors (one pointing posterior to anterior and one pointing caudal to cranial). These three vectors were transferred to each individual t-map (T1) space by using the derived transformation matrices. Using the individual reflection plane and basic linear algebra, the COM locations on the dominant side were automatically reflected on to the nondominant hemisphere (as done before; Wilke et al., 2009). Creation of mask using the estimated locations Around each of the four locations on the left and right hemisphere a spherical region of interest (ROI) was drawn with a radius of 19 mm, corresponding to the same volume as used previously (Wilke et al., 2009). Arcuate fasciculus inclusion mask As described previously by Catani et al. (2005), an inclusion mask to select only the AF/SLF was used. It was medially restricted by the 1564 P.J. Broser et al. / NeuroImage 63 (2012) 1561–1570 corona radiata and laterally by the cortex, extending from MNI Z = 6 to 38 mm. To avoid introducing a bias, the left mask was reflected to the right and the right to the left and a symmetrical conjunction of both masks was used. CSD based probabilistic tractography Tractography for the anatomical and functional based arcuate fasciculus The constructed brain masks were then used as seed masks for the CSD-based probabilistic tractography, using the tool streamtrack from the mrtrix package. Tracking was seeded in the superior temporal brain mask and probabilistic tractography started, with the desired number of tracks set to 100,000. Initially, whole brain tracking was performed, and tracks terminated if they left the brain or if the length of the track exceeded 20 cm or if the minimum curvature was below 1 mm (all default values). In a second step, only tracks were selected that originate from the superior temporal brain mask, pass through the inclusion mask, and propagate into the inferior frontal brain mask (although tracks did not necessarily need to terminate here). In addition, the tool dwi2tensor using a log-linear least square model (Basser et al., 1996) was used to calculate a diffusion tensor map which was used to calculate the mean diffusivity (MD), and fractional anisotropy (FA). Tracks were analyzed with regard to the number of tracks, the mean MD and mean FA along the tracks. Additionally, a lateralization index (LI) was calculated, using the number of tracks: Probability distribution functions As described previously described by Broser et al. (2011), probability distribution functions (PDFs) were calculated from the calculated number of tracks to localize the dominant projection areas and analyze the substructure of the thalamus and caudate nucleus in more detail. Briefly, the left and right thalami and the left and right caudate nucleus were segmented from the T1 Imaging using FIRST from the FSL Toolbox (Smith, 2002). For each voxel in the structure, the number of tracks originating and propagating into the given cortical target structure was counted (as obtained from streamtrack), and the total number of tracks was summed over all voxel (of both sides). Finally, the number in each voxel was divided by the sum, yielding the probability density function. These functions were visualized using transparent volume rendering with a self written program linking against the VTK library. Statistics Due to partially small group sizes, the median and the median absolute deviation (MAD) were used throughout. The total number of tracks (as defined above) was tested for statistically significant differences relating to hemisphericity using a paired Wilcoxon rank test. The same procedure was applied for the thus derived mean FA and mean MD values. All statistical analyses were performed with R (R Development Core Team, 2010). Significance level was set to p ≤.05; values of .05 ≤p ≤ .1 were considered trends. Results LNr−RNr : LNr þ RNr This index is normalized and ranges from +1 (maximum left lateralization) to 0 (symmetric) to −1 (maximum right lateralization; Wilke and Lidzba, 2007). Values of −0.2b LIb 0.2 are considered bilateral (Wilke et al., 2006). Thalamus and caudate nucleus probabilistic tractography To trace the connections between the thalamus and the caudate nucleus and the two cortical language areas, a similar approach was taken. Tractography was initiated in the thalamus and the caudate nucleus as described above. In a second step, tracks were selected and analyzed if they propagated into the desired cortical target mask, in this case without the need to pass through an inclusion mask. Three dimensional visualization For the three dimensional visualization of the data a self-written program (TrackVis) based on the VTK (www.vtk.org) software library was used. The brainstem, the putamen, and the pallidum were extracted from the MNI 1 mm single brain data set using the program FIRST (Patenaude, 2007) from the FSL library. The generated images were loaded as NIFTI (http://nifti.nimh.nih.gov/) images, converted into a 3D vector graphic and visualized. The estimated tracks were visualized as VTK streamlines. Group averages were calculated by creating individual track density images using the program tracks2prob from the mrtrix tool box. The tool calculates the fraction of tracks that enter each voxel, resulting in a density map [tracks per voxel] and transferring the probability images into MNI space. Here, voxelwise median track probability values were calculated. The data was visualized by isosurfaces with a threshold of 50% of the maximum value. Anatomical arcuatus Using the anatomically-defined masks, we could show a lateralization of the arcuate fasciculus to the left side only at trend level in this sample, see Fig. 1 (panel A for an example of a single subject, panels B and C for group average and panel D for a box-whisker plot). The corresponding number of tracks (and their MAD) was 236 (182) for the left and 117 (77) for the right (p≥.05). The resulting LI was .23 (.33), slightly left-lateralized. Remark: Fa values are given in units of 10−1 and MD values are given in units of 10−4 mm2 s−1. The corresponding median FA (and their MAD) was 4.38 (.133) for the left and 4.17 (.197) for the right (p = .0181). The corresponding median MD (and their MAD) was 6.47 (.213) for the left and 6.41 (.284) for the right (p = .0302, see Inline Supplementary Table S-I). Inline Supplementary Table S1 can be found online at http://dx. doi.org/10.1016/j.neuroimage.2012.07.060. Location of fMRI signals Functional MRI reliably activated the two major cortical language areas in the dominant hemisphere. To compare the detected positions with the positions available in the literature, the detected positions were transferred to MNI space. The median inferior frontal position (and their MAD) determined after transformation to MNI space was: [X Y Z] =[−50 (3) 15 (11.9) 16 (10.4)] mm for the frontal and [X Y Z]=[−57 (1.5) −39 (8.9) 9 (8.9)] mm for the posterior temporal area. fMRI driven fasciculus arcuatus tractography Using the functionally-defined masks, a significant lateralization of the AF/SLF was found, see Fig. 2 (panel A for an example of a single subject, panels B and C for a group average and panel D for a box-whisker plot). The corresponding number of tracks was 254 (317) for the left and 93 (105) for the right, p = .018. The resulting LI was .49 (.183), strongly left-lateralized. For quality control, MD and FA were measured along the tracks. P.J. Broser et al. / NeuroImage 63 (2012) 1561–1570 1565 Fig. 1. Visualization of the arcuate fasciculus/superior longitudinal fasciculus (AF/SLF) using standard anatomical seed and target masks. Representative single subject (panel A) and group results (panels B and C), showing tracks on the left in red and tracks on the right in green. Panel D shows a boxplot of the number of tracks; the left/right difference was not statistically significant. The corresponding median FA (and their MAD) was 4.63 (.24) for the left and 4.3 (.217) for the right (p = .000305). The corresponding median MD (and their MAD) was 6.35 (.19) for the left and 6.43 (.141) for the right (p = .0256, see Table 1). Tractography from the thalamus to the inferior frontal-language area Tractography between the thalamus and the ipsilateral cortical inferior frontal-language area did not show any lateralization. The measured number of tracks was 460 (185) on the left and 540 (271) on the right (p≥.05), with a median LI of −.0182 (.253). The corresponding median FA (and their MAD) was 4.42 (.267) for the left and 4.46 (.241) for the right (p ≥.05). The corresponding median MD (and their MAD) was 6.47 (.255) for the left and 6.44 (.153) for the right (p ≥.05, see Inline Supplementary Table S‐II). Inline Supplementary Table S2 can be found online at http://dx. doi.org/10.1016/j.neuroimage.2012.07.060. See Fig. 3 panel A top for a 50% isosurface plot of the average fiber density for the left and right sides. See bottom of Fig. 3 panel A for a boxplot of the number of tracks. Tractography from the thalamus to the superior temporal language area Tractography between the thalamus and the superior temporal language area did not show any significant lateralization. The median number of tracks was 198 (125) on the left and 153 (104) on the right(p ≥ .05), with a median LI of .297 (.56). The corresponding median FA (and their MAD) was 4.26 (.175) for the left and 4.41 (.208) for the right (p = .0353). The corresponding median MD (and their MAD) was 7.53 (.379) for the left and 6.97 (.454) for the right (p = .000183, see Inline Supplementary Table S‐III). Inline Supplementary Table S3 can be found online at http://dx. doi.org/10.1016/j.neuroimage.2012.07.060. See top of Fig. 3 panel B for a 50% iso-surface plot of the average fiber density for the left and right sides. See bottom of Fig. 3 panel B for a boxplot of the number of tracks. Tractography from the caudate nucleus to the inferior frontal language area The tracking of the connectivity between the caudate nucleus and the inferior frontal language area did not show any lateralization. The median number of tracks was 1615 (1090) on the left and 2052 (1420) on the right (p ≥ .05), with a median LI of − .128 (.3). The corresponding median FA (and their MAD) was 3.92 (.122) for the left and 3.85 (.142) for the right (p≥ .05). The corresponding median MD (and their MAD) was 6.67 (.184) for the left and 6.83 (.315) for the right (p =.0413, see Inline Supplementary Table S‐IV). Inline Supplementary Table S4 can be found online at http://dx. doi.org/10.1016/j.neuroimage.2012.07.060. See top of Fig. 3 panel C for a 50% isosurface plot of the average fiber density for the left and right sides. See bottom of Fig. 3 panel C for a boxplot of the number of tracks. Tractography from the caudate nucleus to the superior temporal language area The connectivity between the superior temporal language area and the caudate nucleus shows a significant lateralization with a stronger connectivity on the left compared to the right side. The median number of tracks was: median 88 (86) on the left and 43 (34.1) on the right (p = .0041), with a median LI of .375 (.264). 1566 P.J. Broser et al. / NeuroImage 63 (2012) 1561–1570 Fig. 2. Visualization of the arcuate fasciculus/superior longitudinal bundle (AF/SLF) using individual fMRI-defined seed and target masks. Representative single subject (panel A) and group results (panels B and C), showing tracks on the left in red and tracks on the right in green. Panel D shows a boxplot of the number of tracks; the left/right difference was statistically significant. The corresponding median FA (and their MAD) was 4.06 (.336) for the left and 3.93 (.274) for the right (p = 0.00153). The corresponding median MD (and their MAD) was 6.97 (.344) for the left and 7.35 (.333) for the right (p = .0181, see Inline Supplementary Table S‐V). Inline Supplementary Table S5 can be found online at http://dx. doi.org/10.1016/j.neuroimage.2012.07.060. See top of Fig. 3 panel D for a 50% isosurface plot of the average fiber density for the left and right sides. See bottom of Fig. 3 panel D for a boxplot of the number of tracks. On closer inspection, two separate bundles were detected within these connections on the left side only (see Fig. 4A), tentatively termed superior dorsal bundle and inferior frontal bundle. Within each of these two bundles, mean track density was calculated in a sphere of radius=10 mm to do post-hoc left-right comparisons, placed at MNI (−/+29, −33, 17) for the superior and at MNI (−/+30, −4, −13) for the inferior bundle. The median for the mean track density was .57 (.81) on the left and .30 (.23) on the right (p ≤ .041), with a median LI of .47 (.36) for the superior bundle. The median mean track density for the inferior bundle was .63 (.78) on the left and .13 (.12) on the right (p ≤ .00043), with a median LI of .69 (.28). See Fig. 4B. Probability distribution function Probability distribution functions, showing which portion of the thalamus and caudate nucleus are connecting to the two cortical language areas, are shown in Fig. 5. Discussion Tractography was proven to be a valuable tool to non-invasively analyze axonal connectivity in the brain. Several authors have used tractography to analyze the arcuate fasciculus in the past, using anatomical templates to create seed and target masks for the frontal and temporal language areas (Catani et al., 2005; Tiwari et al., 2011). However, this approach is problematic in children due to ongoing functional (Holland et al., 2001; Szaflarski et al., 2006) and structural brain development (Giedd et al., 1999; Wilke and Lidzba, 2007) and higher inter-individual variability (Wilke et al., 2003). In order to address these issues, we here suggest a different approach, using fMRI to localize the language areas on an individual basis to seed and target tractography. The method was applied to a pediatric cohort to analyze the developing language network, including both corticocortical as well as cortico-subcortical connections, with respect to structural lateralization. With regard to the approach to perform tractography, previous studies investigating the language system have used deterministic diffusion tensor imaging (Catani et al., 2005; Tiwari et al., 2011). However, in recent years the limitations of this method were appreciated (Jones, 2010), one limitation being the assumption of only one major diffusion direction per voxel. This becomes a problem especially for tractography of the language system, where most voxels contain more than one fiber population (Behrens et al., 2007). Recently, new tractography techniques based on high angular resolution diffusion imaging (HARDI) have been introduced to overcome this problem (e.g. Tournier et al., 2011). Therefore, HARDI-based probabilistic tractography has been adopted for the present study in order to investigate more sensitively the structural connectivity of the language system. The data acquired here was first analyzed using the well-established method using anatomical masks to track the arcuate fasciculus. We found that this established method, applied to our pediatric cohort, demonstrated a similar degree of lateralization of the arcuate fasciculus as previously shown in adult (Catani et al., 2005); however, this 1567 P.J. Broser et al. / NeuroImage 63 (2012) 1561–1570 Table 1 Demographic information and number of tracks as well as mean FA and MD (±SD) along the functionally-defined AF/SLF (cf. Fig. 2). Demographics Right Subject no. Gender Age Month Handedness Number of tracks 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Median MAD Male Female Female Male Female Female Female Female Male Male Male Male Female Female Male 95 108 109 112 112 114 127 139 144 147 180 187 189 193 213 139 44.5 R R L R R R R R R R R R R R R 76 212 110 348 72 466 278 93 22 1049 68 66 26 312 6 93 105 Left FA (10−1) MD (10−4 mm2 s−1) Mean SD Mean SD 4.21 4.3 4.37 4.51 4.23 4.41 4.07 4.46 4.07 4.45 4.17 4.51 4.04 4.44 3.95 4.3 0.217 1.33 1.23 1.37 1.3 1.43 1.27 1.43 1.46 1.3 1.32 1.23 1.3 1.28 1.37 1.19 6.71 6.59 6.53 6.09 6.8 6.43 6.45 6.34 6.71 6.4 6.43 6.64 6.35 5.96 6.35 6.43 0.141 0.885 0.8 0.816 0.776 0.811 0.685 0.716 0.928 0.942 0.765 0.799 1.05 0.75 0.767 0.817 lateralization was not significantly different from zero (Fig. 1). It must of course be acknowledged that this lack of statistical significance may well be a sample size effect. However, in contrast to this, a strong lateralization of the arcuate fasciculus was found when using the individually-derived, functional MRI-based locations as seed and target areas (Fig. 2). Of course, when using functional localizers (Saxe et al., 2006), care must be taken to use fMRI tasks well-established and validated in children, as done here (Ebner et al., 2011; Lidzba et al., 2011; Wilke et al., 2005, 2006). Additionally, it avoids the possible bias that is introduced when applying masks derived from adult brains to the pediatric brain, which is different in shape, size, and tissue composition (Giedd et al., 1999; Wilke et al., 2002, 2003). Our results indicate that a lateralization of the arcuate fasciculus already exists in children, Number of tracks 254 708 220 587 210 1164 1516 334 61 793 199 476 40 129 25 254 317 LI FA (10−1) MD (10−4 mm2 s−1) Mean SD Mean SD 4.65 4.63 4.29 4.67 4.12 4.8 4.54 5.07 4.25 4.97 4.55 4.8 4.48 4.91 4.37 4.63 0.248 1.44 1.31 1.37 1.52 1.39 1.16 1.33 1.35 1.34 1.31 1.35 1.31 1.44 1.28 1.43 6.57 6.4 6.64 6.12 6.72 6.28 6.42 6.03 6.83 6.25 6.35 6.34 6.22 5.98 6.35 6.35 0.19 0.728 0.673 0.686 0.881 0.701 0.602 0.631 0.832 0.821 0.692 0.744 0.659 0.751 0.779 0.875 0.54 0.54 0.33 0.26 0.49 0.43 0.69 0.56 0.47 −0.14 0.49 0.76 0.21 −0.41 0.61 0.49 0.18 which is in line with previous studies using the FA as relevant marker (Brauer et al., 2011; Dubois et al., 2009; Tiwari et al., 2011). It should be noted, however, that the definition of laterality here was based on the number of tracks on each side, while other studies used different parameters to assess lateralization (Dubois et al., 2009), making direct comparisons difficult. However, when using fractional anisotropy (FA) or mean diffusivity (MD) to assess lateralization for the structural or functional arcuatus tracking, these differences were also significant (see Results section). When assessing the connectivity of the thalamus with superior temporal language areas, results are significant when using FA/MD but not when using the number of tracks. While, these results suggest that our approach to using the number of tracks, if anything, is more conservative than using FA and/or MD, further studies Fig. 3. Cortico-subcortical language networks: visualization of group results (top panels) and corresponding boxplots (bottom panels). Shown are connections of the thalamus (Thal.)/caudate nucleus (Caud.) with inferior frontal (IFR, panels A, C) and superior temporal (STR, panels B, D) language areas. Median locations of the inferior frontal and superior temporal language areas are visualized as small spheres on the left side. Significant laterality was only found for the caudate/superior temporal language connections (p-value: 0.00412, cf. Fig. 4). 1568 P.J. Broser et al. / NeuroImage 63 (2012) 1561–1570 Fig. 4. Detailed view of the caudate nucleus/superior temporal language connections, suggesting the existence of two separate pathways connecting the two structures on the left side. Two spherical regions of interest were selected (visualized as wireframe spheres) and the track density (number of tracks per cmm) on the left and right sides were compared for the inferior frontal (panel B, left) and the superior dorsal pathway (panel B, right). will have to show which metric is ultimately best-suited to describe the connectivity between the two brain regions. When assessing the connectivity of the major language regions with the thalamus, a strong but not lateralized connectivity was found. In particular, the connectivity between the inferior frontal cortical area and the thalamus was almost perfectly symmetric. The connectivity between the superior temporal area and the thalamus shows a trend towards a positive lateralization index but did not reach statistical significance. Again, this might be due to the group size, but given that both anatomical (Guillery, 1995) as well as neuroimaging studies (Broser et al., 2011) indicate that the thalamus projects into nearly every part of the cortex, there may well be no lateralized effects in absolute connectivity between the left and right language areas. Probability distribution functions (PDF) were calculated in order to analyze which substructure of the thalamus is projecting to or is receiving input from the two cortical language areas. The PDF of the connection from the thalamus to the inferior frontal cortex has its highest density in the ventral anterior part of the thalamus, mainly containing the ventral anterior (VA) nucleus, which was shown before to be connected with inferior frontal brain regions in monkeys (Xiao et al., 2009). It is interesting to speculate that this connectivity may reflect a contribution of the thalamus to a motor output loop, based on anatomical studies implicating this Fig. 5. Probability distribution functions: shown are connections of the thalamus (Thal.)/caudate nucleus (Caud.) with inferior frontal (IFR, panels A, C) and superior temporal (STR, panels B, D) language areas (cf. Fig. 3). Each panel contains a MNI coordinate system. The panels on the top show a view from above while the panels on the bottom show a tilted view from the left side. P.J. Broser et al. / NeuroImage 63 (2012) 1561–1570 brain regions in motor control; however, further research is necessary to confirm and extend these findings. In contrast to this, the mediodorsal nucleus (commonly considered the principal nucleus of the frontal cortex; Herrero et al., 2002) shows much lower connectivity. The PDF of the thalamo‐superior temporal cortical region showed a high density overlapping with the area of the Pulvinar nuclei (Mai, 2003) of the thalamus, as well as the medial genicular nucleus. Especially the medial genicular nucleus is receiving primary auditory input (Herrero et al., 2002), in concordance with the role of the superior temporal cortex in receptive language functions. When assessing the connectivity of the major language regions with the caudate nucleus, we were surprised to find a strong, but symmetrical connectivity between the caudate nucleus and the inferior frontal language area. The PDF showed a projection mainly to the head and anterior part of the body. Based on the classically-assumed role of the caudate in motor control and coordination (White, 2009), we had expected to find a more lateralized projection pattern of the productive language area into this structure. Given the strong connections, we also do not believe that a larger group would show a strikingly different pattern. In contrast to this, the connection between the caudate nucleus and the superior temporal cortical language showed a strong and robust asymmetry. On the left side, there was typically more than twice the number of tracks than on the right. The visualization revealed two bundles connecting the head of the caudate nucleus and the superior temporal language area on the left side, a superior dorsal and an inferior frontal bundle. Both bundles showed a strong lateralization to the left side, which was more pronounced in the inferior bundle (see Fig. 4). While surprising at first this finding may be, other modalities have provided evidence for an anatomical as well as functional interplay between these regions. For example, a strong connection between the inferior parietal and superior temporal cortex to the caudate nucleus has been described in monkeys (Yeterian and Pandya, 1993). Further, functional studies implicate the left head of the caudate nucleus in specific language task (Crinion et al., 2006; Wilke et al., 2009). Moreover, the recently-suggested role of this structure in working memory (White, 2009) may also explain this strong interconnectivity, but further research is necessary to investigate the basis of this interplay. In this context, the almost-absent inferior frontal bundle on the right side in these healthy children is interesting, as it could reflect the previously-described higher reliance of children on this structure (Brauer et al., 2011); alternatively, this pathway could still be lateralized in adults, allowing to shed further light on the structural underpinnings of functional lateralization processes in the human brain. Regarding the implications of our results, it must be noted that the assessment of the lateralization of language functions is one of the most-used applications of fMRI in the clinical context. A multitude of tasks has been designed and applied to this effect, in the hope of avoiding invasive procedures such as the Wada test (Dym et al., 2011). Given the fact that fMRI is notoriously difficult in children (Byars et al., 2002), the mere idea of finding a structural correlate of functional language asymmetry is exciting. We refrained from explicitly comparing structural and functional lateralization in this study as all children were leftdominant in both tasks as measured by a weighted mean LI (Wilke and Schmithorst, 2006; data not shown), making correlations meaningless in this small sample. Further, if more appropriate anatomical masks could be derived, anatomically-informed tractography would even allow assessing language dominance in the absence of cooperation and even in completely non-cooperative children as part of a routine MR examination under sedation. However, for the time being, further research is necessary, comparing functional and structural lateralization in the same children and adults and using both functionally and anatomically informed approaches. It should be noted that generally, diffusion MRI and tractography are not capable of inferring anatomical connectivity at the axonal 1569 level (Lawes et al., 2008). The technique is limited by the intrinsic spatial resolution of the acquired images (2.5 mm 3). This will be more of an issue when it comes to tracking in anatomically highly complex brain regions, such as the central brain structures. However, using a current tracking algorithm shown to be robust to the issues of crossing fibers (Tournier et al., 2008), we believe our results to be valid. A further limitation of the developed method is that the algorithms used to extract the subcortical brain structures were developed on an adult population and so it is possible that features of the pediatric brain may not be respected by these algorithms (i.e. the relation of the size and diameter of the caudate nucleus might be different). Therefore and despite the developer's good experience in applying these tools in children, a bias when using adult prior information cannot wholly be excluded. Consequently, an important future development would be to develop and apply methods (Igual et al., 2011; Wilke et al., 2008) to specifically analyze the pediatric brain. Further the cortical templates used here for anatomical tracing and the search masked used for localizing the seed and target areas for functional driven tractography were also derived from an adult template. So a future project might include the creation of pediatric masks. However, while anatomical localizers have been used in the past and have the merit to be easily and reproducibly applicable to larger groups and across studies, there is an inherent danger to over-interpret the structure–function correspondence (Pouratian and Bookheimer, 2010) even in adults, which is evident with more and more cytoarchitectonic reference data becoming available (Eickhoff et al., 2007). For children, with the brain undergoing substantial maturation both on the structural (Giedd et al., 1999; Wilke et al., 2002, 2003) as well as the functional (Johnson, 2001) level, we believe it is preferable to use functional localizers whenever feasible on theoretical grounds alone, and we suggest our results support this conclusion. Our technique of functionally-driven tractography relies further on the accurate reflection of the localized language area on the left side onto the right. A possible further improvement would be to develop an fMRI localizer for the right/non‐dominant‐hemisphere. In summary, the method developed here is a suitable technique to analyze the structural underpinnings of language networks in children. Using individual functional localizers, the potential bias of using inappropriate masks can be avoided. Using this method, we could show a significant lateralization of the arcuate fasciculus which was not demonstrated using the anatomically-informed approach. Despite what is known about the lateralization of the AF/SLF in adults (Catani et al., 2005) and children (Tiwari et al., 2011), it should be noted here that this cannot be claimed to reflect greater sensitivity of one approach, as no ground truth is available and the difference could be due to false positives in one or false negatives in the other approach. Connections of the major language regions with the thalamus were found to be symmetrical. Further, a strong but perfectly symmetrical connectivity between the inferior frontal brain regions and the caudate nucleus was found, while a strongly lateralized connectivity of the superior temporal brain region with the caudate nucleus was demonstrated for the first time. Acknowledgments This research has been supported by the PATE program of the University of Tübingen (2028-0-0), the DFG (WI3630/1-1) and the Margarete von Wrangell program of the state of Baden-Württemberg. The software used in this paper as generated by the author is available upon request to Dr. Broser. References Andersson, J.L., Hutton, C., Ashburner, J., Turner, R., Friston, K., 2001. Modeling geometric deformations in EPI time series. NeuroImage 13, 903–919. Basser, P.J., Mattiello, J., LeBihan, D., 1996. MR diffusion tensor spectroscopy and imaging. Biophys. J. 66, 256–267. 1570 P.J. Broser et al. / NeuroImage 63 (2012) 1561–1570 Behrens, T.E., Johansen-Berg, H., Woolrich, M.W., Smith, S.M., Wheeler-Kingshott, C.A., Boulby, P.A., Barker, G.J., Sillery, E.L., Sheehan, K., Ciccarelli, O., Thompson, A.J., Brady, J.M., Matthews, P.M., 2003. Non-invasive mapping of connections between human thalamus and cortex using diffusion imaging. Nat. Neurosci. 6, 750–757. Behrens, T.E., Johansen-Berg, H., Jbabdi, S., Rushworth, M.F., Woolrich, M.W., 2007. Probabilistic diffusion tractography with multiple fibre orientations: what can we gain? NeuroImage 34, 144–155. Benedict, H., 1979. Early lexical development: comprehension and production. J. Child Lang. 6, 183–200. Bornkessel, I., Zysset, S., Friederici, A.D., von Cramon, D.Y., Schlesewsky, M., 2005. Who did what to whom? The neural basis of argument hierarchies during language comprehension. NeuroImage 26, 221–233. Brauer, J., Anwander, A., Friederici, A.D., 2011. Neuroanatomical prerequisites for language functions in the maturing brain. Cereb. Cortex 21, 459–466. Broser, P., Vargha-Khadem, F., Clark, C.A., 2011. Robust subdivision of the thalamus in children based on probability distribution functions calculated from probabilistic tractography. NeuroImage 57, 403–415. Burdach, C.F., 1822. Baue und Leben des Gehirns. Leipzig in der Dyk'schen Buchhandlung. Byars, A.W., Holland, S.K., Strawsburg, R.H., Bommer, W., Dunn, R.S., Schmithorst, V.J., Plante, E., 2002. Practical aspects of conducting large-scale functional magnetic resonance imaging studies in children. J. Child Neurol. 17, 885–890. Castellanos, F.X., Lee, P.P., Sharp, W., Jeffries, N.O., Greenstein, D.K., Clasen, L.S., Blumenthal, J.D., James, R.S., Ebens, C.L., Walter, J.M., Zijdenbos, A., Evans, A.C., Giedd, J.N., Rapoport, J.L., 2002. Developmental trajectories of brain volume abnormalities in children and adolescents with attention-deficit/hyperactivity disorder. JAMA 288, 1740–1748. Catani, M., Jones, D.K., Ffytche, D.H., 2005. Perisylvian language networks of the human brain. Ann. Neurol. 57, 8–16. Crinion, J., Turner, R., Grogan, A., Hanakawa, T., Noppeney, U., Devlin, J.T., Aso, T., Urayama, S., Fukuyama, H., Stockton, K., Usui, K., Green, D.W., Price, C.J., 2006. Language control in the bilingual brain. Science 312, 1537–1540. Dubois, J., Hertz-Pannier, L., Cachia, A., Mangin, J.F., Le Bihan, D., Dehaene-Lambertz, G., 2009. Structural asymmetries in the infant language and sensori-motor networks. Cereb. Cortex 19, 414–423. Dym, R.J., Burns, J., Freeman, K., Lipton, M.L., 2011. Is functional MR imaging assessment of hemispheric language dominance as good as the Wada test?: a meta-analysis. Radiology 261, 446–455. Ebner, K., Lidzba, K., Hauser, T.K., Wilke, M., 2011. Assessing language and visuospatial functions with one task: a “dual use” approach to performing fMRI in children. NeuroImage 58, 923–929. Eickhoff, S.B., Paus, T., Caspers, S., Grosbras, M.H., Evans, A.C., Zilles, K., Amunts, K., 2007. Assignment of functional activations to probabilistic cytoarchitectonic areas revisited. NeuroImage 36, 511–521. Frey, S., Campbell, J.S.W., Pike, G.B., Petrides, M., 2008. Dissociating the human language pathways with high angular resolution diffusion fiber tractography. J. Neurosci. 28, 11435–11444. Friston, K.J., Holmes, A.P., Worsley, K.J., Poline, J.B., Frith, C., Frackowiak, R.S.J., 1995. Statistical parametric maps in functional imaging: a general linear approach. Hum. Brain Mapp. 2, 189–210. Giedd, J.N., Blumenthal, J., Jeffries, N.O., Castellanos, F.X., Liu, H., Zijdenbos, A., Paus, T., Evans, A.C., Rapoport, J.L., 1999. Brain development during childhood and adolescence: a longitudinal MRI study. Nat. Neurosci. 2, 861–863. Groeschel, S., Vollmer, B., King, M.D., Connelly, A., 2010. Developmental changes in cerebral grey and white matter volume from infancy to adulthood. Int. J. Dev. Neurosci. 28, 481–489. Guillery, R.W., 1995. Anatomical evidence concerning the role of the thalamus in corticocortical communication: a brief review. J. Anat. 187, 583–592. Herrero, M.T., Barcia, C., Navarro, J.M., 2002. Functional anatomy of thalamus and basal ganglia. Child Nerv. Syst. 18, 386–404. Hertz-Pannier, L., Chiron, C., Jambaqu, I., Renaux-Kieffer, V., Van de Moortele, P.F., Delalande, O., Fohlen, M., Brunelle, F., Le Bihan, D., 2002. Late plasticity for language in a child's non-dominant hemisphere: a pre- and post-surgery fMRI study. Brain 125, 361–372. Holland, S.K., Plante, E., Byars, A.W., Strawsburg, R.H., Schmithorst, V.J., Ball, W.S., 2001. Normal fMRI brain activation patterns in children performing a verb generation task. NeuroImage 14, 837–843. Igual, L., Soliva, J.C., Hernandez-Vela, A., Escalera, S., Jimanez, X., Vilarroya, O., Radeva, P., 2011. A fully-automatic caudate nucleus segmentation of brain MRI: application in volumetric analysis of pediatric attention-deficit/hyperactivity disorder. Biomed. Eng. Online 5 (10), 105. Jenkinson, M., Smith, S., 2001. A global optimisation method for robust affine registration of brain images. Med. Image Anal. 5, 143–156. Johnson, M.H., 2001. Functional brain development in humans. Nat. Rev. Neurosci. 2, 475–483. Jones, D.K., 2010. Precision and accuracy in diffusion tensor magnetic resonance imaging. Top. Magn. Reson. Imaging 21, 87–99. Karunanayaka, P.R., Holland, S.K., Schmithorst, V.J., Solodkin, A., Chen, E.E., Szaflarski, J.P., Plante, E., 2007. Age-related connectivity changes in fMRI data from children listening to stories. NeuroImage 34, 349–360. Lawes, I.N., Barrick, T.R., Murugam, V., Spierings, N., Evans, D.R., Song, M., Clark, C.A., 2008. Atlas-based segmentation of white matter tracts of the human brain using diffusion tensor tractography and comparison with classical dissection. NeuroImage 39, 62–79. Lidzba, K., Schwilling, E., Grodd, W., Krägeloh-Mann, I., Wilke, M., 2011. Language comprehension vs. language production: age effects on fMRI activation. Brain Lang. 119, 6–15. Macey, P.M., Macey, K.E., Kumar, R., Harper, R.M., 2004. A method for removal of global effects from fMRI time series. Neuroimage 22 (1), 360–366 (May). Mai, J.K., 2003. Atlas of the Human Brain. Elsevier. Makris, N., Kennedy, D.N., McInerney, S., Sorensen, A.G., Wang, R., Caviness, V.S., Pandya, D.N., 2005. Segmentation of subcomponents within the superior longitudinal fascicle in humans: a quantitative, in vivo, DT-MRI study. Cereb. Cortex 15, 854–869. Marslen-Wilson, W.D., Tyler, L.K., 2007. Morphology, language and the brain: the decompositional substrate for language comprehension. Philos. Trans. R. Soc. Lond. B Biol. Sci. 362, 823–836. Mori, S., Zhang, J., 2006. Principles of diffusion tensor imaging and its applications to basic neuroscience research. Neuron 51, 527–539. Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 9, 97–113. Patenaude, B., 2007. Bayesian Statistical Models of Shape and Appearance for Subcortical Brain Segmentation. D.Phil. Thesis. University of Oxford. Petermann, F., Petermann, U. (Eds.), 2007. Hamburg-Wechsler Intelligenztest für Kinder — IV (HAWIK-IV). Huber, Bern. Pouratian, N., Bookheimer, S.Y., 2010. The reliability of neuroanatomy as a predictor of eloquence: a review. Neurosurg. Focus. 28, E3. Price, C.J., 2010. The anatomy of language: a review of 100 fMRI studies published in 2009. Ann. N. Y. Acad. Sci. 1191, 62–88. R Development Core Team, 2010. R: a language and environment for statistical computing, R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org. Riecker, A., Ackermann, H., Wildgruber, D., Meyer, J., Dogil, G., Haider, H., Grodd, W., 2000. Articulatory/phonetic sequencing at the level of the anterior perisylvian cortex: a functional magnetic resonance imaging (fMRI) study. Brain Lang. 75 (2), 259–276 (Nov). Saxe, R., Brett, M., Kanwisher, N., 2006. Divide and conquer: a defense of functional localizers. NeuroImage 30, 1088–1096. Smith, S.M., 2002. Fast robust automated brain extraction. Hum. Brain Mapp. 17, 143–155. Smith, S.M., Jenkinson, M., Woolrich, M.W., Beckmann, C.F., Behrens, T.E., JohansenBerg, H., Bannister, P.R., De Luca, M., Drobnjak, I., Flitney, D.E., Niazy, R.K., Saunders, J., Vickers, J., Zhang, Y., De Stefano, N., Brady, J.M., Matthews, P.M., 2004. Advances in functional and structural MR image analysis and implementation as FSL. NeuroImage 23, 208–219. Szaflarski, J.P., Holland, S.K., Schmithorst, V.J., Byars, A.W., 2006. fMRI study of language lateralization in children and adults. Hum. Brain Mapp. 27, 202–212. Tiwari, V.N., Jeong, J.W., Asano, E., Rothermel, R., Juhasz, C., Chugani, H.T., 2011. A sensitive diffusion tensor imaging quantification method to detect language laterality in children: correlation with the Wada test. J. Child Neurol. 26, 1516–1521. Tournier, J.D., Calamante, F., Gadian, D.G., Connelly, A., 2004. Direct estimation of the fiber orientation density function from diffusion-weighted MRI data using spherical deconvolution. NeuroImage 23, 1176–1185. Tournier, J.D., Calamante, F., Connelly, A., 2007. Robust determination of the fibre orientation distribution in diffusion MRI: non-negativity constrained super-resolved spherical deconvolution. NeuroImage 35, 1459–1472. Tournier, J.D., Yeh, C.H., Calamante, F., Cho, K.H., Connelly, A., Lin, C.P., 2008. Resolving crossing fibres using constrained spherical deconvolution: validation using diffusion-weighted imaging phantom data. NeuroImage 42, 617–625. Tournier, J.D., Mori, S., Leemans, A., 2011. Diffusion tensor imaging and beyond. Magn. Reson. Med. 65, 1532–1556. Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard, O., Delcroix, N., Mazoyer, B., Joliot, M., 2002. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. NeuroImage 15, 273–289. White, N.M., 2009. Some highlights of research on the effects of caudate nucleus lesions over the past 200 years. Behav. Brain Res. 199, 3–23. Wilke, M., Lidzba, K., 2007. LI-tool: a new toolbox to assess lateralization in functional MR-data. J. Neurosci. Methods 163, 128–136. Wilke, M., Schmithorst, V.J., 2006. A combined bootstrap/histogram analysis approach for computing a lateralization index from neuroimaging data. NeuroImage 33, 522–530. Wilke, M., Schmithorst, V.J., Holland, S.K., 2002. Assessment of spatial normalization of whole-brain magnetic resonance images in children. Hum. Brain Mapp. 17, 48–60. Wilke, M., Holland, S.K., Myseros, J.S., Schmithorst, V.J., Ball, W.S., 2003. Functional magnetic resonance imaging in pediatrics. Neuropediatrics 34, 225–233. Wilke, M., Lidzba, K., Staudt, M., Buchenau, K., Grodd, W., Krägeloh-Mann, I., 2005. Comprehensive language mapping in children, using functional magnetic resonance imaging: what's missing, counts. Neuroreport 16, 915–919. Wilke, M., Lidzba, K., Staudt, M., Buchenau, K., Grodd, W., Krägeloh-Mann, I., 2006. An fMRI task battery for assessing hemispheric language dominance in children. NeuroImage 32, 400–410. Wilke, M., Krägeloh-Mann, I., Holland, S.K., 2007. Global and local development of gray and white matter volume in normal children and adolescents. Exp. Brain Res. 178 (3), 296–307 (Apr) (Epub 2006 Oct 19). Wilke, M., Holland, S.K., Altaye, M., Gaser, C., 2008. Template-O-Matic: a toolbox for creating customized pediatric templates. NeuroImage 41, 903–913. Wilke, M., Lidzba, K., Krägeloh-Mann, I., 2009. Combined functional and causal connectivity analyses of language networks in children: a feasibility study. Brain Lang. 108, 22–29. Xiao, D., Zikopoulos, B., Barbas, H., 2009. Laminar and modular organization of prefrontal projections to multiple thalamic nuclei. Neuroscience 161, 1067–1081. Yeterian, E.H., Pandya, D.N., 1993. Striatal connections of the parietal association cortices in rhesus monkeys. J. Comp. Neurol. 332, 175–197.