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.