A pilot study of basal ganglia and thalamus structure by high dimensional mapping in children with Tourette syndrome

Background: Prior brain imaging and autopsy studies have suggested that structural abnormalities of the basal ganglia (BG) nuclei may be present in Tourette Syndrome (TS). These studies have focused mainly on the volume differences of the BG structures and not their anatomical shapes. Shape differences of various brain structures have been demonstrated in other neuropsychiatric disorders using large-deformation, high dimensional brain mapping (HDBM-LD). A previous study of a small sample of adult TS patients demonstrated the validity of the method, but did not find significant differences compared to controls. Since TS usually begins in childhood and adult studies may show structure differences due to adaptations, we hypothesized that differences in BG and thalamus structure geometry and volume due to etiological changes in TS might be better characterized in children. Objective: Pilot the HDBM-LD method in children and estimate effect sizes. Methods: In this pilot study, T1-weighted MRIs were collected in 13 children with TS and 16 healthy, tic-free, control children. The groups were well matched for age. The primary outcome measures were the first 10 eigenvectors which are derived using HDBM-LD methods and represent the majority of the geometric shape of each structure, and the volumes of each structure adjusted for whole brain volume. We also compared hemispheric right/left asymmetry and estimated effect sizes for both volume and shape differences between groups. Results: We found no statistically significant differences between the TS subjects and controls in volume, shape, or right/left asymmetry. Effect sizes were greater for shape analysis than for volume. Conclusion: This study represents one of the first efforts to study the shape as opposed to the volume of the BG in TS, but power was limited by sample size. Shape analysis by the HDBM-LD method may prove more sensitive to group differences.


Abstract
Prior brain imaging and autopsy studies have suggested that : Background structural abnormalities of the basal ganglia (BG) nuclei may be present in Tourette Syndrome (TS). These studies have focused mainly on the volume differences of the BG structures and not their anatomical shapes. Shape differences of various brain structures have been demonstrated in other neuropsychiatric disorders using large-deformation, high dimensional brain mapping (HDBM-LD). A previous study of a small sample of adult TS patients demonstrated the validity of the method, but did not find significant differences compared to controls. Since TS usually begins in childhood and adult studies may show structure differences due to adaptations, we hypothesized that differences in BG and thalamus structure geometry and volume due to etiological changes in TS might be better characterized in children.
Pilot the HDBM-LD method in children and estimate effect sizes. :

Objective
In this pilot study, T1-weighted MRIs were collected in 13 children : Methods with TS and 16 healthy, tic-free, control children. The groups were well matched for age. The primary outcome measures were the first 10 eigenvectors which are derived using HDBM-LD methods and represent the majority of the geometric shape of each structure, and the volumes of each structure adjusted for whole brain volume. We also compared hemispheric right/left asymmetry and estimated effect sizes for both volume and shape differences between groups.
We found no statistically significant differences between the TS : Results subjects and controls in volume, shape, or right/left asymmetry. Effect sizes

Introduction
Tourette syndrome (TS) is a chronic idiopathic syndrome characterized by the appearance of both vocal and motor tics during childhood or adolescence 1,2 . Tics are repetitive, stereotyped, suppressible movements or vocalizations that may include blinking, abdominal tensing, sniffing, or throat clearing 3 . TS affects approximately 0.5% of school-age children, but its causes and pathophysiology are not yet well understood 4 .
It has been suggested that problems with activity modulation in the basal ganglia and thalamus may contribute to the inability of TS patients to exercise behavioral inhibition 5,6 as a result of these structures' effects on behavioral inhibition via the prefrontal, parietal, temporal, and cingulate cortices 7 . The basal ganglia and thalamus modulate cortical activity through cortico-basal ganglia-thalamo-cortical loops, composed of connections from the frontal cortex to the striatum, the striatum to the globus pallidus, substantia nigra, and thalamus, and the thalamus back to the cortex 8 .
Several lines of evidence support the presence of structural abnormalities in basal ganglia nuclei in individuals with TS 4 . Autopsy studies have found abnormalities within the basal ganglia, including increased number of neurons in the globus pallidus interna, decreased density and number of neurons in the globus pallidus pars externa, and decreased parvalbumin and choline acetyltransferase staining cholinergic interneurons in the caudate nucleus and putamen 9,10 . However, since TS is rarely a fatal disease, the number of autopsied cases is limited 11 . Case studies of focal brain lesions have demonstrated new tic onset after lesions to the prefrontal cortex, thalamus, and basal ganglia 12 . In addition, encephalitis lethargica, frontal lobe degeneration, Huntington disease, Wilson disease, and other degenerative illnesses are associated with tics 12 . Further, some TS patients have benefitted from deep brain stimulation of the globus pallidus and thalamus in TS 13-16 . Collectively, these observations suggest a role for the basal ganglia, thalamus, and frontal cortex in tics.
Neuroimaging studies can be especially beneficial for studying structural abnormalities because they allow longitudinal study design, reduced investigator and sampling bias, and are relatively non-invasive. A number of MRI studies have examined anatomical volumes and cortical thickness in children and adults with TS and reported significant differences in various brain regions, including the caudate, sensorimotor and prefrontal cortex, and corpus callosum 17 . Most consistently, basal ganglia volumes were found to be smaller in TS subjects compared with healthy controls, but neuroanatomical shape differences and asymmetry abnormalities have not yet been consistently described 18-24 .
Large-deformation high dimensional brain mapping (HDBM-LD) is a computational anatomy tool that reduces the potential for human error in image analysis by further automating elements of image analysis. It has been successfully employed in characterizing shape and volume abnormalities of the hippocampus in neuropsychiatric disorders such as schizophrenia 25-27 , dementia of the Alzheimer type 28-31 , depression 32 and epilepsy 33 . It has also been applied to examine the thalamus in schizophrenia 34 .
HDBM-LD was applied to assess volume and shape differences in putamen, caudate nucleus, nucleus accumbens, globus pallidus, and thalamus in 15 adults with TS and 15 matched controls. No differences in volume or shape were found 35 . However, TS begins before adulthood. Several structural imaging studies in TS have found an interaction between regional brain volumes and age 21,22 . It has been suggested that differences seen in adult studies may reflect adaptations or selection bias rather than changes etiologically relevant to TS 20 . Thus the present study applied HDBM-LD to investigate the volume and shape of these structures in children. We hypothesized that we would find reduced volume, abnormal shape, or abnormal right-to-left asymmetry in one or more of these structures, compared to age-matched controls. Given that there were no prior studies using the HDBM-LD method to analyze brain structures of children with TS in the literature, another goal of this pilot study was to estimate the effect size of these measures in this population.

Materials and methods
Ethics statement A parent of each subject gave written informed consent to participate in the study, and each subject assented to participation. The study was approved by the Washington University Human Studies Committee (approval # 03-1282).

Participants
This study included 13 children with TS (mean age (SD) = 12.44 (2.22), 3 female, 12 right-handed) and 16 healthy controls (mean age (SD) = 12.39 (1.92), 2 female, 15 right-handed). A movement disorders-trained physician examined all TS subjects and 10 of the control subjects. The remaining control subjects underwent neuropsychological evaluation as described previously 36 . Exclusion criteria were: inability to give informed consent, contraindication to MRI, currently symptomatic major depression, or lifetime history of mental retardation, autism, psychosis, mania, anorexia, bulimia, or drug abuse. All TS subjects met DSM-IV-TR criteria either for Tourette's Disorder or Chronic Tic Disorder. Disease duration and severity and other clinical characteristics are summarized in Table 1.
Image acquisition and preprocessing A 1.5 T Siemens Vision system with a standard head receiver coil was used to collect T1-weighted MR structural images. Prior to scanning sessions, the transmitter was tuned and the main field was shimmed. Anatomic images used a 3D T1-weighted sequences (MPRAGE, 1x1x1.25 mm 3 voxels) 37 . Individual MPRAGE collections lasted approximately 6.5 minutes.
Initial image processing was done as described previously 35,38 . Using Analyze TM software (Rochester, Minnesota), images were linearly rescaled so that voxels with intensity two standard deviations above the mean in the corpus callosum were mapped to 255, and voxels with intensity levels two standard deviations below the mean in the lateral ventricles were mapped to 0.
Large-Deformation High-Dimensional Brain Mapping (HDBM-LD) HDBM-LD was used to determine the volume and shape of the brain structures of interest in all subject scans, as described in detail elsewhere 35 . Briefly, on each subject's brain image, a single rater (MEMcN) marked 27 points on the boundaries of the basal ganglia and thalamus in each hemisphere, which were used as an initial step to roughly align the brain image to a labeled standard brain image (template). From this starting point a differentiable, invertible transformation was computed that mapped all voxels of the subject's image to the template. Using this transformation, the labels on the template image are automatically assigned to the corresponding voxels of each subject's image. The authors checked the segmentation of each subject's MR image by visual inspection. This method is extremely reliable and has been validated against expert manual tracings 35 .
Brain structure volume and shape analysis All brain structure volume and shape analysis methods were conducted as described previously 35 . We examined five structures: caudate nucleus, nucleus accumbens, globus pallidus, putamen, and thalamus. Volume for each structure was analyzed using a repeated measures ANCOVA, with diagnostic group as the between-subjects factor, brain hemisphere as the within-subjects factor, and age and whole brain volume as covariates. The degree of volumetric asymmetry was examined with the hemisphere effect, and group differences in volumetric asymmetry were assessed by examining the group-by-hemisphere interactions. We also analyzed the total (left and right hemisphere) structure volumes using an ANCOVA. The volume ANCOVAs were repeated with other covariates and factors, including estimated total intracranial volume, sex and handedness, none of which substantively changed the results.
Brain structure shapes were determined from the inter-subject deformation vector fields provided by the HDBM-LD transformations. Eigenvalues and a complete orthonormal set of eigenvectors representing shape variation were obtained using singular value decomposition (SVD) of the pooled covariance in the population studied. The coefficients (eigenscores) associated with the eigenvalues and eigenvectors were calculated for each subject and for each structure in each hemisphere 35,40 . We used the eigenscores based on the first ten eigenvectors for each structure in each hemisphere in a multivariate ANCOVA to test for group differences in  shape. These first ten eigenscores explained 81-92% of the total variance for each structure.
Data file (subject characteristics, volume and shape)

Volume
Repeated-measures ANCOVAs showed no significant group effect for any structure. Structural volumes and ANCOVA statistics are shown in Table 2. Additionally, no significant hemisphere effects or group by hemisphere interactions were seen for any of the five structures examined (see Table 2).

Shape
MANCOVAs (using the first ten eigenscores as dependent variables) for each structure in each hemisphere showed no significant group effect (see Table 3). Effect sizes (Cohen's ƒ 2 ) for both volume and shape are provided in Table 4; the effect sizes for the shape comparisons were larger than those for the volume comparisons.

Discussion
Using HDBM-LD, a validated method for automatic, highdimensional mapping of basal ganglia and thalamic structures, we found no significant differences in basal ganglia volumes or shape between children with TS and matched control children. For most basal ganglia regions, this reflects the conclusions of a recent review 17 . For instance, two groups found increased putamen volume in TS 41,42 , but a larger study found decreased volume 43 . However, the majority of these studies found no abnormality in putamen, similar to the current study. Three other studies, including the HDBM-LD  The present study and the HDBM-LD study in adults represent some of the first efforts to study the shape (as opposed to the volume) of basal ganglia nuclei in TS, and provide effect size estimates for planning a study with larger samples.

Author contributions
LW and KJB designed the study. JAC, BLS, JMH and KJB collected the data. MEMcN, ACW, DJG, SLW and LW analyzed the data. MEMcN, ACW and KJB wrote the first draft of the manuscript. All authors were involved in revising the manuscript and agreed to the final content.

Competing interests
No competing interests were disclosed.

Grant information
This study was funded by a research grant from the Tourette Syndrome Association to LW (Morphological abnormalities of the thalamus and basal ganglia in Tourette syndrome by computational anatomy). Manuscript preparation was supported in part by the National Institutes of Health grant K24 MH087913 to KJB.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.