f t

c,,Timothy J. DeVito , Kiralee M. Hayashia, Jennifer A. Geaga , Dick J. Drost ,

Peter C. Williamsonb,c, Nagalingam Rajakumarb, Arthur W. Togaa, Paul M. Thompsona

RM, Clermont-Ferrand, France

the thickness of the lateral ventricles between patients and controls. Although ventricular volumes measured using traditionalmethods did not differ significantly between groups, statistical surface maps revealed subtle, highly localized reductions in

hood or whether it is limited to childhood (Nicolson andSzatmari, 2003).

The overall excess in brain volume in patients with

Available online at www.sciencedirect.com

Psychiatry Research: Neuroimaging Corresponding author. London Health Sciences Centre, 800ventricular size in patients with autism in the left frontal and occipital horns. These localized reductions in the lateral ventricles mayresult from exaggerated brain growth early in life. 2007 Elsevier Ireland Ltd. All rights reserved.

Keywords: Autism; Lateral ventricles; Brain; MRI

1. Introduction

Autism is a developmental disorder characterized bysocial deficits, impaired communication, and restrictedand repetitive patterns of behavior (American Psychia-

tric Association, 2000). There is strong evidence thatautism has a neurobiological basis, and while manystudies suggest that total brain volume is increased inchildren with autism, there is disagreement about wheth-er this enlargement in brain volume persists into adult-eDepartment of Computer Sciences, ISIMA Campus Scientifique des Cezaux, Aubiere, France

Received 16 November 2006; received in revised form 22 October 2007; accepted 7 November 2007

Abstract

In this study, a computational mapping technique was used to examine the three-dimensional profile of the lateral ventricles inautism. T1-weighted three-dimensional magnetic resonance images of the brain were acquired from 20 males with autism (age:10.13.5 years) and 22 male control subjects (age: 10.72.5 years). The lateral ventricles were delineated manually andventricular volumes were compared between the two groups. Ventricular traces were also converted into statistical three-dimensional maps, based on anatomical surface meshes. These maps were used to visualize regional morphological differences indFaculte du Medicine, INSEaLaboratory of Neuro Imaging, Brain Mapping Division, Dept. Neurology, UCLA School of Medicine, Los Angeles, California, USAbDepartment of Psychiatry, The University of Western Ontario, London, Ontario, Canada

cDepartment of Medical Biophysics, The University of Western Ontario, London, Ontario, CanadaThree-dimensional mapping o

Christine N. Vidala, Rob Nicolsonb,cCommissioners Road East, E2-601, London, Ontario, Canada N6C-2V5. Tel.: +1 519 685 8427; fax: +1 519 685 8595.

E-mail address: Rnicolso@uwo.ca (R. Nicolson).

0925-4927/$ - see front matter 2007 Elsevier Ireland Ltd. All rights resedoi:10.1016/j.pscychresns.2007.11.002he lateral ventricles in autism

Jean-Yves Boired, Vincent Barrae,a b,c

163 (2008) 106115www.elsevier.com/locate/psychresnsautismmay result from abnormally high rates of growthfor both gray and white matter in early childhood

rved.

rch: N(Courchesne et al., 2001; Herbert et al., 2003;Hazlett et al.,2005, 2006). This early hypertrophy might be expected toinfluence the volume or shape of the lateral ventricles, butmagnetic resonance imaging (MRI) studies of lateralventricular anatomy in autism have yielded inconsistentresults. Several studies have reported an increase in lateralventricle volume (Piven et al., 1995; Howard et al., 2000;Palmen et al., 2005) or area (Gaffney et al., 1989), whileother studies usingMRI (Filipek et al., 1992; Hardan et al.,2001; McAlonan et al., 2002) or computed tomographicimaging (Campbell et al., 1982; Harcherik et al., 1985;Creasey et al., 1986; Jacobson et al., 1988) have found nodifference in the total volume of the lateral ventriclesbetween patients and controls. Among those studies whichcontrolled for group differences in brain size, a majorityfound no group differences in ventricular size (Campbellet al., 1982; Harcherik et al., 1985; Jacobson et al., 1988;Hardan et al., 2001; McAlonan et al., 2002), although twostudies did find an increase in ventricular size even aftercorrecting for brain size (Gaffney et al., 1989; Palmenet al., 2005).

Factors related to subject selection may have con-tributed to the inconsistency of these past studies. How-ever, each of these studies assessed the ventricles inautism with traditional volumetric methods, examiningonly the total volume (or area) of the lateral ventricles.Total volume measures can be insensitive to anatomicalshape variability and are unlikely to identify subtleregional differences in anatomy between groups. Assuch, regional methods, which can examine more local-ized abnormalities in the ventricles, may be more appro-priate. These regional differences may provide importantclues about the nature and location of the aberrant braingrowth in autism. Only one study to date has used thisapproach: Bigler et al. (2003) examined the temporalhorns of the lateral ventricles specifically, and noted anon-significant trend for patients with normal head sizes(i.e., without macrocephaly) to have smaller temporalhorns than controls (P=0.09). Other regional methodsthat can examine more highly localized abnormalities inthe ventricles may reveal irregularities that would remainundetected by traditional volumetric analyses. Three-dimensional (3D) anatomical mapping methods, whichhave not previously been used in studies of the lateralventricles in autism, may be beneficial for detectingsubtle regional morphological abnormalities. Unliketraditional volumetric methods, surface mesh modelsand statistical maps can provide better anatomicallocalization of group differences in lateral ventricularmorphology while preserving information on subtlevariability patterns within groups (Thompson et al.,

C.N. Vidal et al. / Psychiatry Resea2004a,b). Surface-based anatomical modeling is com-plementary to voxel-based statistical methods that assessdifferences in tissue types at each voxel of stereotaxicspace. The shapemodeling approachworks by averagingthe geometry of the anatomical models across subjects,rather than comparing segmented images. These com-putational methods have detected regional alterations ofventricular morphology in conditions such as schizo-phrenia and Alzheimer's disease, and have associatedventricular shape changes with immunological declineand psychomotor impairment in HIV/AIDS (Narr et al.,2006; Thompson et al., 2004a, 2006). Ventricular mapshave also been used to quantify the effects of progressivebrain atrophy, and relate them to deteriorating cognitivefunction in longitudinal scans of patients with dementia(Thompson et al., 2004a).

In the present study, we used computational mappingmethods to visualize shape and volume abnormalities ofthe lateral ventricles in autism. Given reports of earlyrapid growth of the brain in autism (Courchesne et al.,2001, 2003; Hazlett et al., 2005), we hypothesized thatpatients with autism would have localized ventricularabnormalities, although the direction of the effect(smaller or larger) was not predicted a priori. Becauseof the relative ease of ventricular measurement in T1-weighted brain images, we were interested in determin-ing the 3D profile of ventricular shape differencesassociated with autism, which might provide importantclues regarding the nature and timing of the abnormalneurodevelopment in the disorder.

2. Methods

2.1. Subjects

Twenty males with autism (age: 10.13.5 years; range:617 years) were included in the study. The diagnosis ofautism was made using the Autism Diagnostic Interview-Revised (ADI-R; Lord et al., 1994), the AutismDiagnosticObservation Schedule (ADOS-G; Lord et al., 2000), andclinical observation. All patients met DSM-IV-TR criteriafor autism (American Psychiatric Association, 2000) aswell as ADI-R and ADOS-G algorithm criteria. Patientswere also assessed using the Wechsler Intelligence Scalefor Children, 3rd Edition (WISC-III) or the LeiterInternational Performance Scale. Exclusionary criteriafor patients included a non-verbal IQ below 70, a seizuredisorder or any other neurological condition, or a knowncytogenetic abnormality. All patients had a physicalexamination prior to participation in this study, includingmeasurement of height, weight, and head circumference.Handedness was determined by clinical observation and

107euroimaging 163 (2008) 106115the report of the patients and their parents. At the time of

rch: Nthe scan, 10 patients were being treated with psychotropicmedication: six were taking stimulants, three werereceiving dopamine antagonists, two were taking selectiveserotonin reuptake inhibitors, and one was being treatedwith a cholinesterase inhibitor. Among the 10 patients whowere not taking medications at the time of the study, sixwere medication-nave, three had discontinued theirprevious medications at least 4 weeks before imaging,and one subject who had been treated with a stimulant haddiscontinued it 1 week before his scan.

Twenty-two healthy male control subjects (age: 10.72.5 years) were recruited from the local communitythrough advertisement and word of mouth. They wereassessed with the Schedule for Affective Disorders andSchizophrenia-Childhood Version (Kaufman et al., 1997)to ensure that none had a major psychiatric disorder.Additionally, none had a personal history of neurologicaldisorders. Controls were also assessed with the WISC-IIIor the Wechsler Abbreviated Scale of Intelligence, and afull-scale IQ of less than 70 was exclusionary.

This study was approved by the Health SciencesResearch Ethics Board at the University of WesternOntario. The parents or legal guardians of all subjectsprovided written consent for participation in this study,while the subjects provided written assent.

2.2. Magnetic resonance imaging

All subjects were scanned on a 3.0 T head-onlyscanner (IMRIS, Winnipeg, Canada) with a quadraturehead coil. All patient scans and the majority of controlsubject scans were acquired in the late evening.

Standard T1-weighted localizer images were acquiredinitially. Images used for volumetric analysis were thenacquired using a T1-weighted 3DMP-RAGE (Magnetiza-tion PreparedRapidGradient Echo) sequence (TI=200ms,TR=11 ms, TE=5 ms, flip-angle 12 degrees, total scantime: 8 min) with 1.2-mm isotropic voxels.

2.3. Image processing and analysis

Each brain volume was corrected for intensityinhomogeneities (Sled et al., 1998) and resliced into astandard orientation as follows. Twenty standard ana-tomical landmarks were identified by a trained operator(C.V.) in all three planes and matched with a set ofcorresponding point locations defined on the ICBM53stereotaxic brain template (Mazziotta et al., 2001).These landmarks were used to compute a three-translation and three-rotation rigid-body transformationfor each brain volume to align it to the standardized

108 C.N. Vidal et al. / Psychiatry Reseacoordinate system of the ICBM53 average brain(Mazziotta et al., 2001). Each brain volume wasreoriented to correct for head alignment and resampledto 1.0-mm isotropic voxels using trilinear interpolation.In addition, a second set of analyses was performed toadjust the data for overall differences in brain scale. Todo this, each brain was uniformly scaled into theICBM53 stereotaxic space using a 9-parameter lineartransformation, allowing the brain to be scaled to matchthe standardized average brain template. Becausedifferences in overall brain scale were anticipatedbetween patients and controls, analyses were performedboth with and without global brain scaling.

2.4. Lateral ventricle modeling

The lateral ventricles were manually traced bilater-ally by a single image analyst (C.V.) blind to diagnosisand hemisphere. The lateral ventricles were manuallytraced by following the CSF/gray or CSF/white tissueboundaries in the gray scale image volumes correctedfor head-tilt and orientation (as described previously).The ventricles were traced in successive coronal brainslices and the digitized surface contours were displayedsimultaneously in all three viewing planes to facilitatethe accurate identification of neuroanatomic boundaries(see Fig. 1a,b). Contours were drawn on imagesmagnified four-fold to allow sub-voxel precision andfaithful tracking of small-scale features. Anatomicaldelineation was performed using a standardized ven-tricular segmentation protocol (Narr et al., 2001).Previous studies in our laboratory have demonstratedhigh interrater (Intraclass Correlation Coefficient (ICC)=0.91) and intrarater reliability (ICCN0.89) using thismethodology (Narr et al., 2001, 2006). Partial volumeeffects of CSF with the adjacent white matter andhippocampal gray matter in the inferior horns made itdifficult to identify them consistently; therefore,analysis was confined to the frontal and occipitalhorns of the ventricles. The posterior aspect of theatrium of the lateral ventricles was used as the boundarybetween the frontal and occipital ventricular horns.Volumes obtained from these tracings were retained forstatistical analyses.

2.5. Anatomical surface averaging

Anatomical mesh modeling methods (Narr et al., 2001;Thompson et al., 2004a,b) were then used to match equi-valent ventricular surface points, obtained from manualtracings, across subjects and groups (Fig. 1d,e). To matchthe digitized points representing the ventricle surface traces

euroimaging 163 (2008) 106115in each brain volume, the manually derived contours were

rch: NC.N. Vidal et al. / Psychiatry Reseamade spatially uniform by modeling them as a three-dimensional parametric surface mesh (Fig. 1c). That is, thespatial frequency of digitized points making up the ven-tricular surface traceswas equalizedwithin and across brainslices. These procedures allowmeasurements to bemade atcorresponding surface locations in each subject that maythen be compared statistically. The matching proceduresalso allow the averaging of ventricular surface morphologyacross all individuals belonging to a group (Fig. 1d) andthey record the amount of variation between correspondingsurface points relative to the group averages (Fig. 1e).

2.6. Mapping radial ventricular size

The three-dimensional parametric mesh models of eachindividual's lateral ventricles were analyzed to estimate the

Fig. 1. Ventricular surface modeling. The lateral ventricles were manually deliconverted into three-dimensional parametric surface meshes (c). Surfaces made usame group to produce an average anatomical model (c). Amedial curve, equidistaventriclesmeasured from the centerline to each surface point (e). Arrows in (e) reprThese distancemeasures are then averaged across subjects at each boundary pointcontraction of the ventricles.109euroimaging 163 (2008) 106115regional specificity of any group differences in ventricularvolume. As in our prior work (Thompson et al., 2004a,2006), amedial curvewas defined as the three-dimensionalcurve traced out by the centroid of the ventricular boundaryin each section (Fig. 1e). This medial curve, computedseparately for each individual, threads down the center ofeach individual's ventricular model. The radial size of eachventricle at each boundary point was assessed bymeasuring the radial distances from homologous ventri-cular surface points to the central core of the individual'sventricular surface model. These distances can be thoughtof as a map of radial expansion or compression, assigningnumbers to each ventricular boundary point, which recordhow far it is from the medial curve of the lateral ventricle(Fig. 1e). The distances of each ventricular surface pointto its respective medial curve are represented on the

neated on consecutive coronal magnetic resonance images (a and b) andp of spatially uniform triangular tiles were averaged across subjects in thent from each surface, was derived for each subject, and the radial size of theesent vectors from the centerline to various points on the ventricular surface.(d) and plotted in color to produce a regional measure of radial expansion or

locations on the ventricular surface maps. Uncorrectedt-values and their corresponding two-tailed probabilityvalues were mapped onto the averaged ventricularsurface models of the entire group and displayed in threedimensions in scaled and raw space.

For tests of overall volume differences and forstatistical mapping of surface-based measures, a two-tailed alpha level of Pb0.05 was used as thesignificance threshold. However, for statistical mapping,comparisons were made at thousands of ventricularsurface points. Permutation testing was therefore used toconfirm the overall significance of the statistical map-ping results, adjusting for multiple comparisons. Thisaccounts for the spatial autocorrelation of the residualsof the statistical model while adjusting for the multiplecomparisons implicit in conducting multiple statisticaltests at each point on a surface (Nichols and Holmes,

rch: Neuroimaging 163 (2008) 106115ventricular surface as a map. Since all ventricular surfacesare represented using the same parametric mesh structure,corresponding surface traces can be matched acrosssubjects and averaged across a group, together with theirassociated distance measures.

2.7. Statistical analysis and permutation tests

Age, race, handedness, height, head circumference,the scaling factor needed to transform each brain volumeinto stereotaxic space, and intelligence quotients (verbal,non-verbal, and full-scale) were compared using t-testsor chi-squared analyses.

Thickness measures improve the localization ofdeficits but may be less sensitive to differences in thelength of the ventricles. Traditional volumetric measuresmay thus be more sensitive than thickness maps fordetecting group differences in some instances, whilethickness maps may be more sensitive in others.Therefore, we used both traditional volumetric methodsand statistical maps in order to compare the ability ofboth methods to detect ventricular abnormalities amongpatients with autism.

Ventricular volumes measured using traditional volu-metric methods were compared between groups usingrepeatedmeasuresAnalysis ofVariance, with diagnosis asthe between-subjects factor and side (left and right) andventricular region (frontal and occipital) as repeatedmeasures. This analysis was applied to both the raw andscaled ventricular volumes (i.e., both before and aftercontrolling for group differences in brain size throughscaling; see section on brain size correction below).Although the two groups differed significantly in IQscores, we chose not to use IQ scores as a covariate asreductions in IQ may reflect the neurobiological abnorm-alities underlying autism (Cochrane, 1957). Further, IQscores were not correlated with ventricular volumes forboth groups. Although the age range in this study waswide, we did not covary for age in the analyses asventricular volumes were not correlated with age in eithergroup or when the two groups were combined.

The possible effects of psychotropic medications onventricular volumes determined from traditional mea-sures in patients with autism were assessed with arepeated measures Analysis of Variance. This analysiswas similar to the analysis described above with theexception that the between-groups factor consisted ofthree levels (patients receiving medication, patients notreceiving medication, and controls).

To evaluate regional differences in ventricular vol-ume as indexed by measures of ventricular radial dis-

110 C.N. Vidal et al. / Psychiatry Reseatance, Student's t-tests were performed at equivalent2002). For permutations, the assignment of subjects todiagnostic groups was randomized 100,000 times,performing statistical tests at each hippocampal surfacepoint for each random assignment. Using a threshold ofPb0.05, uncorrected, the number of significant resultsbetween patients with autism and controls in the actualdata was compared to the number of significant resultsproduced during the permutation testing to produce acorrected overall significance value for each map. Thisensures that any reported patterns of group differenceswere not observed by chance alone.

2.8. Brain size correction

For statistical mapping of ventricular surface para-meters, image volumes and the ventricular contours

Table 1Demographic characteristics of patients with autism and controlsubjects

Measures Patients Controlsubjects

Teststatistic

df P

(n=20) (n=22)

Age (years) 10.13.5 10.72.5 t=0.7 40 0.5Race (# Caucasian) 19 22 2=1.1 1 0.3Handedness (R:L) 14:6 22:0 2=7.7 1 0.007Height (cm) 145.520.0 146.213.1 t=0.1 36 0.9Head circumference(cm)

55.42.7 54.91.8 t=0.7 40 0.5

Verbal IQ 90.811.1 104.810.6 t=3.8 33 0.001Non-verbal IQ 97.814.4 101.511.6 t=0.9 38 0.4Full-scale IQ 93.811.9 103.09.6 t=2.5 33 0.02

All continuous data presented as meanstandard deviation. Degrees offreedom vary across tests due to missing data for some subjects, e.g.,

the lack of a verbal or full-scale IQ for subjects tested using the LeiterInternational Performance Scale.

were mapped into stereotaxic space, adjusting fordifferences in brain size. However, uncertainty existsas to whether or not brain size corrections increase ordecrease error variance for region of interest compar-isons (Arndt et al., 1991; Mathalon et al., 1993). Toallow for either possibility, we built the same anatomicalmaps and performed the same comparisons using raw

(descaled) ventricular volumes, which were derivedby dividing each ventricular volume by the scalingfactor used to transform it to the ICBM53 average brain.To descale the three-dimensional maps created pre-viously, the inverse of the global scaling transformationmatrix was applied to the ventricular surface points. Aleast squares rigid transform with 6 parameters (withoutscaling) was then applied to the resulting ventriculartraces to align them rigidly with the ICBM53 averagebrain dataset. Maps of group differences were createdboth before and after adjusting for any individual andgroup differences in brain size.

3. Results

3.1. Subjects

The groups did not differ significantly in terms ofage, race, height, head circumference, or total brainvolume (see Table 1). While there was no significantdifference in non-verbal IQ between the two groups,patients did have a significantly lower verbal IQ

Table 2Lateral ventricle volume, after scaling to control for brain size, inpatients with autism and control subjects

Measure Patients Control subjects

(n=20) (n=22)

Total brain volume 1576.3135.3 1572.8104.2Total ventricle 12.06.0 15.8367.7Left ventricle 6.33.6 8.43.9Right ventricle 5.82.6 7.44.2Left frontal horn 3.01.6 4.22.2Left occipital horn 2.81.6 3.22.1Right frontal horn 3.83.4 4.72.0Right occipital horn 2.51.1 3.72.0

Volumes are given in cc. All data are presented as meanstandarddeviation.

111C.N. Vidal et al. / Psychiatry Research: Neuroimaging 163 (2008) 106115Fig. 2. Scaled ventricular surface probability maps for patients with autism apatients with autism relative to controls (ac, red colors). Significant localizedhorns bilaterally as well as in the mid-portion of the left ventricle (df, whiteonly the regional reductions in the left frontal (P=0.03) and occipital horns (Pfor a reduction in a portion of the right frontal horn as well (P=0.052). (For inreferred to the web version of this article.)nd controls. The upper row shows regions of volumetric reductions inreductions in ventricular size were detected in the frontal and occipitalcolors). After permutation testing to control for multiple comparisons,=0.02) remained significant, although there was a non-significant trend

terpretation of the references to color in this figure legend, the reader is

rch: N(Pb0.0007) and full-scale IQ (P=0.02). Consistent withprevious studies (Escalante-Mead et al., 2003), therewas a significantly greater proportion of left-handedsubjects in the patient group (P=0.007).

3.2. Ventricular volume

Using traditional volumetric methods with the scaledimages, repeated measures ANOVA did not reveal anysignificant main effects of diagnosis (F1,40 = 3.1,P=0.09) or interactions of diagnosis-by-ventricularregion (F1,40=0.2, P=0.7), diagnosis-by-side (F1,40=0.7, P=0.4), or diagnosis-by-ventricular region-by-side(F1,40=1.2, P=0.3) in the overall volumetric measures(see Table 2). The pattern of results was similar when thedescaled (native space) images were used.

Repeated measures ANOVA used to determine theeffects of medication on the results from traditionalvolumetric measurements did not reveal a significantmain effect of group (patients receiving medication,patients not receiving medication, and controls). Therewere no significant group-by-side or group-by-ventri-cular region-by-side interactions. There was a signifi-cant group-by-ventricular region interaction (F2,39=3.4,P=0.04), but post-hoc univariate tests did not reveal anysignificant differences among the three groups.

3.3. Ventricular distance maps

The statistical maps revealed several localized groupdifferences in the size of the lateral ventricles afterscaling (i.e., after controlling for total brain volume; seeFig. 2). Patients with autism exhibited localizedreductions in regional ventricular volumes in the rightand left frontal and occipital horns of the ventricles aswell as in the mid-portion of the left lateral ventricle (redcolors in Fig. 2a,b,c). Permutation tests on these specificregions of interest to control for multiple testingconfirmed the significant local volume reduction in theleft frontal (P=0.03) and occipital horns (P=0.02;white colors in Fig. 2d,e,f). Although the other regionsdid not reach statistical significance after permutationtesting, there was a non-significant trend for localizedreduction in the right frontal horn to be significantlygreater among the patients (P=0.052). The native spacemaps (i.e., descaled) were essentially identical to thescaled maps, and so are not reported here.

4. Discussion

To our knowledge, this is the first study to use com-

112 C.N. Vidal et al. / Psychiatry Reseaputational mapping methods to investigate ventricularabnormalities in autism. While traditional methods didnot reveal any significant group differences in total orregional ventricular volume, computational methodsrevealed highly localized reductions in volume in theleft frontal and occipital horns of the lateral ventricles inpatients with autism. This difference remained signifi-cant both with and without controlling for total brainsize. Patients also had a strong trend for a localizeddeficit in volume in the right frontal horn.

As discussed earlier, previous MRI studies of theventricles in autism have yielded inconsistent results. Incontrast to the computational mapping methods usedhere, these earlier studies examined the total volume ofthe ventricles using traditional methods of analysis.Such an approach may be less sensitive than computa-tional methods that permit examination of regionalshape differences. As seen in previous studies ofAlzheimer's disease and HIV/AIDS (Thompson et al.,2004a, 2006), computational maps, in conjunction withpermutation testing, have more power than traditionalvolume measurements to detect group differences thatare spatially heterogeneous and that do not affect theentire volume of a structure uniformly. If deficit patternsare not spatially uniform, computational shape mappingis more likely to detect deficits that volumetric measuresmay miss. Therefore, computational mapping methodsmake it possible to detect subtle regional abnormalitiesin the lateral ventricles even in the absence of globalvolumetric differences, as in the present study. Thelower power of traditional volumetric measures mayexplain why studies using them have had inconsistentresults, with the majority finding no significant groupdifferences, as was noted here as well.

The spatial localization of volumetric abnormali-ties of the lateral ventricles provided by computationalmapping methods permits possible insight into the brainregions potentially involved in the pathophysiology ofautism and the timing of the aberrant neurodevelopmentin the disorder. A highly localized reduction in thevolume of the lateral ventricles suggests compression byassociated white and/or gray matter. Aberrant, excessivebrain growth during the first few years of life would beexpected to lead to an exaggeration in head circumfer-ence, which has been reported frequently in autism(Lord, 2006). However, if this exaggerated brain growthcontinues after the sutures have closed, then a reductionin size of the lateral ventricles would be a potentialresult. This reduction in size would be most apparent inventricular regions anatomically related to those brainregions with the greatest amount of aberrant growthafter the close of the sutures. While most human sutures

euroimaging 163 (2008) 106115do not close until adulthood, the metopic, or frontal

rch: Nsuture, closes by approximately age two (Cohen, 1993).Several authors have reported that the regions with themost abnormally increased volume in autism are frontalgray and white matter (Carper and Courchesne, 2005;Carper et al., 2002; Herbert et al., 2004). The regionalreduction in the volume of the frontal horn of the leftventricle (and the strong trend to a regional reduction inthe right frontal ventricular volume) could therefore be areflection of excessive frontal growth. Similarly, theregional reduction in the size of the left occipital horncould be secondary to excessive growth of the temporalor occipital lobes, which has been noted in some studiesof autism (Piven et al., 1996; Hazlett et al., 2006).

Alternatively, the reduction in size of the left frontalhorn could be related to abnormalities of the caudatenucleus. While this could be secondary to the use ofantipsychotic medications by some patients (Chakoset al., 1994), this seems unlikely given that ventricularsize did not differ between patients taking medicationand those not receiving medication. However, a recentmeta-analysis of volumetric brain imaging studies inautism noted that patients had an increase in caudatevolume (Stanfield et al., in press), suggesting that thisalso could potentially have resulted in a regionalreduction of the frontal horn of the ventricles.

Patients in this study had a larger ventricular occipitalhorn on the left side, while controls showed the opposite.While this could be related to abnormalities of asymmetryreported in other brain imaging studies of autism (e.g.,Herbert et al., 2002), statistical assessment of asymmetrydid not show significant group differences.

The results of this study should be interpretedcautiously due to several limitations. The sample sizein this study was relatively small and thus there mayhave insufficient power to detect group differences oflesser magnitude than those seen here. The lack offemale subjects also limits the conclusions that can bedrawn. However, given the known gender differences inthe prevalence and severity of autism (Yeargin-Allsop etal., 2003), and in the developmental trajectory of thelateral ventricles (Giedd et al., 1997), the inclusion ofmales only in this study may have highlighted groupdifferences by removing gender variables affecting thesize of the ventricles. A number of patients in the presentstudy were taking psychotropic medication, which couldpotentially have influenced the results. This seemsunlikely, however, as MRI studies of other patientgroups taking similar medications to those used inpatients with autism have not reported reductions in thesize of the lateral ventricles (Kumra et al., 2000;Peterson et al., 2001; Scherk and Falkai, 2006).

C.N. Vidal et al. / Psychiatry ReseaAdditionally, control subjects in the present study didnot differ on traditional ventricular measures frompatients receiving medication and medication-freepatients, further suggesting that the results of thisstudy are not solely due to the effects of psychotropicmedication. However, given the small sample size andthe diversity of medications used by the subjects in thepatient group, possible effects of medications on theresults cannot be ruled out conclusively. Although weassessed handedness in our subjects, it was not doneusing an objective measure.

The methodology used here requires some manualinteraction with the images to trace the lateral ventricleswhereas some other methods, such as voxel-basedmorphometry, are completely automated. If automatedventricular segmentation methods become as reliable asmanual tracing, these shape analysis methods may beefficiently extended to larger samples. We are currentlyvalidating more automated approaches for creating thesesurface maps (Carmichael et al., 2007). In this study, a 9-parameter global scaling transform was applied to eachindividual brain, to match the standardized averagebrain template. Although, in our sample, there were nosystematic group differences in overall brain volume,we still performed brain scaling to adjust for the largeinter-individual variability in brain scale, which canmake it easier to detect subtle group differences in brainsubstructures, such as the ventricles. However, it ispossible in principle that a 9-parameter scaling canintroduce effects due to possibly independent scalingsalong each of the 3 cardinal axes. To examine whetherthis scaling transformation may have affected theinferences made regarding volume reductions based onthe maps, we also created non-scaled maps, as in ourpast studies (e.g., Thompson et al., 2005a,b), in whichthe individual datasets were brought into mutualalignment using a rotational alignment but no scaling.To do this, the 9-parameter scaling transformationmatrix was inverted, applied to the surface meshes,and a 6-parameter rigid-body alignment was thenapplied to align the surface meshes before averagingand group comparison. As noted in Section 3, the non-scaled maps were essentially the same, so the maps arenot shown.

In this study, a computational mapping strategy wasused to visualize patterns of localized ventricularcontraction in children with autism. The greatestabnormalities were noted in the left frontal and occipitalhorns. Shape and computational mapping may revealthe spatial pattern of ventricular deficits with greatervisual and statistical power than traditional volumetricmethods and may thus provide important insights into

113euroimaging 163 (2008) 106115the timing of neurodevelopmental abnormalities in

114 C.N. Vidal et al. / Psychiatry Research: Neuroimaging 163 (2008) 106115autism. Future studies are required to confirm thesedifferences in larger samples, and longitudinal studieswill be needed to assess whether ongoing develop-mental changes in the ventricles among patients withautism differ from those seen in typically developingchildren and adolescents.

Acknowledgments

This work was presented at the Annual Meeting ofthe Organization for Human Brain Mapping, Toronto,2005.

This work was funded by grants from the NationalInstitute for Biomedical Imaging and Bioengineering,the National Center for Research Resources, and theNational Institute on Aging (to P.T.: R21 EB01651, R21RR019771, P50 AG016570). Additional support wasprovided by the Child and Parent Resource Institute andgrants from the London Health Sciences Foundation andthe Ontario Mental Health Foundation (R.N.), and by aHuman Brain Project grant to the InternationalConsortium for Brain Mapping, funded jointly byNIMH and NIDA (P20 MH/DA52176; P41 RR13642,and P20 MH65166 to A.W.T.).

References

American Psychiatric Association, 2000. Diagnostic and StatisticalManual of Mental Disorders, 4th Edition. Text revision. APA,Washington, D.C.

Arndt, S., Cohen, G., Alliger, R.J., Swayze, V.W., 1991. Problems withratio and proportion measures of imaged cerebral structures.Psychiatry Research. Neuroimaging 40, 7989.

Bigler, E.D., Tate, D.F., Neeley, E.S., Wolfson, L.J., Miller, M.J., Rice,S.A., Cleavinger, H., Anderson, C., Coon, H., Ozonoff, S.,Johnson, M., Dinh, E., Lu, J., Mc Mahon, W., Lainhart, J.E., 2003.Temporal lobe, autism, and macrocephaly. American Journal ofNeuroradiology 24, 20662076.

Campbell, M., Rosenbloom, S., Perry, R., George, A.E., Kricheff, I.I.,Anderson, L., Small, A.M., Jennings, S.J., 1982. Computerizedaxial tomography in young autistic children. American Journal ofPsychiatry 139, 510512.

Carmichael, O.T., Kuller, L.H., Lopez, O.L., Thompson, P.M., Dutton,R.A., Lu, A., Lee, S.E., Lee, J.Y., Aizenstein, H.J., Meltzer, C.C.,Liu, Y., Toga, A.W., Becker, J.T., 2007. Ventricular volume anddementia progression in the Cardiovascular Health Study.Neurobiology of Aging 28, 389397.

Carper, R.A., Courchesne, E., 2005. Localized enlargement of thefrontal cortex in early autism. Biological Psychiatry 57, 126133.

Carper, R.A., Moses, P., Tigue, Z.D., Courchesne, E., 2002. Cerebrallobes in autism: early hyperplasia and abnormal age effects. Neu-roImage 12, 10381051.

Chakos, M.H., Lieberman, J.A., Bilder, R.M., Borenstein, M., Lerner,G., Bogerts, B., Wu, H., Kinon, B., Ashtari, M., 1994. Increase incaudate nuclei volumes of first-episode schizophrenic patientstaking antipsychotic drugs. American Journal of Psychiatry 151,

14301436.Cochrane, W.G., 1957. Analysis of covariance: its nature and uses.Biometrics 13, 261281.

Cohen Jr., M.M., 1993. Sutural biology and the correlates of cra-niosynostosis. American Journal of Medical Genetics 47, 581616.

Courchesne, E., Karns, C.M., Davis, H.R., Ziccardi, R., Carper, R.A.,Tigue, Z.D., Chisum, H.J., Moses, P., Pierce, K., Lord, C., Lincoln,A.J., Pizzo, S., Schreibman, L., Haas, R.H., Akshoomoff, N.A.,Courchesne, R.Y., 2001. Unusual brain growth patterns in earlylife in patients with autistic disorder: an MRI study. Neurology 57,245254.

Courchesne, E., Carper, R., Akshoomoff, N., 2003. Evidence of brainovergrowth in the first year of life in autism. Journal of theAmerican Medical Association 290, 337344.

Creasey, H., Rumsey, J.M., Schwartz, M., Duara, R., Rapoport, J.L.,Rapoport, S.I., 1986. Brain morphometry in autistic men asmeasured by volumetric computed tomography. Archives ofNeurology 43, 669672.

Escalante-Mead, P.R., Minshew, N.J., Sweeney, J.A., 2003. Abnormalbrain lateralization in high-functioning autism. Journal of Autismand Developmental Disorders 33, 539543.

Filipek, P.A., Richelme, C., Kennedy, D.N., Rademacher, J., Pitcher,D.A., Zidel, S., Caviness, V.A., 1992. Morphometric analysis ofthe brain in developmental language disorders and autism. Annalsof Neurology 32, 475.

Gaffney, G.R., Kuperman, S., Tsai, L.Y., Minchin, S., 1989. Forebrainstructure in infantile autism. Journal of the American Academy ofChild and Adolescent Psychiatry 28, 534537.

Giedd, J.N., Castellanos, F.X., Rajapakse, J.C., Vaituzis, A.C.,Rapoport, J.L., 1997. Sexual dimorphism of the developinghuman brain. Progress in Neuro-Psychopharmacolology & Biolo-gical Psychiatry 21, 11851201.

Harcherik, D.F., Cohen, D.J., Ort, S., Paul, R., Shaywitz, B.A.,Volkmar, F.R., Rothman, S.L., Leckman, J.F., 1985. Computedtomographic brain scanning in four neuropsychiatric disorders ofchildhood. American Journal of Psychiatry 142, 731734.

Hardan, A.Y., Minshew, N.J.,Mallikarjuhn,M., Keshevan,M.S., 2001.Brain volume in autism. Journal of Child Neurology 16, 421424.

Hazlett, H.C., Poe, M.D., Gerig, G., Smith, R.G., Provenzale, J., Ross,A., Gilmore, J., Piven, J., 2005. Magnetic resonance imaging andhead circumference study of brain size in autism: birth through age2 years. Archives of General Psychiatry 62, 13661376.

Hazlett, H.C., Poe, M.D., Gerig, G., Smith, R.G., Piven, J., 2006.Cortical gray and white brain tissue volume in adolescents andadults with autism. Biological Psychiatry 59, 16.

Herbert, M.R., Harris, G.J., Adrien, K.T., Ziegler, D.A., Makris, N.,Kennedy, D.N., Lange, N.T., Chabris, C.F., Bakardjiev, A.,Hodgson, J., Takeoka, M., Tager-Flusberg, H., Caviness Jr., V.S.,2002. Abnormal asymmetry in language association cortex inautism. Annals of Neurology 52, 588596.

Herbert, M.R., Ziegler, D.A., Deutsch, C.K., O'Brien, L.M., Lange,N., Bakardjiev, A., Hodgson, J., Adrien, K.T., Steele, S., Makris,N., Kennedy, D., Harris, G.J., Caviness Jr., V.S., 2003. Dissocia-tions of cerebral cortex, subcortical and cerebral white mattervolumes in autistic boys. Brain 126, 11821192.

Herbert, M.R., Ziegler, D.A., Makris, N., Filipek, P.A., Kemper, T.L.,Normandin, J.J., Sanders, H.A., Kennedy, D.N., Caviness Jr., V.S.,2004. Localization of white matter volume increase in autism anddevelopmental language disorder. Annals of Neurology 55, 530540.

Howard, M.A., Cowell, P.E., Boucher, J., Broks, P., Mayes, A.,Farrant, A., Roberts, N., 2000. Convergent neuroanatomical andbehavioural evidence of an amygdala hypothesis of autism.

NeuroReport 11, 29312935.

Jacobson, R., Le Couteur, A., Howlin, P., Rutter, M., 1988. Selectivesubcortical abnormalities in autism. Psychological Medicine 18,

Nicolson, R., Szatmari, P., 2003. Genetic and neurodevelopmentalinfluences in autistic disorder. Canadian Journal of Psychiatry 48,526537.

Palmen, S.J., Hulshoff Pol, H.E., Kemner, C., Schnack, H.G., Durston, S.,Lahuis, B.E., Kahn, R.S., Van Engeland, H., 2005. Increased gray-matter volume in medication-naive high-functioning children withautism spectrum disorder. Psychological Medicine 35, 561570.

Peterson, B.S., Staib, L., Scahill, L., Zhang, H., Anderson, C.,

115C.N. Vidal et al. / Psychiatry Research: Neuroimaging 163 (2008) 106115Kaufman, J., Birmaher, B., Brent, D., Rao, U., Flynn, C., Moreci, P.,Williamson, D., Ryan, N., 1997. Schedule for Affective Disordersand Schizophrenia for School-Age Children-Present and LifetimeVersion (K-SADS-PL): initial reliability and validity data. Journalof the American Academy of Child and Adolescent Psychiatry 36,980988.

Kumra, S., Giedd, J.N., Vaituzis, A.C., Jacobsen, L.K., McKenna, K.,Bedwell, J., Hamburger, S., Nelson, J.E., Lenane,M., Rapoport, J.L.,2000. Childhood-onset psychotic disorders: magnetic resonanceimaging of volumetric differences in brain structure. AmericanJournal of Psychiatry 157, 14671474.

Lord, C., 2006. Autism spectrum disorders: phenotype and diagnosis.In:Moldin, S.O., Rubenstein, J.L.R. (Eds.), UnderstandingAutism:From Basic Neuroscience to Treatment. Taylor and Francis Group,Boca Raton, pp. 123.

Lord, C., Rutter,M., LeCouteur, A., 1994. AutismDiagnostic Interview-Revised: a revised version of a diagnostic interview for caregivers ofindividuals with possible pervasive developmental disorders. Journalof Autism and Developmental Disorders 24, 659685.

Lord, C., Risi, S., Lambrecht, L., Cook Jr., E.H., Leventhal, B.L.,DiLavore, P.C., Pickles, A., Rutter,M., 2000. TheAutismDiagnosticObservation Schedule-Generic: a standard measure of social andcommunication deficits associated with the spectrum of autism.Journal of Autism and Developmental Disorders 30, 205223.

Mathalon, D.H., Sullivan, E.V., Rawles, J.M., Pfefferbaum, A., 1993.Correction for head size in brain-imaging measurements. Psy-chiatry Research. Neuroimaging 50, 121139.

Mazziotta, J., Toga, A., Evans, A., Fox, P., Lancaster, J., Zilles, K.,Woods, R., Paus, T., Simpson, G., Pike, B., Holmes, C., Collins,L., Thompson, P., MacDonald, D., Iacoboni, M., Schormann, T.,Amunts, K., Palomero-Gallagher, N., Geyer, S., Parsons, L., Narr,K., Kabani, N., Le Goualher, G., Boomsma, D., Cannon, T.,Kawashima, R., Mazoyer, B., 2001. A probabilistic atlas andreference system for the human brain: International Consortium forBrain Mapping (ICBM). Philosophical Transactions of the RoyalSociety of London. Series B, Biological Sciences 356, 12931322.

McAlonan, G.M., Daly, E., Kumari, V., Critchley, H.D., vanAmelsvoort, T., Suckling, J., Simmons, A., Sigmundsson, T.,Greenwood, K., Russell, A., Schmitz, N., Happe, F., Howlin, P.,Murphy, D.G., 2002. Brain anatomy and sensorimotor gating inAsperger's syndrome. Brain 125, 15941606.

Narr, K.L., Thompson, P.M., Sharma, T., Moussai, J., Blanton, R., Anvar,B., Edris, A., Krupp, R., Rayman, J., Khaledy, M., Toga, A.W., 2001.Three-dimensional mapping of temporo-limbic regions and the lateralventricles in schizoprenia: gender effects. Biological Psychiatry 50,8497.

Narr, K.L., Bilder, R.M., Woods, R.P., Thompson, P.M., Szeszko, P.,Robinson, D., Ballmaier, M.,Messenger, B.,Wang, Y., Toga, A.W.,2006. Regional specificity of cerebrospinal fluid abnormalities infirst episode schizophrenia. Psychiatry Research. Neuroimaging146, 2133.

Nichols, T.E., Holmes, A.P., 2002. Nonparametric permutation tests forfunctional neuroimaging: a primer with examples. Human BrainMapping 15, 125.Leckman, J.F., Cohen, D.J., Gore, J.C., Albert, J., Webster, R.,2001. Regional brain and ventricular volumes in Tourettesyndrome. Archives of General Psychiatry 58, 427440.

Piven, J., Arndt, S., Bailey, J., Havercamp, S., Andreasen, N., Palmer,P., 1995. An MRI study of brain size in autism. American Journalof Psychiatry 152, 11451149.

Piven, J., Arndt, S., Bailey, J., Andreasen, N., Palmer, P., 1996.Regional brain enlargement in autism: a magnetic resonanceimaging study. Journal of the American Academy of Child andAdolescent Psychiatry 35, 530536.

Scherk, H., Falkai, P., 2006. Effects of antipsychotics on brainstructure. Current Opinion in Psychiatry 19, 145150.

Sled, J.G., Zijdenbos, A.P., Evans, A.C., 1998. A non-parametricmethod for automatic correction of intensity non-uniformity inMRI data. IEEE Transactions on Medical Imaging 17, 8797.

Stanfield, A.C., McIntosh, A.M., Spencer, M.D., Philip, R., Gaur, S.,Lawrie, S.M., in press. Towards a neuroanatomy of autism: asystematic review and meta-analysis of structural magnetic re-sonance imaging studies. European Psychiatry (30 August 2007,Electronic publication ahead of print).

Thompson, P.M., Hayashi, K.M., De Zubicaray, G.I., Janke, A.L., Rose,S.E., Semple, J., Hong, M.S., Herman, D.H., Gravano, D., Doddrell,D.M., Toga, A.W., 2004a. Mapping hippocampal and ventricularchange in Alzheimer disease. NeuroImage 22, 17541766.

Thompson, P.M., Hayashi, K.M., Simon, S.L., Geaga, J.A., Hong, M.S.,Sui, Y., Lee, J.Y., Toga, A.W., Ling, W., London, E.D., 2004b.Structural abnormalities in the brains of human subjects who usemethamphetamine. Journal of Neuroscience 24, 60286036.

Thompson, P.M.,Dutton,R.A., Hayashi, K.M., Toga, A.W., Lopez, O.L.,Aizenstein, H.J., Becker, J.T., 2005a. Thinning of the cerebral cortexin HIV/AIDS reflects CD4+ T-lymphocyte decline. Proceedings ofthe National Academy of Sciences 102, 1564715652.

Thompson, P.M., Lee, A.D., Dutton, R.A., Geaga, J.A., Hayashi, K.M.,Eckert,M.A., Bellugi, U., Galaburda, A.M., Korenberg, J.R.,Mills,D.L., Toga, A.W., Reiss, A.L., 2005b. Abnormal cortical complex-ity and thickness profiles mapped inWilliams syndrome. Journal ofNeuroscience 25, 41464158.

Thompson, P.M., Dutton, R.A., Hayashi, K.M., Lu, A., Lee, S.E., Lee,J.Y., Lopez, O.L., Aizenstein, H.J., Toga, A.W., Becker, J.T., 2006.3D mapping of ventricular and corpus callosum abnormalities inHIV/AIDS. NeuroImage 31, 1223.

Yeargin-Allsop, M., Rice, C., Karapurkar, T., Doernber, N., Boyle, C.,Murphy, C., 2003. Prevalence of autism in a US metropolitan area.Journal of the American Medical Association 289, 4855.3948.

Three-dimensional mapping of the lateral ventricles in autismIntroductionMethodsSubjectsMagnetic resonance imagingImage processing and analysisLateral ventricle modelingAnatomical surface averagingMapping radial ventricular sizeStatistical analysis and permutation testsBrain size correction

ResultsSubjectsVentricular volumeVentricular distance maps

DiscussionAcknowledgmentsReferences