Psychiatry Research: Neuroimaging 122 (2003) 153–167

Searching for a structural endophenotype in using computational morphometry

Machteld Marcelisa , John Suckling b,c,fd , Peter Woodruff e , Paul Hofman f , Ed Bullmore , Jim van Osa,g, *

aDepartment of Psychiatry and Neuropsychology, European Graduate School of Neuroscience, Maastricht University, P.O. Box 616, 6200 MD Maastricht, The Netherlands bClinical Age Research Unit, Department of Health Care of the Elderly, Guy’s King’s and St. Thomas’ Medical School, London, UK cDepartment of Biostatistics and Computing, Institute of Psychiatry, London, UK dUniversity of Sheffield, Sheffield, UK eDepartment of Radiology, University Hospital Maastricht, Maastricht, The Netherlands fDepartment of Psychiatry, University of Cambridge, UK gDivision of Psychological Medicine, Institute of Psychiatry, De Crespigny Park, London SE5 8AF, UK

Received 26 November 2001; received in revised form 17 September 2002; accepted 14 November 2002

Abstract

Structural cerebral abnormalities are frequently observed in . These abnormalities may indicate vulnerability for the disorder, as evidenced by reports of familial clustering of measures identified through region-of- interest analyses using manual outlining procedures. We used computational morphometry to detect structural differences within the entire brain to further examine possible structural endophenotypes. Magnetic resonance imaging scans were obtained in 31 psychotic patients, 32 non-psychotic first-degree relatives of psychotic patients and 27 healthy controls. The images were processed using an automated procedure, yielding global grey matter, white matter, CSF and total brain volume. The relative distribution ofgrey matter was compared between groups on a clustered- voxel basis. Global grey matter and total brain volume did not differ between the groups. White matter volume was significantly higher and CSF volume significantly lower in relatives compared to both cases and controls. The clustered-voxel based group comparison yielded evidence for significant grey matter deficits in fronto-thalamic- cerebellar regions, in psychotic patients, whereas the most prominent deficits in relatives involved the cerebellum. Patients with psychosis and first-degree healthy relatives of patients with psychosis show cerebellar abnormalities, which may constitute a marker ofgenetic transmission. ᮊ 2002 Elsevier Science Ireland Ltd. All rights reserved.

Keywords: Schizophrenia; Structural magnetic resonance imaging; Voxel-based; Cerebellum; Marker; Family study; Brain

*Corresponding author. Tel.: q31-43-3299-773; fax: q31-20-877-9249. E-mail address: [email protected] (J. van Os).

0925-4927/03/$ - see front matter ᮊ 2002 Elsevier Science Ireland Ltd. All rights reserved. doi:10.1016/S0925-4927(02)00125-7 154 M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167

1. Introduction which we hypothesised that structural abnormali- ties would be present in psychotic patients, and Many questions remain regarding the origins of not necessarily in the same regions given the the structural cerebral abnormalities associated possible effects of the illness and its treatment, in with schizophrenia. The structural alterations have non-psychotic first-degree relatives of patients with been directly related to the clinical phenotype, but psychosis. have also been found to be indicators of (genetic) liability for the disorder (endophenotypic markers 2. Methods ofliability ). In addition, certain structural brain alterations may be the result ofmedication and the 2.1. Study sample illness itself, thus not reflecting a possible cause but rather a consequence ofthe disorder. MRI scans were acquired from 31 patients with Studies of first-degree relatives are useful in the psychosis, 32 non-psychotic first-degree relatives search for endophenotypes. In addition to evidence ofpatients with psychosis and 27 healthy controls. for familial aggregation of ventricular enlargement The present subsample is part ofa larger study, typically derived from computed tomography scan- the Maastricht Psychosis Study (Krabbendam et ning (reviewed in Cannon and Marco, 1994), al., 2001). recent magnetic resonance imaging (MRI) studies Patients with a lifetime history of clinical psy- show more diverse patterns ofcortical and yor chosis (ofat least 2 weeks ) according to the subcortical alterations in both patients with schiz- Rearch Diagnostic Criteria (RDC)(Spitzer et al., ophrenia and their first-degree relatives (Seidman 1978), who were not in need ofin-patient treat- et al., 1997; Sharma et al., 1997; Cannon et al., ment, were recruited at the community mental 1998; Staal et al., 1998; Seidman et al., 1999; health centre in Maastricht, the Netherlands. Non- Staal et al., 2000; Wright et al., 2000). Traditional psychotic first-degree relatives were recruited MRI studies have been based on a priori defined through the participating patients, as well as regions ofinterest (ROI) and manual outlining through a local relative association. Relatives had procedures to assess volumetric measurements. to be free from a lifetime history of psychosis. This method, however, may preclude the observa- Unrelated healthy controls were sampled from the tion of significant but unexpected findings, and general population, using a mailing procedure to may have contributed to inconsistencies in the randomly selected households in the local catch- literature and publication bias (Wolkin et al., ment area. Controls were excluded whenever they 1998). The availability ofcomputational mor- had a personal or family history of psychosis or phometric techniques that permit the detection of other psychiatric disorder requiring hospital structural differences within the entire brain admission. (Andreasen et al., 1994a; Collins et al., 1994; The present sample included 78 families, of Wright et al., 1995; Wolkin et al., 1998; Bullmore which 11 families contributed one patient and one et al., 1999) may possibly lead to more consistent discordant sibling, and one family contributed two results regarding the origins ofcerebral abnormal- non-psychotic relatives (parent and sib). From the ities in schizophrenia and their possible role in the remaining 66 families, 20 independent patients, 19 pathophysiology. In addition, these techniques are independent first-degree relatives, and 27 controls generally more automated and faster than tradition- were included. al ROI methods ofanalysis, providing the oppor- Other inclusion criteria for all participants (cas- tunity to investigate larger samples. In the present es, relatives and controls) were being in the age family study, such a computational morphometric range of18–55 years and being in good health as technique was used, which comprised both global determined by a physical examination, electrocar- and regional (clustered-voxel) comparisons ofthe diography and routine laboratory investigations. relative distributions ofthe separate brain tissues. Individuals with a history ofsevere head trauma The present article focuses on grey matter, for with loss ofconsciousness, neurological disorders M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167 155 andyor other medical disorders that might have receiving antipsychotic medication (14 patients significantly affected brain function or structure received atypical, 13 received typical and one were excluded, as well as individuals who used patient received a combination ofatypical and alcohol in excess offivestandard units per day or typical antipsychotic medication). The groups did illicit drugs on a weekly basis. not differ on mean alcohol intake (in units per week) over the last year wpatients: 9.6 (S.D.s 2.2. Clinical and diagnostic procedures 12.5), relatives: 5.0 (S.D.s8.0), controls: 5.5 (S.D.s5.8), Fs2.29, d.f.s2, 87, Ps0.11). There Patients, relatives and controls were interviewed were four patients and one relative who had used with the BriefPsychiatric Rating Scale (Overall drugs within the past year (two patients and one and Gorham, 1962; Lukoff et al., 1986) and the relative had used cannabis, one patient had used Positive and Negative Syndrome Scale (Kay et al., both cannabis and a stimulant drug, and one patient 1987). They were additionally screened for symp- had used cocaine). All ofthem stopped using toms listed in the OCCPI (McGuffin et al., 1991). drugs at least 1 month prior to study participation. Where necessary, additional information was All the subjects gave written informed consent derived from case notes and interviews with the after the procedures had been fully explained in responsible medical officer. Using the combined conformity with the local ethics committee information, the computerised program OPCRIT guidelines. (McGuffin et al., 1991) yielded the following RDC diagnoses in the cases: schizophrenia (ns25) and 2.3. Image acquisition schizo-affective disorder (ns6). Among the first- degree relatives, there were four lifetime diagnoses MRI scans were obtained at the Department of and in four first-degree relatives a (life-time) of Radiology, University Hospital Maastricht, The major depression. No other psychiatric disorders Netherlands, with a Gyroscan NT T-I1 (Philips were detected in the relatives, nor were any ofthe Medical Systems) operating at 1.5 Tesla. Inter- controls diagnosed with psychiatric illness. Hand- leaved two-dimensional dual-echo fast spin-echo edness was assessed using the Annett Handedness images (60 slices, 3-mm thick, 0.3-mm gap Scale (Annett, 1970). To determine lifetime history between slices) were acquired and angled parallel ofalcohol and drug use, the appropriate sections to the clivus, covering the entire brain. Proton ofthe Composite International Diagnostic Inter- density (PD) weighted and T2-weighted images view (CIDI)(Smeets and Dingemans, 1993) were were acquired simultaneously wecho time (TE)1s used. 20 ms, TE2s100 ms, repetition time (TR)s4000 Demographic characteristics did not significant- ms, echo train lengths6, total acquisition times ly differ among the three groups in terms of age, 10 min 12 sx. The matrix size and field of view sex, handedness, height and paternal level ofoccu- were set at 256=205 and 22 cm, respectively. The pation (see Table 1). Significant differences number ofsignal averages was one. between the three groups were found for the number ofyears in education (Fs3.31, d.f.s2,87, 2.4. Image processing Ps0.04) and for educational achievement (Fs 4.65, d.f.s2,87, Ps0.01). Mean number ofyears Image processing and computations were done offormaleducation was higher in the group of on a SUN Ultra 10 (Sun MicroSystems Inc., relatives compared to both cases and controls, who Mountain View, CA, USA) workstation with the were balanced in this respect. Mean level of BAMM software (Brain Activation and Morpho- educational achievement was lowest in the patient logical Mapping, University ofCambridge, UK ). group with no difference between relatives and Initially, a mask ofparenchymal tissue was gen- controls. erated from linear scale-space features derived The mean duration ofillness in the patients was from the PD weighted images (Suckling et al., 8.5 years (S.D.s5.8). Twenty-eight patients were 1999a). Each voxel in the mask was then catego- 156 M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167

Table 1 Demographic characteristics ofthe study sample

Cases Relatives Controls F or x2 P Ns31 Ns32 Ns27 statistic Mean (S.D.) Mean (S.D.) Mean (S.D.) Gender Male 15 14 12 0.2 0.93 Female 16 18 15 Age (years) 30.7 (7.5) 35.5 (10.0) 35.5 (9.8) 2.8 0.07 Handedness (total of14-item 23.7 (6.0) 23.1 (8.2) 23.5 (6.7) 0.1 0.93 questionnaire score) Years ofeducation 13.4 (2.9) 15.0 (3.0) 13.3 (2.6) 3.31 0.04 Level ofeducational 3.7 (1.3) 4.8 (1.7) 4.3 (1.5) 4.65 0.01 achievement Paternal level ofoccupational 4.1 (1.7) 3.8 (1.1) 3.4 (1.5) 1.64 0.20 achievement rised in terms ofthe proportion occupied by grey brain depicted in a standard stereotactic atlas matter, white matter, CSF or durayblood vessels. (Talairach and Tournoux, 1988). This algorithm partitioned the feature space formed The five transformed PD images were then by the two MR echoes (PD and T2 weighting) averaged to produce a single template image in using a four-class modified fuzzy clustering standard space. The affine transformation, which scheme, and assigned continuous membership of minimised the sum of grey level differences each tissue class to every voxel (Suckling et al., between each individual’s PD weighted image and 1999b). Axial non-uniformity of image contrast the template image, was identified by the Fletcher– due to the reduction in sensitivity at the edges of Davidson–Powell algorithm (Press et al., 1992; the transmittingyreceiving coil was corrected with Brammer et al., 1997). This individually estimated a moving window scheme. Classifying data in this transformation matrix was applied, in turn, to each manner allows for changes in the distribution of ofthat subject’s three tissue probability maps to voxels in the feature space. For a detailed descrip- register them in standard space. This differs from tion, see Suckling et al. (1999b). Total cerebral the ‘optimised VBM’ method recently developed tissue volumes were obtained by summing over by Good et al. (2001a,b). In this method, an image all proportions and multiplying by the voxel is segmented in native space, and these segmented volume. maps are then separately registered to segmented The next step was the transformation of the template images in standard space. A novel feature three maps obtained from each individual into a is that it incorporates a step to adjust images for standard stereotactic space. To do this, a template the effect of the transformations necessary to map image was first constructed by piecewise linear images into standard space. The linear registration rescaling ofa subset offivePD-weighted images used in the present study applies a constant scaling from the control group. Using AFNI software factor across the image, and does not produce local (Cox, 1994), anatomical landmarks were identi- volumetric changes. fied, including the anterior and posterior commis- sures and the lateral, superior and inferior 2.5. Statistical analysis of global tissue convexities ofthe cerebral surface.The distances between landmarks were linearly rescaled to Both cases and relatives were compared sepa- approximate each individual image to the reference rately to the control reference group, in line with M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167 157

our expectation that differences between groups Here, Tj denotes the proportional volume of might arise in different parts of the brain. Group grey matter, white matter or CSF estimated at a differences in the three main brain tissue volumes given voxel for the jth individual and ej is random ( ) grey, white and CSF volume and total corrected variation. The independent variables Groupjj , Age , brain volume in the native space ofeach individual Handjj , Gender and Global j denote the group mem- (TCBVsgreyqwhiteqCSF volumes in cubic bership, age, handedness, gender and global tissue centimetres) were analysed using multiple regres- volume, respectively, ofthe jth individual. sion techniques (STATA, 2001). Effect sizes were This model was fitted at each intra-cerebral expressed as the regression coefficient (b), statis- voxel ofthe observed data, with each class of tically evaluated by the Wald test. In order to take proportional volume taken in turn as the dependent into account the fact that some individuals in the variable, to yield a set of three ‘effect maps’ of sample ofrelatives and patients were clustered in coefficient a standardised by its standard error: i.e. the same families, compromising the statistical a*sayS.E.(a). This model was also fitted 10 independence ofthe observations, the CLUSTER times at each voxel for each tissue class after and ROBUST options were used in the STATA permutation ofthe elements ofthe factorcoding regression analyses. The CLUSTER option combined group membership. The order ofpermutation was with the ROBUST option allows for the use of either entirely random or between pairs ifthe data observations that are not independent within clus- were considered repeated measures. This generated ters (in this case, within families) and obtains the 10 randomised or permuted effect maps for each HuberyWhiteySandwich estimator ofvariance tissue class. Both observed and permuted effect instead ofthe traditional variance estimator. These maps were then thresholded such that ifthe abso- procedures result in standard errors that are adjust- lute value of a* was less than 1.96 (2 S.D.s from ed for clustering within families. As described the mean ofthe normal distribution ), the value of above, there were 12 families with a patient– that voxel was set to zero, and ifthe absolute relative (or relative–relative) pair, leaving 66 fam- value of a* was greater than 1.96, the value of ilies that contributed only one case, relative or that voxel was set to a* equal to y1.96. This control. procedure generates several clusters ofsuprathres- The association between the group variables hold voxels that are spatially contiguous in three (comparing relatives and cases to the reference spatial dimensions. The sum ofsuprathreshold group ofcontrols ) on the one hand, and brain voxel statistics, or ‘mass’, ofeach three-dimen- tissue volumes (outcome variables) on the other, sional cluster was measured in each ofthe 10 was investigated. Age, sex, handedness and TCBV permuted effect maps generated for each tissue were used as covariates. class; and these measurements were ordered to sample the permutation distribution ofcluster mass 2.6. Statistical analysis of tissue probability maps under the null hypothesis of zero difference between groups. The mass ofeach cluster in the Before the calculation of between-group differ- observed effect maps was then tested against two- ences, all images were smoothed with a two- tailed critical values obtained from the correspond- dimensional Gaussian filter of 4.2-mm FWHM. ing permutation distribution. This non-parametric An ANCOVA model was then fitted at each voxel or distribution-free hypothesis testing procedure in standard space where there were N proportional was adopted because there is considerable evi- volume (probability) estimates for each tissue dence from functional imaging that cluster level class. The model is written below with tissue statistics, incorporating information about the spa- proportional volume as the dependent variable: tial neighbourhood ofeach voxel, may be more sensitive than voxel test statistics (Poline and Mazoyer, 1993; Rabe-Hesketh et al., 1997), but s q q q q Tj b0 aGroupjlb Age jb2Handj b3Genderj theoretical distributions for cluster statistics may q q b4Globaljje be intractable or oflimited generalisability (Friston 158 M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167

Table 2 Mean volumes ofgrey matter, white matter, CSF and total brain volume in cubic centimetres afterautomated segmentation procedures

Cases Relatives Controls Mean (S.D.) Mean (S.D.) Mean (S.D.) Grey matter 559.5 (64.0) 572.3 (56.4) 564 (45.1) White matter** 545.4 (63.4) 588.3 (71.2) 552.7 (53.4) CSF* 169.2 (32.0) 157.5 (30.4) 166.3 (38.3) TCBV 1274.1 (127.2) 1317.9 (133.4) 1283.7 (116.9) Results from the regression analyses were adjusted for age, sex, handedness and TCBV. ** Relatives vs. controls: bs21.4, Fs8.51, d.f.s1,77, Ps0.0046; relatives vs. patients: bs22.2, Fs7.19, d.f.s1,77, Ps0.0090. * Relatives vs. controls: bsy15.0, Fs5.88, d.f.s1,77, Ps0.018; relatives vs. patients: bsy23.4, Fs13.26, d.f.s1,77, P- 0.001. et al., 1994; Poline, 1997). Cluster-level inference extracerebral) CSF volume was significantly also mitigates the multiple comparisons problem decreased compared to both the control and the associated with voxel-level analysis, simply by patient group (relatives vs. controls: bsy15.0, reducing the total number oftests by one or two Fs5.88, d.f.s1,77, Ps0.018; relatives vs. orders ofmagnitude. For greater procedural detail, patients: bsy23.4, Fs13.26, d.f.s1,77, P- and a comparative validation ofnominal Type I 0.001). (Table 2.) error control by this method, see Bullmore et al. (1999). 3.2. Comparison of images In order to deal with the fact that the sample of relatives and patients was partly dependent as The results ofusing cluster mass to test (by described above, paired statistical analyses were non-parametric inference) for differences between conducted in the patient–relative comparison, this psychotic patients and controls are shown in Fig. representing the more conservative approach 1A. Differences between non-psychotic relatives (although results were similar when unpaired tests and controls are shown in Fig. 1B, and differences were used). between patients and relatives are shown in Fig. 1C. The clustered voxels are superimposed on the 3. Results PD-weighted grey matter template image. Clusters indicating deficits in tissue density are coloured 3.1. Comparison of global volumes yellow, whereas clusters indicating excesses are coloured purple. Clusters were conservatively There was no significant effect of group on thresholded with PF0.005, the level at which the either TCBV (patients vs. controls: b equal to number ofexpected false-positiveclusters per y8.8, Fs0.11, d.f.s1,77, Ps0.74; relatives vs. image equals one (i.e. P-value times the number controls: bs35.0, Fs1.49, d.f.s1,77, Ps0.23) ofobserved clusters ). or grey matter volume, although the patients and The location in Talairach coordinates and the relatives tended to have lower grey matter volumes size ofeach cluster’s centroid are summarised for than the control group (patients vs. controls: b all comparisons in Tables 3–5. equal to y7.8, Fs1.32, d.f.s1,77, Ps0.26; rel- atives vs. controls: bsy6.5, Fs1.26, d.f.s1,77, 3.3. Case–control comparison Ps0.27). In the relatives, white matter volume was significantly increased compared to both the In the case–control comparison, the number of control group (relatives vs. controls: bs21.4, Fs observed suprathreshold clusters by thresholding 8.51, d.f.s1,77, Ps0.0046) and the patient group with PF0.005 (estimated number offalse-posi- (relatives vs. patients: bs22.2, Fs7.19, d.f.s tivess1) was: excesss1, deficits6. Thus, a sig- 1,77, Ps0.0090), and total (intraventricular and nificant difference between controls and patients M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167 159

Fig. 1. Significant differences using cluster mass to test (by randomisation) for grey matter differences between: (a) 31 psychotic patients and 27 control subjects; (b) 32 non-psychotic first-degree relatives of psychotic patients and 27 control subjects; and (c) 31 psychotic patients and 32 non-psychotic first-degree relatives. Clusters indicating deficits in tissue density are coloured yellow, whereas clusters indicating excess areas are coloured purple. The cluster-wise probability ofType I error was at Ps0.005. Results are adjusted for total tissue volume, age, sex, and handedness. Numbers indicate approximate Talairach y-coordinates. in grey matter density was identified at seven including putamen and globus pallidum (right) spatially extensive three-dimensional voxel clus- (Fig. 1A). ters. Six clusters indicated reduced grey matter density in the patients: (i) a cluster including the 3.4. Relative–control comparison caudate nucleus and the (left); (ii) a cluster including the cingulate gyrus, central gyrus In the relative–control comparison, the results and medial frontal gyrus (left); (iii) a cluster ofthresholding cluster mass with PF0.005 (esti- extending from the inferior frontal gyrus (opercular mated number offalsepositives s1) was 6 part) to insula (right); (iv) a cluster in the right (excesss1; deficitss5). One cluster indicating thalamus (centroid in the dorsal medial nucleus); grey matter deficit was found in the temporal lobe and (v) two clusters in the cerebellum (hemi- (fusiform gyrus)(right) and the other four deficit spheres, bilateral). Increased grey matter density clusters were found bilateral in the cerebellar in patients vs. controls was found in one cluster hemispheres. There was one cluster indicating grey 160 M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167

Table 3 Summary of regional grey matter density differences—Case–control comparison

Cerebral region Side N x y z Grey matter deficit Caudate nucleus, amygdala L 1051 4.8 18.6 6.6 Cingulate gyrus, central gyrus, L 576 0.5 17.2 42.2 medial frontal gyrus Inferior frontal gyrus R 745 y35.7 9.9 12.9 (opercular part), insula Thalamus: dorsal R 208 y7.3 y16.1 7.2 medial nucleus Cerebellum R 1605 y26.0 y53.6 y55.9 Cerebellum L 1253 22.0 y55.9 y59.1 Grey matter excess Putamen, globus pallidum R 1649 y25.4 y1.8 14.8 The location ofeach cluster’s centroid is given in Talairach coordinates wx, y and z (mm)x, Nsnumber ofvoxels in each cluster. The cluster-wise probability threshold was Ps0.005.

Table 4 Relative–control comparison

Cerebral region Side N x y z Grey matter deficit Temporal lobe R 269 y36.7 1.6 y22.6 (fusiform gyrus) Cerebellum R 2034 y25.0 y42.2 y62.2 Cerebellum R 893 y19.4 y55.9 y48.8 Cerebellum L 1519 20.9 y56.1 y55.3 Cerebellum R 609 y2.2 y78.6 y32.3 Grey matter excess Superior frontal gyrus L 447 1.3 y14.8 54.7 The location ofeach cluster’s centroid is given in Talairach coordinates wx, y and z (mm)x, Nsnumber ofvoxels in each cluster. The cluster-wise probability threshold was Ps0.005.

Table 5 Caseyrelative comparison

Cerebral region Side N x y z Grey matter deficit Frontal limbic area, superior L 2256 1.6 3.1 54.3 frontal gyrus Insula, inferior frontal gyrus R 784 y38.8 3.9 13.0 (opercular part) Cingulate gyrus L 471 3.2 y41.1 45.5 Paracentral lobule L 533 4.0 y59.5 21.9 Grey matter excess Putamen, globus pallidum R 1966 y25.5 y4.6 13.6 The location ofeach cluster’s centroid is given in Talairach coordinates wx, y and z (mm)x, Nsnumber ofvoxels in each cluster. The cluster-wise probability threshold was Ps0.005. M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167 161 matter excess in the superior frontal gyrus (left). lum as being located in the lower end ofthe coil. (Fig. 1B). However, ifexistent, such an error is likely to be non-systematic, especially since the subjects ofthe 3.5. Case–relative comparison three groups were scanned in random order throughout the study period. This would make it When the group ofpatients were compared with difficult to explain why grey matter alterations the group ofrelatives using paired statistical tests were found in patients and relatives, but not in (threshold PF0.005, estimated number offalse control subjects. positivess1), five suprathreshold clusters were More generally, the reliance ofcomputational observed (four deficits; one excess). Thus, a sig- morphometrics on image registration has brought nificant difference between relatives and patients into question its validity. As pointed out by Book- in grey matter density was identified at five spa- stein (2001), one important consequence ofthe tially extensive three-dimensional voxel clusters. affine transformation used for this work is that The clusters representing deficits included: (i) the between-group differences at a given voxel may frontal limbic area and superior frontal gyrus (left); represent a mismatch ofcortical locations due to (ii) insula and inferior frontal gyrus; (iii) cingulate locally imperfect registration rather than volumet- gyrus (left); and (iv) paracentral lobule (left). One ric changes. For example, ifthere is pathological cluster indicating increased grey matter density deformation or displacement of an anatomical was found which included the putamen and the structure in patients, then affine transformation globus pallidum (Fig. 1C). will not generally correct this local deformity matching only global size and shape. This will be 4. Discussion manifest as a discrepancy at the edges or bounda- ries ofthe structure due to its local misalignment Using computational morphometrics with struc- with the corresponding structure in the template tural MRI to investigate global brain tissue volume image. Bookstein (2001) is concerned that these and regional group differences in tissue density residual misregistration signals cannot be disam- with clustered-voxel statistics, the current family biguated from volumetric differences between study provided evidence for detectable cortical and groups in proportion ofgrey matter, say, at a subcortical grey matter deficits being present not perfectly registered voxel representing precisely only in psychotic patients, but also in non-psy- the same anatomical structure in all subjects. chotic first-degree relatives of psychotic patients. Although this argument is mathematically sound, Results showed substantial cerebellar grey matter we suggest that misregistration signals can often deficits in both groups. In addition, there was be empirically recognised as such by the existence evidence for grey matter deficits in the frontal and ofcomplementary changes in adjacent voxels rep- temporal lobe, thalamus, insula, cingulate gyrus resenting different tissue classes. For example, a and caudate nucleus in patients, as well as for focus of cortical grey matter deficit immediately temporal grey matter deficits in relatives. adjoining a focus of subcortical white matter excess seems likely to be due to local misregistra- 4.1. Methodological considerations tion ofthe cortical boundary with subjacent white matter. In any case, systematic differences between Scans were segmented according to the moving groups in proportion or probability ofgrey or window implementation (Suckling et al., 1999a), white matter are relevant to a comprehensive which reduces local misclassification by increasing localisation ofpathological brain changes, whether grey matter in areas where it is systematically they represent misregistration ofa deformedstruc- under-represented due to non-uniformity of image ture or volumetric differences in a perfectly regis- contrast. Nevertheless, it cannot be completely tered structure. We accept that affine ruled out that none ofthis error remained, and if transformation does not always allow us to be so, this may have especially affected the cerebel- confident in making this distinction, but we main- 162 M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167 tain that computational morphometric methods to date is that there is hardly any suggestion for such as ours remain invaluable for screening the irreversible structural brain damage resulting from whole brain for evidence of distributed abnormal- (chronic) cannabis and other substance abuse ities (deformities or volumetric differences). (Wert and Raulin, 1986; Cascella et al., 1991; Structural abnormalities may be non-specific for Wiesbeck and Taeschner, 1991; Liu et al., 1995; schizophrenia and can, for example, be found in Castle and Ames, 1996). Moreover, the present mood disorders (Elkis et al., 1995). Four relatives results replicate earlier structural MRI findings in had a lifetime diagnosis of major depressive dis- patients and relatives who had never used any order according to the RDC. Although it is unlikely illicit drug (see Section 4.2). that the similar findings in psychotic patients and non-psychotic relatives would be fully explained 4.2. Findings by this small subgroup ofrelatives having (had) a depressive illness, we tested this by excluding Relatives had significantly higher white matter them in the cluster analyses. This did not affect volumes and lower CSF volumes than both cases the pattern or extent ofbrain changes. and controls, whereas cases and controls did not Although groups were matched on age, sex and significantly differ on these measures. As enlarge- educational achievement, patients were slightly, ment ofCSF spaces in schizophrenia is one ofthe though not significantly, younger than both rela- most consistent findings in the neuroimaging lit- tives and controls. In addition, educational erature (Johnstone et al., 1976; Shelton et al., achievement in patients was lower than that in 1988; Raz and Raz, 1990), the only non-significant relatives and controls. As higher age is known to tendency ofhigher CSF volumes in cases com- be associated with lower grey matter density and pared to controls was perhaps unexpected. A likely therefore is a potential confounder, age was con- explanation for this is chance, neuroimaging stud- trolled for in the statistical analyses. Moreover, the ies sampling around a small effect size. The results fact that relatives exhibited more structural abnor- pertaining to white matter will be the subject of malities as compared to controls without being further investigations using a cluster-statistic based different on the variable age or educational approach. Global grey matter reduction in patients achievement, argues against age- or education- has been demonstrated in several studies (Breier related findings. et al., 1992; Zipursky et al., 1992, 1998; Lim et Excessive alcohol consumption has been found al., 1996; Sullivan et al., 1996, 1998; Gur et al., to be associated with cerebellar tissue deficits 1999). In addition, Cannon et al. (1998) reported (Sullivan et al., 2000). In the present study, a grey matter reduction in the relatives ofpatients history ofalcohol abuse or dependence was part with schizophrenia, particularly in the frontal and ofthe exclusion criteria. Moreover, there were no temporal lobes. In the present study, there were no significant differences between the three study significant differences in grey matter volume groups in the amount ofalcohol used per week, between patients and relatives on the one hand, which makes differential alcohol consumption pat- and controls on the other. However, the direction terns an unlikely explanation for the cerebellar of the effect was towards lower volume in the deficits that were seen in patients and relatives. patient and relative groups. Similarly, current weekly drug use was one of The current findings provided evidence for cer- the exclusary criteria. Four patients and one rela- ebellar grey matter deficits, in both psychotic tive did, however, use illicit drugs in the past year patients and non-psychotic first degree relatives. prior to participation in the study. None ofthem In patients, there was additional evidence for used any drugs in at least the last month prior to frontal and thalamic deficits. The presence of study participation. The probability that the struc- fronto-thalamic-cerebellar deficits suggests a tural brain alterations found in the patient and the dysfunctional cortico-cerebellar-thalamic-cortical relative groups are induced by drug use is assumed (CCTC) circuit, as proposed by Andreasen et al. to be negligible as the tendency from the literature (1998, 1999). A disruption in this circuitry may M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167 163 possibly lead to ‘cognitive dysmetria’, or difficulty prevalence as well as moderate effect sizes of in coordinating the processing, prioritisation, frontal and temporal lobe deficits in schizophrenia. retrieval and expression ofinformation (Andreasen In the majority ofMRI studies investigating the et al., 1998). Non-psychotic first-degree relatives thalamus, an alteration in volume or shape ofthis in this study also exhibited grey matter deficits in structure was noted in patients (Andreasen et al., the cerebellum. As described above, the cerebellum 1990, 1994a; Flaum et al., 1995; Buchsbaum et may represent a fundamental element of the CCTC al., 1996; Hazlett et al., 1999; Konick and Fried- or any other cortico-(sub)cortical) circuit, and man, 2001) and in relatives ofpatients with schiz- alterations therein may lead to dysconnectivity. ophrenia (Seidman et al., 1997, 1999; Staal et al., Whatever the precise mechanism, however, the 1998). finding of cerebellar abnormalities in patients and The cerebellum has been less well investigated, relatives may, at least partly, reflect an association but has received renewed interest due to the with the genetic factors that predispose for psy- recognition ofits role, besides motor functions,in chosis vulnerability. cognition (Rapoport et al., 2000), the disturbance The presence ofstructural deficitsin elements ofwhich could be implicated in the pathophysiol- ofcortico- (sub)cortical circuits leading to impaired ogy and aetiology ofschizophrenia (Andreasen et information processing is compatible with evi- al., 1998; Wassink et al., 1999). Evidence for dence for generalised cognitive deficits in schizo- structural and functional abnormalities of the cer- phrenia (Mohamed et al., 1999), and with ebellum in schizophrenia has come from several neuropsychological findings in the present sample studies (Jacobsen et al., 1997; Levitt et al., 1999; showing that both cases and relatives perform Loeber et al., 1999, 2001; Nopoulos et al., 1999; diffusely worse than controls on a broad range of for reviews, see Martin and Albers, 1995; Katsetos cognitive tasks (Krabbendam et al., 2001). The et al., 1997). One group investigated cerebellar latter study replicates the findings of Cannon et abnormalities in relatives using structural MRI, al. (1994). and found a tendency towards volume reduction The observed regional deficits or excesses in in the group ofall relatives, and significantvolume grey matter correspond well with previously reduction in siblings only (Seidman et al., 1999). detected regions in many neuroimaging studies To our knowledge, the present study is the first to investigating schizophrenia. For instance, structural demonstrate structural cerebellar grey matter defi- frontal lobe abnormalities have been found repeat- cits in non-psychotic first-degree relatives using a edly in patients with schizophrenia (Andreasen et voxel based whole-brain analysis. al., 1986; Breier et al., 1992; Andreasen et al., A recent structural imaging study by Volz et al. 1994b; Buchanan et al., 1998; Goldstein et al., (2000) using an automatic whole-brain analysis 1999; Gur et al., 2000a), and additional evidence demonstrated reduced volumes simultaneously in suggests similar abnormalities in non-psychotic the frontal lobe, the temporal lobe, the thalamus, siblings (Cannon et al., 1998). the left cerebellar hemisphere and the right cere- Temporal lobe abnormalities, particularly in the bellar vermis in patients with schizophrenia com- hippocampalyamygdala area, have been reported pared with controls. In contrast, two other recent in numerous studies investigating patients with voxel-based studies investigating patients with schizophrenia (Breier et al., 1992; Woodruff et al., schizophrenia showed cerebellar increases in grey 1997; Nelson et al., 1998; Velakoulis et al., 1999; matter (Wilke et al., 2001; Suzuki et al., 2002), Gur et al., 2000b; Wright et al., 2000) and also in whereas they found concurring evidence for grey first-degree relatives of psychotic patients (Cannon matter deficits in the frontal and temporal lobe, et al., 1998; Seidman et al., 1999). However, recent anterior cingulate, and insula. meta-analyses on frontal lobe (Zakzanis and Hein- There is conflicting evidence with regard to richs, 1999) and temporal lobe (Zakzanis et al., alterations in the basal ganglia derived from neu- 2000) studies using structural and functional neu- ropathological studies (Heckers, 1997), but several roimaging techniques suggest a rather moderate MRI studies point towards increased volumes of 164 M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167 the and globus pallidum (Jernigan et al., References 1991; Breier et al., 1992; Swayze et al., 1992; Buchanan et al., 1993). Also, in the whole-brain Andreasen, N.C., Ehrhardt, J.C., Swayze, V.W.D., Alliger, R.J., ( ) Yuh, W.T., Cohen, G., Ziebell, S., 1990. Magnetic resonance analysis study by Volz et al. 2000 , a volume imaging ofthe brain in schizophrenia. The pathophysiologic increase in the right putamen was found in patients significance of structural abnormalities. Archives of General with schizophrenia. These increased volumes may Psychiatry 47, 35–44. be positively associated with exposure to typical Andreasen, N.C., Arndt, S., Swayze II, V., Cizadlo, T., Flaum, antipsychotics (Chakos et al., 1994; Gur et al., M., O’Leary, D., Ehrhardt, J.C., Yuh, W.T., 1994. Thalamic ) abnormalities in schizophrenia visualized through magnetic 1998; Corson et al., 1999 . Whether other mech- resonance image averaging. Science 266, 294–298. anisms, like defective synaptic pruning of subcor- Andreasen, N.C., Flashman, L., Flaum, M., Arndt, S., Swayze tical brain structures (Feinberg, 1982) or increased II, V., O’Leary, D.S., Ehrhardt, J.C., Yuh, W.T., 1994. synaptic density as a compensatory mechanism for Regional brain abnormalities in schizophrenia measured decreased input from other brain regions (Graybiel, with magnetic resonance imaging. JAMA 272, 1763–1769. 1990), may additionally affect basal ganglia size Andreasen, N.C., Nasrallah, H.A., Dunn, V., Olson, S.C., Grove, W.M., Ehrhardt, J.C., Coffman, J.A., Crossett, J.H., needs further investigation. A structural MRI study 1986. Structural abnormalities in the frontal system in ofrelatives ofschizophrenia patients founda schizophrenia. A magnetic resonance imaging study. (marginally significant) enlargement in the palli- Archives ofGeneral Psychiatry 43, 136–144. dum and a decreased volume ofthe putamen, but Andreasen, N.C., Nopoulos, P., O’Leary, D.S., Miller, D.D., the patient group was not investigated (Seidman Wassink, T., Flaum, M., 1999. Defining the phenotype of ) ( schizophrenia: cognitive dysmetria and its neural mecha- et al., 1999 . The patients in the present study but nisms. Biological Psychiatry 46, 908–920. not the relatives) displayed grey matter excesses Andreasen, N.C., Paradiso, S., O’Leary, D.S., 1998. ‘Cognitive in the globus pallidum and putamen, concurring dysmetria’ as an integrative theory ofschizophrenia: a with the majority ofstudies. dysfunction in cortical-subcortical-cerebellar circuitry? In summary, structural cerebral alterations, par- Schizophrenia Bulletin 24, 203–218. Annett, M., 1970. A classification of hand preference by ticularly in the cerebellum, can be identified in association analysis. British Journal ofPsychology 61, non-psychotic first-degree relatives. As these alter- 303–321. ations resemble those in the patients, the sug- Bookstein, F.L., 2001. Voxel-based morphometry should not gestion is that they are not illness- or be used with imperfectly registered images. Neuroimage 14, medication-related and likely to be present before 1454–1462. illness onset, thereby favouring a neurodevelop- Brammer, M.J., Bullmore, E.T., Simmons, A., et al., 1997. ( Generic brain activation mapping in functional magnetic mental origin Murray and Lewis, 1987; Weinber- resonance imaging: a nonparametric approach. Magnetic ger, 1987). Therefore, the grey matter deficits Resonance Imaging 15, 763–770. found in the present study may constitute markers Breier, A., Buchanan, R.W., Elkashef, A., Munson, R.C., ofgenetic transmission, and need furtherreplica- Kirkpatrick, B., Gellad, F., 1992. Brain morphology and tion both from structural and functional neuroim- schizophrenia. A magnetic resonance imaging study of limbic, , and caudate structures. Archives aging studies in order to elucidate the ofGeneral Psychiatry 49, 921–926. pathophysiological mechanisms underlying psy- Buchanan, R.W., Breier, A., Kirkpatrick, B., Elkashef, A., chosis vulnerability. In addition, structural endo- Munson, R.C., Gellad, F., Carpenter Jr., W.T., 1993. Struc- phenotypes may become useful candidates as tural abnormalities in deficit and nondeficit schizophrenia. quantitative measures in genetic studies. American Journal ofPsychiatry 150, 59–65. Buchanan, R.W., Vladar, K., Barta, P.E., Pearlson, G.D., 1998. Structural evaluation ofthe prefrontal cortex in schizophre- Acknowledgments nia. American Journal ofPsychiatry 155, 1049–1055. Buchsbaum, M.S., Someya, T., Teng, C.Y., Abel, L., Chin, S., We thank Truda Driesen for her assistance in Najafi, A., Haier, R.J., Wu, J., Bunney Jr., W.E., 1996. PET several aspects ofthe study and Marc Geerlings and MRI ofthe thalamus in never-medicated patients with schizophrenia. American Journal ofPsychiatry 153, for his technical assistance. This research was 191–199. supported by The Dutch Brain Society and The Bullmore, E.T., Suckling, J., Overmeyer, S., Rabe-Hesketh, S., Dutch Prevention Fund. Taylor, E., Brammer, M.J., 1999. Global, voxel, and cluster M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167 165

tests, by theory and permutation, for a difference between Faraone, S.V., Tsuang, M.T., 1999. Cortical abnormalities in two groups ofstructural MR images ofthe brain. IEEE schizophrenia identified by structural magnetic resonance Transactions on Medical Imaging 18, 32–42. imaging. Archives ofGeneral Psychiatry 56, 537–547. Cannon, T.D., Marco, E., 1994. Structural brain abnormalities Good, C.D., Johnsrude, I.S., Ashburner, J., Henson, R.N., as indicators ofvulnerability to schizophrenia. Schizophrenia Friston, K.J., Frackowiak, R.S., 2001. A voxel-based mor- Bulletin 20, 89–102. phometric study ofageing in 465 normal adult human Cannon, T.D., Zorrilla, L.E., Shtasel, D., Gur, R.E., Gur, R.C., brains. Neuroimage 14, 21–36. Marco, E.J., Moberg, P., Price, R.A., 1994. Neuropsychol- Good, C.D., Johnsrude, I., Ashburner, J., Henson, R.N., Fris- ogical functioning in siblings discordant for schizophrenia ton, K.J., Frackowiak, R.S., 2001. Cerebral asymmetry and and healthy volunteers. Archives ofGeneral Psychiatry 51, the effects of sex and handedness on brain structure: a 651–661. voxel-based morphometric analysis of465 normal adult Cannon, T.D., van Erp, T.G., Huttunen, M., Lonnqvist, J., human brains. Neuroimage 14, 685–700. Salonen, O., Valanne, L., Poutanen, V.P., Standertskjold- Graybiel, A.M., 1990. Neurotransmitters and neuromodulators Nordenstam, C.G., Gur, R.E., Yan, M., 1998. Regional gray in the basal ganglia. Trends in Neuroscience 13, 244–254. matter, white matter, and cerebrospinal fluid distributions in Gur, R.E., Maany, V., Mozley, P.D., Swanson, C., Bilker, W., schizophrenic patients, their siblings, and controls. Archives Gur, R.C., 1998. Subcortical MRI volumes in neuroleptic- ofGeneral Psychiatry 55, 1084–1091. naive and treated patients with schizophrenia. American Cascella, N.G., Pearlson, G., Wong, D.F., Broussolle, E., Journal ofPsychiatry 155, 1711–1717. Nagoshi, C., Margolin, R.A., London, E.D., 1991. Effects ofsubstance abuse on ventricular and sulcal measures Gur, R.E., Turetsky, B.I., Bilker, W.B., Gur, R.C., 1999. assessed by computerised tomography. British Journal of Reduced gray matter volume in schizophrenia. Archives of Psychiatry 159, 217–221. General Psychiatry 56, 905–911. Castle, D.J., Ames, F.R., 1996. Cannabis and the brain. Gur, R.E., Cowell, P.E., Latshaw, A., Turetsky, B.I., Grossman, Australian and New Zealand Journal ofPsychiatry 30, R.I., Arnold, S.E., Bilker, W.B., Gur, R.C., 2000. Reduced 179–183. dorsal and orbital prefrontal gray matter volumes in schiz- Chakos, M.H., Lieberman, J.A., Bilder, R.M., Borenstein, M., ophrenia. Archives ofGeneral Psychiatry 57, 761–768. Lerner, G., Bogerts, B., Wu, H., Kinon, B., Ashtari, M., Gur, R.E., Turetsky, B.I., Cowell, P.E., Finkelman, C., Maany, 1994. Increase in caudate nuclei volumes offirst-episode V., Grossman, R.I., Arnold, S.E., Bilker, W.B., Gur, R.C., schizophrenic patients taking antipsychotic drugs. American 2000. Temporolimbic volume reductions in schizophrenia. Journal ofPsychiatry 151, 1430–1436. Archives ofGeneral Psychiatry 57, 769–775. Collins, D.L., Neelin, P., Peters, T.M., Evans, A.C., 1994. Hazlett, E.A., Buchsbaum, M.S., Byne, W., Wei, T.C., Spiegel- Automatic 3D intersubject registration ofMR volumetric Cohen, J., Geneve, C., Kinderlehrer, R., Haznedar, M.M., data in standardized Talairach space. Journal ofComputer Shihabuddin, L., Siever, L.J., 1999. Three-dimensional anal- Assisted Tomography 18, 192–205. ysis with MRI and PET ofthe size, shape, and functionof Corson, P.W., Nopoulos, P., Miller, D.D., Arndt, S., Andreasen, the thalamus in the schizophrenia spectrum. American Jour- N.C., 1999. Change in basal ganglia volume over 2 years nal ofPsychiatry 156, 1190–1199. in patients with schizophrenia: typical versus atypical neu- Heckers, S., 1997. Neuropathology ofschizophrenia: cortex, roleptics. American Journal ofPsychiatry 156, 1200–1204. thalamus, basal ganglia, and neurotransmitter-specific pro- Cox, R.W., 1994. Analysis ofFunctional Neuroimages, version jection systems. Schizophrenia Bulletin 23, 403–421. 1.01. Medical College ofWisconsin, Madison. Jacobsen, L.K., Giedd, J.N., Berquin, P.C., Krain, A.L., Ham- Elkis, H., Friedman, L., Wise, A., Meltzer, H.Y., 1995. Meta- burger, S.D., Kumra, S., Rapoport, J.L., 1997. Quantitative analyses ofstudies ofventricular enlargement and cortical morphology ofthe cerebellum and fourthventricle in child- sulcal prominence in mood disorders. Comparisons with hood-onset schizophrenia. American Journal ofPsychiatry controls or patients with schizophrenia. Archives ofGeneral 154, 1663–1669. Psychiatry 52, 735–746. Feinberg, I., 1982. Schizophrenia: caused by a fault in pro- Jernigan, T.L., Zisook, S., Heaton, R.K., Moranville, J.T., grammed synaptic elimination during adolescence? Journal Hesselink, J.R., Braff, D.L., 1991. Magnetic resonance ofPsychiatric Research 17, 319–334. imaging abnormalities in lenticular nuclei and cerebral Flaum, M., Swayze II, V.W., O’Leary, D.S., Yuh, W.T., Ehr- cortex in schizophrenia. Archives ofGeneral Psychiatry 48, hardt, J.C., Arndt, S.V., Andreasen, N.C., 1995. Effects of 881–890. diagnosis, laterality, and gender on brain morphology in Johnstone, E.C., Crow, T.J., Frith, C.D., Husband, J., Kreel, schizophrenia. American Journal ofPsychiatry 152, L., 1976. Cerebral ventricular size and cognitive impairment 704–714. in chronic schizophrenia. Lancet 2, 924–926. Friston, K.J., Jezzard, P., Turner, R., 1994. Analysis offunc- Katsetos, C.D., Hyde, T.M., Herman, M.M., 1997. Neuropa- tional MRI time-series. Human Brain Mapping 1, 153–171. thology ofthe cerebellum in schizophrenia—an update: Goldstein, J.M., Goodman, J.M., Seidman, L.J., Kennedy, 1996 and future directions. Biological Psychiatry 42, D.N., Makris, N., Lee, H., Tourville, J., Caviness Jr., V.S., 213–224. 166 M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167

Kay, S.R., Fiszbein, A., Opler, L.A., 1987. The Positive and Overall, J.E., Gorham, D.R., 1962. The BriefPsychiatric Negative Syndrome Scale (PANSS) for schizophrenia. Rating Scale. Psychological Reports 10, 779–812. Schizophrenia Bulletin 13, 261–276. Poline, J.B., Mazoyer, B.M., 1993. Analysis ofindividual Konick, L.C., Friedman, L., 2001. Meta-analysis ofthalamic positron emission tomography activation maps by detection size in schizophrenia. Biological Psychiatry 49, 28–38. ofhigh signal-to-noise pixel clusters. Journal ofCerebral Krabbendam, L., Marcelis, M., Delespaul, P., Jolles, J., Os, J. Blood Flow and Metabolism 13, 325–437. van, 2001. Are there multiple cognitive endophenotypes in Poline, J., 1997. Combining spatial extent and peak intensity schizophrenia? American Journal ofMedical (Neu- to test for activation in functional imaging. NeuroImage 5, ropsychiatric Genetics) 105(2), 183–188. 83–96. Levitt, J.J., McCarley, R.W., Nestor, P.G., Petrescu, C., Don- Press, W.H., Teukolsky, S.A., Vetterling, W.T., Flannery, B.P., nino, R., Hirayasu, Y., Kikinis, R., Jolesz, F.A., Shenton, 1992. Numerical Recipes in C: The Art ofScientific M.E., 1999. Quantitative volumetric MRI study ofthe Computing. Cambridge University Press, Cambridge. cerebellum and vermis in schizophrenia: clinical and cog- Rabe-Hesketh, S., Bullmore, E.T., Brammer, M.J., 1997. Anal- nitive correlates. American Journal ofPsychiatry 156, ysis offunctionalmagnetic resonance images. Statististical 1105–1107. Methods in Medical Research 6, 215–237. Lim, K.O., Harris, D., Beal, M., Hoff, A.L., Minn, K., Rapoport, M., van Reekum, R., Mayberg, H., 2000. The role Csernansky, J.G., Faustman, W.O., Marsh, L., Sullivan, E.V., ofthe cerebellum in cognition and behavior: a selective Pfefferbaum, A., 1996. Gray matter deficits in young onset review. Journal ofNeuropsychiatry and Clinical Neurosci- schizophrenia are independent ofage ofonset. Biological ence 12, 193–198. Psychiatry 40, 4–13. Raz, S., Raz, N., 1990. Structural brain abnormalities in the Liu, X., Phillips, R.L., Resnick, S.M., Villemagne, V.L., Wong, major psychoses: a quantitative review ofthe evidence from D.F., Stapleton, J.M., London, E.D., 1995. Magnetic reso- computerized imaging. Psychological Bulletin 108, 93–108. nance imaging reveals no ventriculomegaly in polydrug Seidman, L.J., Faraone, S.V., Goldstein, J.M., Goodman, J.M., abusers. Acta Neurologica Scandinavica 92, 83–90. Kremen, W.S., Matsuda, G., Hoge, E.A., Kennedy, D., Loeber, R.T., Cinton, C.M., Yurgelun-Todd, D.A., 2001. Mor- Makris, N., Caviness, V.S., Tsuang, M.T., 1997. Reduced phometry ofindividual cerebellar lobules in schizophrenia. subcortical brain volumes in nonpsychotic siblings of American Journal ofPsychiatry 158 (6), 952–954. schizophrenic patients: a pilot magnetic resonance imaging Loeber, R.T., Sherwood, A.R., Renshaw, P.F., Cohen, B.M., study. American Journal ofMedical Genetics 74, 507–514. Yurgelun-Todd, D.A., 1999. Differences in cerebellar blood Seidman, L.J., Faraone, S.V., Goldstein, J.M., Goodman, J.M., volume in schizophrenia and . Schizophrenia Kremen, W.S., Toomey, R., Tourville, J., Kennedy, D., Research 37, 81–89. Makris, N., Caviness, V.S., Tsuang, M.T., 1999. Thalamic Lukoff, D., Nuechterlein, K.H., Ventura, J., 1986. Manual for and amygdala-hippocampal volume reductions in first- the Expanded BriefPsychiatric Rating Scale. Schizophrenia degree relatives ofpatients with schizophrenia: an MRI- Bulletin 12, 594–602. based morphometric analysis. Biological Psychiatry 46, Martin, P., Albers, M., 1995. Cerebellum and schizophrenia: a 941–954. selective review. Schizophrenia Bulletin 21, 241–250. Sharma, T., du Boulay, G., Lewis, S., Sigmundsson, T., McGuffin, P., Farmer, A., Harvey, I., 1991. A polydiagnostic Gurling, H., Murray, R., 1997. The Maudsley Family Study application ofoperational criteria in studies ofpsychotic I: Structural brain changes on magnetic resonance imaging illness. Development and reliability ofthe OPCRIT system. in familial schizophrenia. Progress in Neuropsychopharma- Archives ofGeneral Psychiatry 48, 764–770. cology and Biological Psychiatry 21, 1297–1315. Mohamed, S., Paulsen, J.S., O’Leary, D., Arndt, S., Andreasen, Shelton, R.C., Karson, C.N., Doran, A.R., Pickar, D., Bigelow, N., 1999. Generalized cognitive deficits in schizophrenia: a L.B., Weinberger, D.R., 1988. Cerebral structural pathology study offirst-episodepatients. Archives ofGeneral Psychi- in schizophrenia: evidence for a selective prefrontal cortical atry 56, 749–754. defect. American Journal of Psychiatry 145, 154–163. Murray, R.M., Lewis, S.W., 1987. Is schizophrenia a neuro- Smeets, R., Dingemans, P., 1993. Composite International developmental disorder? Editorial. British Medical Journal Diagnostic Interview (CIDI), version 1.1. World Health (Clinical Research Edition) 295, 681–682. Organization, Geneva. Nelson, M.D., Saykin, A.J., Flashman, L.A., Riordan, H.J., Spitzer, R.L., Endicott, J., Robins, E., 1978. Research diag- 1998. Hippocampal volume reduction in schizophrenia as nostic criteria: rationale and reliability. Archives ofGeneral assessed by magnetic resonance imaging: a meta-analytic Psychiatry 35, 773–782. study. Archives ofGeneral Psychiatry 55, 433–440. Staal, W.G., Hulshoff Pol, H.E., Schnack, H., van der Schot, Nopoulos, P.C., Ceilley, J.W., Gailis, E.A., Andreasen, N.C., A.C., Kahn, R.S., 1998. Partial volume decrease ofthe 1999. An MRI study ofcerebellar vermis morphology in thalamus in relatives ofpatients with schizophrenia. Amer- patients with schizophrenia: evidence in support ofthe ican Journal ofPsychiatry 155, 1784–1786. cognitive dysmetria concept. Biological Psychiatry 46, Staal, W.G., Hulshoff Pol, H.E., Schnack, H.G., Hoogendoorn, 703–711. M.L., Jellema, K., Kahn, R.S., 2000. Structural brain abnor- M. Marcelis et al. / Psychiatry Research: Neuroimaging 122 (2003) 153–167 167

malities in patients with schizophrenia and their healthy Wassink, T.H., Andreasen, N.C., Nopoulos, P., Flaum, M., siblings. American Journal ofPsychiatry 157, 416–421. 1999. Cerebellar morphology as a predictor ofsymptom STATA Corporation, 2001. STATA Statistical Software, and psychosocial outcome in schizophrenia. Biological Psy- Release 7.0. College Station, TX. chiatry 45, 41–48. Suckling, J., Brammer, M.J., Lingford-Hughes, A., Bullmore, Weinberger, D.R., 1987. Implications ofnormal brain devel- E.T., 1999. Removal ofextracerebral tissues in dual-echo opment for the pathogenesis of schizophrenia. Archives of magnetic resonance images via linear scale-space features. General Psychiatry 44, 660–669. Magnetic Resonance Imaging 17, 247–256. Wert, R.C., Raulin, M.L., 1986. The chronic cerebral effects ofcannabis use. I. Methodological issues and neurological Suckling, J., Sigmundsson, T., Greenwood, K., Bullmore, E.T., findings. International Journal of 21, 605–628. 1999. A modified fuzzy clustering algorithm for operator Wiesbeck, G.A., Taeschner, K.L., 1991. A cerebral computed independent brain tissue classification of dual echo MR tomography study ofpatients with drug-induced psychoses. images. Magnetic Resonance Imaging 17, 1065–1076. European Archives ofPsychiatry and Clinical Neuroscience Sullivan, E.V., Deshmukh, A., Desmond, J.E., Mathalon, D.H., 241, 88–90. Rosenbloom, M.J., Lim, K.O., Pfefferbaum, A., 2000. Con- Wilke, M., Kaufmann, C., Grabner, A., Putz, B., Wetter, T.C., tribution ofalcohol abuse to cerebellar volume deficitsin Auer, D.P., 2001. Gray matter-changes and correlates of men with schizophrenia. Archives ofGeneral Psychiatry 57, disease severity in schizophrenia: a statistical parametric 894–902. mapping study. Neuroimage 13, 814–824. Sullivan, E.V., Lim, K.O., Mathalon, D., Marsh, L., Beal, Wolkin, A., Rusinek, H., Vaid, G., Arena, L., Lafargue, T., D.M., Harris, D., Hoff, A.L., Faustman, W.O., Pfefferbaum, Sanfilipo, M., Loneragan, C., Lautin, A., Rotrosen, J., 1998. A., 1998. A profile of cortical gray matter volume deficits Structural magnetic resonance image averaging in schizo- characteristic ofschizophrenia. Cerebral Cortex 8, 117–124. phrenia. American Journal ofPsychiatry 155, 1064–1073. Sullivan, E.V., Shear, P.K., Lim, K.O., Zipursky, R.B., Pfeffer- Woodruff, P.W., Wright, I.C., Shuriquie, N., Russouw, H., baum, A., 1996. Cognitive and motor impairments are Rushe, T., Howard, R.J., Graves, M., Bullmore, E.T., Mur- related to gray matter volume deficits in schizophrenia. ray, R.M., 1997. Structural brain abnormalities in male Biological Psychiatry 39, 234–240. schizophrenics reflect fronto-temporal dissociation. Psycho- logical Medicine 27, 1257–1266. Suzuki, M., Nohara, S., Hagino, H., Kurokawa, K., Yotsutsuji, Wright, I.C., McGuire, P.K., Poline, J.B., Travere, J.M., Mur- T., Kawasaki, Y., Takahashi, T., Matsui, M., Watanabe, N., ray, R.M., Frith, C.D., Frackowiak, R.S., Friston, K.J., 1995. Seto, H., Kurachi, M., 2002. Regional changes in brain gray A voxel-based method for the statistical analysis of gray and white matter in patients with schizophrenia demonstrat- and white matter density applied to schizophrenia. Neuroim- ed with voxel-based analysis ofMRI. Schizophrenia age 2, 244–252. Research 55, 41–54. Wright, I.C., Rabe-Hesketh, S., Woodruff, P.W., David, A.S., Swayze II, V., Andreasen, N.C., Alliger, R.J., Yuh, W.T., Murray, R.M., Bullmore, E.T., 2000. Meta-analysis of Ehrhardt, J.C., 1992. Subcortical and temporal structures in regional brain volumes in schizophrenia. American Journal affective disorder and schizophrenia: a magnetic resonance ofPsychiatry 157 (1), 16–25. imaging study. Biological Psychiatry 31, 221–240. Zakzanis, K.K., Heinrichs, R.W., 1999. Schizophrenia and the Talairach, J., Tournoux, P., 1988. Co-Planar Stereotaxic Atlas frontal brain: a quantitative review. Journal of International ofthe Human Brain. Thieme, New York. Neuropsychology and Sociology 5, 556–566. Velakoulis, D., Pantelis, C., McGorry, P.D., Dudgeon, P., Zakzanis, K.K., Poulin, P., Hansen, K.T., Jolic, D., 2000. Searching the schizophrenic brain for temporal lobe deficits: Brewer, W., Cook, M., Desmond, P., Bridle, N., Tierney, P., a systematic review and meta-analysis. Psychological Med- Murrie, V., Singh, B., Copolov, D., 1999. Hippocampal icine 30, 491–504. volume in first-episode psychoses and chronic schizophre- Zipursky, R.B., Lambe, E.K., Kapur, S., Mikulis, D.J., 1998. nia: a high-resolution magnetic resonance imaging study. Cerebral gray matter volume deficits in first episode psy- Archives ofGeneral Psychiatry 56, 133–141. chosis. Archives ofGeneral Psychiatry 55, 540–546. Volz, H., Gaser, C., Sauer, H., 2000. Supporting evidence for Zipursky, R.B., Lim, K.O., Sullivan, E.V., Brown, B.W., the model ofcognitive dysmetria in schizophrenia—a struc- Pfefferbaum, A., 1992. Widespread cerebral gray matter tural magnetic resonance imaging study using deformation- volume deficits in schizophrenia. Archives of General Psy- based morphometry. Schizophrenia Research 46, 45–56. chiatry 49, 195–205.