Abstract

Quantitative magnetic resonance imaging (MRI) studies in patients with schizophrenia have shown reliable deficits in global tissue volume as well as some regionally specific changes, particularly in the temporal and frontal lobes. Recent technical advances have enabled automated voxel-wise analyses, which have the advantage of facilitating whole brain coverage without the restrictions of anatomically defined regions of interest and imperfect rater reliability. We used such a method to estimate voxel composition from segmentation of bivariate, dual-echo spin-echo data in 72 men with schizophrenia. Of these, 41 had a prominent history of auditory– verbal hallucinations and 31 had no such history. The patients were compared with 32 age, gender, handedness and IQ matched healthy controls. The study revealed localized areas of reduced grey-matter tissue proportion aggregating around the medial temporal lobes, the insulae, orbito-frontal cortex including anterior cingulate, and the precuneus (and lingual) gyri, in the schizophrenia patients as a whole. There were also reductions in white-matter tissue proportion extending along much of the large anterior–posterior frontal tracts in the right hemisphere. Small regions of increased grey matter were also noted in the right inferior parietal lobe. A contrast between the hallucinator and non-hallucinator patient groups showed a single region of reduced grey-matter tissue proportion affecting the left insula and adjacent temporal lobe. These data confirm the utility of voxel-based morphometric methods in schizophrenia research and point towards disruption to a ‘paralimbic’ neural network, as underlying schizophrenic psychopathology in general, with abnormalities of the left insula specifically related to hallucinations.

Introduction

Non-invasive brain imaging has been used in schizophrenia research for over two decades, first with computerized X-ray tomography and more recently, magnetic resonance imaging (MRI). The superior soft tissue contrast of MRI has allowed such research to progress beyond the measurement of diffuse atrophy to the demonstration of regionally and tissue-specific abnormalities (McCarley et al., 1999). Meta-analytic reviews have confirmed greater grey than white-matter tissue loss, as well as pronounced reduction in medial temporal lobe volumes bilaterally in schizophrenic patients (Woodruff et al., 1995a; Lawrie and Abukmeil, 1998; Nelson et al., 1998; Wright et al., 2000).

Traditional quantitative morphometric methods have the following properties. They require a prior hypothesis or ‘region of interest’ (ROI), usually based on lobar topography or traditional structural neuroanatomy (e.g. the hippocampus, Sylvian fissure, planum temporale, etc.). Therefore, ROI morphometry is rarely comprehensive (Goldstein et al., 1999). Such methods are labor intensive, paradoxically more so as the technology improves, since there are more and more thinner slices to trace around. As a result, intraand inter-operator reliabilities are <1. On the positive side, the selection of ROI’s may be theoretically motivated and makes use of expert information. Finally, regions may be defined in native space, obviating the problems of standardization.

Another approach is computational morphometrics. This requires no prior hypothesis and so is not bound by topographical regions with clearly identifiable boundaries but perhaps little functional specificity or validity. These methods are comprehensive, covering the entire brain including diffuse white-matter tracts. From a practical point of view, such methods can be automated and are robust and reliable. Current techniques depend on normalization and mapping into a standard space, which may introduce error and noise. Many solutions have been proposed for this, though none is universally accepted (see below).

The first application of computational morphometrics in schizophrenia research was by Andreasen et al. (Andreasen et al., 1994a), who used a bounding box technique and linear transformation for coregistration. This revealed differences in the sub-cortical grey matter, including the right thalamus and surrounding tissues between patients and normal controls. ROI measures on an overlapping group of participants highlighted bilateral frontal lobe volume decrements (Andreasen et al., 1994b). Others (Wolkin et al., 1998) used the same technique and found instead pixel-wise signal intensity differences between the schizophrenics and controls in both cortical and periventricular areas, affecting grey and white matter.

Voxel-based morphometry (VBM) is being used increasingly in schizophrenia research. It involves statistical inference of values representing the composition of a voxel, in terms of cerebrospinal fluid and grey and white matters, in a standard space such as that of Talairach and Tournoux (Talairach and Tournoux, 1988). Estimates of tissue composition have been most widely taken from univariate, T1-weighted spoilt GRASS, or similar high resolution MR scans as in studies using statistical parametric mapping (Ashburner and Friston, 2000). The first application of this technique (Wright et al., 1995) was in a within-group study of 15 patients which showed negative correlations between left temporal grey-matter density and symptoms of reality distortion (Liddle et al., 1992) — auditory hallucinations and delusions. Another study (Chua et al., 1997), using a similar design and methods on a more chronic sample (n = 12), found a different pattern of brain-symptom correlations. Wright and colleagues (Wright et al., 1999) looking at grey matter only, later showed significant reductions in the temporal pole and insula, bilaterally, extending to the left dorsolateral prefrontal cortex. More recently, imaging studies of unselected schizophrenic in-patients (Wilke et al., 2001) and early onset cases (Paillère-Martinot et al., 2001) found grey-matter reductions in the medial frontal gyri, left insula and other left-sided temporal lobe structures. The latter group also reported white-matter reductions in both frontal lobes. Finally, Hulshoff-Pol et al. (Hulshoff-Pol et al., 2001) studied patients with schizophrenia and schizophreniform disorder from across the lifespan and showed extensive greymatter density reductions, particularly in the left temporal lobe but also insula, precuneus and posterior cingulate bilaterally.

Two studies compared automated whole brain with semiautomated volumetric measures (Wright et al., 1999; Sowell et al., 2000). The latter study group was nine cases of early onset schizophrenia. Statistical parametric maps pointed to ventricular abnormalities that were used to guide individual ROI analyses in native space, which offered partial confirmation.

Gaser et al. (Gaser et al., 1999) studied a group of schizophrenia patients and healthy volunteers, but instead of composition, group averaged deformations were calculated from displacement vectors of non-linear transformations (Ashburner and Friston, 2000) required to produce anatomical correspondence between subjects and a template brain image. They have also reported preliminary evidence of temporal lobe abnormalities specific to auditory hallucinators (Sauer et al., 2000). Other averaging techniques have been used to detect regional shape (rather than positional) differences between patients with schizophrenia and controls (DeQuardo et al., 1999).

Alternative estimates of voxel composition can be made from segmentation of bivariate, dual-echo spin-echo data (Bullmore et al., 1995; Suckling et al., 1999b), which trade potential improvements in tissue specificity against thicker slices from two-dimensional MR sequences. After standard space mapping with affine transforms, statistical inference at the voxel level and contiguous multi-voxel or cluster level are made using non-parametric statistics and randomization testing (Bullmore et al., 1999). Sigmundsson et al. (Sigmundsson et al., 2001) studied 27 negative symptom patients and normal controls and demonstrated significant greyand white-matter deficits in the left superior and medial temporal cortices and insula, as well as medial frontal structures. Associated white-matter deficits were also demonstrated.

The methods used in the study of Sigmundsson et al. have also been applied to the study of a number of other disorders, including attention deficit hyperactivity disorder or ADHD (Overmeyer et al., 2001), tuberous sclerosis (Ridler et al., 2001), amytrophic lateral sclerosis (Ellis et al., 2001) and velocardiofacial syndrome (van Amelsvoort et al., 2001). Other voxel-based methods have been used to investigate brain structure in both normal (Good et al., 2001a,b) and diseased subjects (Mummery et al., 2000; Gitelman et al., 2001; Keller et al., 2002). The clinical studies confirmed the expected distribution of atrophy detected with conventional imaging methods, thereby providing concurrent validity.

The current study uses some of the same basic tools as described by Sigmundsson et al. (Sigmundsson et al., 2001), but has a number of additional design strengths. First, the sample is larger and includes a within-group comparison: patients with an established history of prominent auditory–verbal hallucinations versus non-hallucinators. Secondly, cortical regions thought to be relevant to such symptoms (Barta et al., 1990; Shenton et al., 1992; Flaum et al., 1995; Levitan et al., 1999; Rajarethinam et al., 2000; Stephane et al., 2001), including the planum temporale and Sylvian fissure (Shapleske et al., 1999) of both the right and left hemispheres, as well as the corpus callosum (Günther et al., 1991), have already been quantified manually in the same subjects using clearly defined protocols, but found not to differ significantly between the groups (Rossell et al., 2001; Shapleske et al., 2001).

The specific aims of the current study were to determine whether:

  1. the regional abnormalities uncovered by Sigmundsson et al. (Sigmundsson et al., 2001) can be generalized to other types of schizophrenia patients;

  2. there are structural brain differences related to the propensity to experience auditory hallucinations and, if so, whether they are in areas outside candidate regions and whose topography fails to conform to conventional anatomical boundaries.

A subsidiary aim was to use data on greyand white-matter differences between groups (deficits and excesses) to begin to infer mechanisms underlying the pathophysiology of some of the features of schizophrenia and to guide future research.

Materials and Methods

Subjects

Male rightand left-handed schizophrenics, who met DSM-IV diagnostic criteria according to their attending physician, were screened for study eligibility by an experienced research psychiatrist (J.S.) on the basis of clinical interview and exhaustive chart review. Patients with a score of >3 for auditory hallucinations on the Scale for Assessment of Positive Symptoms or SAPS (Andreassen and Olsen, 1982) for at least 2 weeks at some point during their illness were classed as ‘hallucinators’ (n = 41) and those who scored <2 for auditory hallucinations on SAPS for all but 1 week, in the entirety of their illness, were classed as ‘non-hallucinators’ (n = 31). Patients were recruited from a variety of settings, both in-patient and community based, from South London, UK. Control subjects (left and right handers) were recruited by advertisement in a local job centre and from hospital ancillary staff (n = 32). All subjects gave written informed consent. The local research ethics committee approved the research.

Handedness was assessed using the Edinburgh Handedness Inventory. Right handers were defined scoring >0.75 and left handers were defined as scoring <0.4 (all except one non-right-handed subject scored <–0.1). This was validated objectively with a timed finger-tapping test from a computerized testing battery (NeuroScan Inc., VA). The distribution of handedness in each group is shown in Table 1.

Exclusion criteria included: history of neurological disorders, major medical illnesses, drug or alcohol dependency and, in the control population, any history of psychiatric disorder. All three groups had similar ages, handedness, IQ and years of education (Table 1), as well as ethnicity and drug and alcohol history. The hallucinators had higher SAPS scores than the other patients, but this was accounted for entirely by the hallucinations items. The only other difference between the two patient groups was that the hallucinators had higher negative symptoms scores or SANS (Andreassen and Olsen, 1982), but both groups had low scores.

Magnetic Resonance Imaging Acquisition

MR imaging was performed on a 1.5 T General Electric (Milwaukee, WI) Signa Advantage MR system. A dual-echo, fast-spin echo-pulse sequence was acquired (repetition time length TR = 4000 ms; effective echo times TE1 = 20 ms and TE2 = 100 ms) with an echo train length of 8. This was a near-coronal pulse sequence angled parallel to the clivus, of 60 contiguous 3 mm thick slices with an in-plane field of view of 22 cm across a 256 × 256 pixel matrix. This sequence was previously optimized for voxel morphometrics using a simulation tool specifically designed for the purpose (Simmons et al., 1996).

Image Analysis

Initially, extracerebral tissues were removed with an automated computational algorithm which uses linear scale-space features derived from the proton density weighted images (Suckling et al., 1999a). Probabilistic classification of the composition of each intracerebral voxel began with an operator (J.S.) selecting a small training set of voxels representing each of the three main tissue classes from each MRI dataset. These training data were used to estimate the parameters of a polychotomous logistic discriminant function by maximum likelihood; and the trained or parameterized discriminant function was used to estimate the probability for each voxel belonging to each tissue class (Bullmore et al., 1995). As a result, a set of three maps was obtained from each MRI dataset, representing the voxel-wise probabilities of belonging to grey matter, white matter and cerebrospinal fluid (CSF) tissue classes. Some segmentation algorithms for VBM (Hulshoff-Pol et al., 2001; Watkins et al., 2002) have used binary segmentation. The segmentation algorithm implemented in statistical parametric mapping is a probabilistic segmentation based on a modified mixture-model clustering algorithm (Ashburner and Friston, 1997, 2000).

Based on prior results, we equate these probabilities to the proportional volumes of each tissue class in the often heterogeneous volume of tissue represented by each voxel. So, for example, if the probability of grey-matter class membership was 0.8 for a given voxel, it was assumed that 80% of the tissue represented by that voxel was grey matter. Given the voxel dimensions in millimetres, it was then straightforward to estimate the volume in millilitres of grey matter, or any other tissue class, at each voxel. Summing over voxels, the tissue class volumes can be estimated over the whole brain. The next step involved transformation of the three probability maps obtained from each individual dataset into a standard stereotactic space. To do this, a template image was first constructed by proportional rescaling of a subset of five PD-weighted images from the control group. Using AFNI software (Cox, 1994) anatomical landmarks were identified, including the anterior and posterior commissures and lateral, superior and inferior convexities of the cerebral surface. The distances between landmarks were linearly rescaled to approximate each individual image to the reference brain depicted in a standard stereotactic atlas (Talairach and Tournoux, 1988).

The five transformed images were then averaged to produce a single template image in standard space. The affine transformation, which minimized the sum of grey-level differences between each individual’s PD weighted image and the template image, was identified by the Fletcher–Davidson–Powell algorithm (Press et al., 1992; Brammer et al., 1997). This individually estimated transformation matrix was applied in turn to each of that subject’s three tissue probability maps to register them in standard space. This differs from the ‘optimized VBM’ method recently developed (Good et al., 2001a,b). In this method, an image is segmented in native space and these segmented maps are then separately registered to segmented template images in standard space. A novel feature is that it incorporates a step to adjust images for the effect of the transformations necessary to map images into standard space. The linear registration used in the present study applies a constant scaling factor across the image and does not produce local volumetric changes.

After registration the images were written out with the same resolution as the original scans (0.859 mm in-plane, 3 mm slice thickness). Although it is possible to create images with isotropic resolution (e.g. 1 mm), this would involve subvoxel interpolation. We believe that this is not appropriate given the type of voxel-based analysis employed here. Smoothing was then performed with a two-dimensional Gaussian filter of ∼4.2 mm FWHM. Tworather than three-dimensional filtering was chosen to take account of the anisotropic voxel dimensions of the fast spin echo sequence. Using the more conservative two-dimensional smoothing avoids assumptions about the relationship of in-plane data to those acquired out of plane.

An analysis of covariance (ANCOVA) model was then fitted at each voxel in standard space where there were N proportional volume (probability) estimates for each tissue class. The model is written below with grey-matter proportional volume as the dependent variable: 

\[\mathit{G}_{\mathit{kj}}\ {=}\ \mathit{m}\ {+}\ \mathit{a}_{\mathit{k}}\ {+}\ \mathit{b}Age_{\mathit{kj}}\ {+}\ \mathit{g}Hand_{\mathit{kj}}\ {+}\ \mathit{e}_{\mathit{kj}}.\]
Here, Gkj denotes the proportional volume of grey matter estimated at a given voxel for the jth individual in the kth group; m is the overall mean; m + ak is the mean of the kth group; and ekj is random variation. The independent variables Agekj and Handkj denote the age and handedness of the jth individual in the kth group.

This model was fitted at each intra-cerebral voxel of the observed data, with each class of proportional volume taken in turn as the dependent variable, to yield a set of three ‘effect maps’ of coefficient a divided by its standard error, i.e. a* = a/SE(a). This model was also fitted 10 times at each voxel for each tissue class after random permutation of the elements of the factor coding group membership. This generated 10 randomized or permuted effect maps for each tissue class. Both observed and permuted effect maps were then thresholded such that if the absolute value of a* was <1.96 (two standard deviations from the mean of normal distribution), the value of that voxel was set to zero and if the absolute value of a* was >1.96, the value of that voxel was set to a*–1.96. This procedure generates several clusters of suprathreshold voxels that are spatially contiguous in three spatial dimensions. The sum of suprathreshold voxel statistics, or ‘mass’, of each three-dimensional cluster was measured in each of the 10 permuted effect maps generated for each tissue class; these measurements were then ordered to sample the permutation distribution of cluster mass under the null hypothesis of zero difference between groups. The mass of each cluster in the observed effect maps was then tested against critical values obtained from the corresponding permutation distribution. This non-parametric or distribution free hypothesis testing procedure was adopted because there is considerable evidence from functional imaging that cluster level statistics, incorporating information about the spatial neighbourhood of each voxel, may be more sensitive than voxel test statistics (Poline and Mazoyer, 1993; Rabe-Hesketh et al., 1997), but theoretical distributions for cluster statistics may be intractable or of limited generalizability (Friston et al., 1994; Poline, 1997). Cluster-level inference also mitigates the multiple comparisons problem associated with voxel-level analysis, simply by reducing the total number of tests by one or two orders of magnitude. For greater procedural detail and a comparative validation of nominal type I error control by this method, see elsewhere (Bullmore et al., 1999).

Given that images were mapped into standard space, a common coordinate system (with 0, 0, 0 set at the anterior commissure) could be used. A computer program was written in which all coordinates in standard space were assigned an anatomical label. The coordinates of the centroid of each three-dimensional cluster were automatically determined and assigned an anatomical label. The label was then manually checked with the Talairach atlas.

Results

Total tissue volumes

When total brain tissue volumes were compared, there was a significant difference between the groups in white matter only (Table 2). Post hoc comparisons revealed that this was due to reductions in both schizophrenia groups in comparison with controls. The two patients groups did not differ. Grey-matter and CSF volumes did not differ significantly between any of the groups.

Controls versus Combined Schizophrenia Group

Automated voxel-based analysis showed significant regional differences when comparing schizophrenics with controls. Statistical thresholds were set such that the expected number of false positive clusters (P-value times the number of tests) was <1. Hence, any cluster reported exceeded this threshold. For the grey-matter comparisons, the significance level was P ≤ 0.005, while for white matter it was P ≤ 0.01 (see Table 3 and Fig. 1, row A, upper).

There were nine clusters of grey-matter deficits in the schizophrenic group as a whole. These clusters themselves can be grouped together into four main anatomical regions around which they coalesce: (i) bilateral insula; (ii) orbito-frontal, including anterior cingulate; (iii) bilateral medial temporal; and (iv) precuneus (and lingual) gyri.

There was an isolated region of increased grey-matter tissue proportion in the schizophrenia group in the right inferior parietal lobe. Regarding white matter, there were no excesses and only one deficit, which extended over a considerable length of the right frontal lobe (from y = 26 to y = 0; Fig. 1, row A, lower).

Controls versus Hallucinators

There were several regions where there was reduced grey-matter tissue proportion in the patient sub-group compared with normal controls (Table 4 and Fig. 1, row B, upper). These were located in the same regions as in the total patient group comparison, with the exception of the medial temporal lobe abnormalities, which were right-sided only. This anatomical consistency applies to the single region of grey-matter proportion excess and the whitematter deficit (Fig. 1, row B, lower). However, the two sets of comparisons do diverge, with the hallucinators showing an apparent white-matter excess in left temporal–parietal connecting tracts, extending to the claustrum.

Controls versus Non-hallucinators

Again, there were several regions of reduced grey-matter tissue proportion in the patient sub-group compared with normal controls (Fig. 1, row C, upper). On inspection, the deficits were less extensive and more right-sided than in the hallucinators versus controls comparison. This applied to both medial temporal and insula abnormalities. Furthermore, the grey-matter deficits did not extend to the inferior frontal region, or to the posterior cingulate cortex. There was no region of excess. There was a single white-matter deficit cluster, in the right frontal lobe, but this was less extensive and inferior to that seen in the hallucinators (Fig. 1, row C, lower). The region of white-matter excess was, however, very differently located, being the right temporal lobe in the peri-hippocampal region.

Hallucinators versus Non-hallucinators

There was only one significant cluster identified in this comparison. This was a deficit confined to the left hemisphere grey matter and included the insula and adjacent temporal lobe (Fig. 1, row D). A correlation was calculated between the size of the left insula cluster and the hallucinations total score on the SAPS. The Pearson r was 0.40 (P = 0.002).

Discussion

Methodological Issues

In evaluating these results it is important to bear in mind the relative novelty of the morphometric techniques we have used and the possible bearing these methods might have on the results and their interpretation. Essentially, we have used an affine transformation to match individual images to a common template image, then tested between-group differences in the proportion or probability of grey and white matter at the level of suprathreshold voxel clusters. This approach is broadly similar to the ‘voxel-based morphometry (VBM)’ methods implemented in statistical parametric mapping (Ashburner and Friston, 2001), but differs in a number of details. For example, we have used logistic discriminant analysis to make probabilistic tissue classifications at each voxel based on dual echo MRI data and prior knowledge of global tissue class proportions; whereas current implementations of VBM typically make a binary classification of grey or white matter on the basis of finite mixture modelling of a single T1-weighted MR image. As shown by Bullmore et al. (Bullmore et al., 1995), partial volume effects are common in digital images of brain structure and a probabilistic approach to classification provides a natural way of dealing with them.

Another point of difference is in terms of the registration algorithm we have used to achieve spatial co-registration. An affine transformation linearly matches grey-scale images broadly in terms of global size and shape; it does not effect more local or non-linear transformations and it is not informed by any expert knowledge concerning locations of anatomical landmarks on the cortical surface. This is in contrast to more locally adaptive, more highly parameterized warping algorithms used elsewhere (Thompson et al., 2001; Ashburner and Friston, 2001). As pointed out by Bookstein (Bookstein, 2001), one important consequence of using an affine transformation for spatial registration in this context is that differences in grey-scale values between groups at a given voxel may represent a mismatch of cortical locations due to locally imperfect registration. For example, if there is pathological deformation or displacement of a discrete anatomical structure in patients, then affine transformation will not generally correct this local deformity and it will be manifest following registration as a discrepancy in voxel values at the edges or boundaries of the structure due to its local misalignment with the corresponding structure in the template image. Bookstein (Bookstein, 2001) is concerned that these residual misregistration signals cannot be disambiguated from volumetric differences between groups in proportion of grey matter, say, at a perfectly registered voxel representing precisely the same anatomical structure in all subjects. Although this argument is mathematically sound, we suggest, misregistration signals can often be empirically recognized as such by the existence of complementary changes in adjacent voxels representing different tissue classes. For example, a focus of cortical grey-matter deficit immediately adjoining a focus of subcortical white-matter excess seems likely to be due to local misregistration of the cortical boundary with subjacent white matter. In any case, systematic differences between groups in proportion or probability of grey or white matter are relevant to a comprehensive localization of pathological brain changes, whether they represent misregistration of a deformed structure or volumetric differences in a perfectly registered structure. We accept that affine transformation does not always allow us to be confident in making this distinction, but we maintain that computational morphometric methods such as ours remain invaluable for screening the whole brain for evidence of distributed abnormalities (deformities or volumetric differences), perhaps involving regions, such as the insula or anterior cingulate cortex, which have not traditionally been regarded as ‘regions of interest’ in schizophrenia imaging research (Wright et al., 2000), but which have been repeatedly implicated in schizophrenia following the advent of computationally intensive methods.

A potentially more troublesome implication of the misregistration issue is in relation to type 1 error control by theoretical hypothesis testing. If we imagine applying a range of warping algorithms to two groups of structural images, we can envisage that the voxel statistics for a group difference will generally be larger for the lower-dimensional, globally adaptive algorithms, because of locally imperfect registration and smaller for the higher-dimensional, locally adaptive algorithms. If these statistic maps are tested for significance against the fixed critical values of a theoretical distribution for standardized mean difference under the null hypothesis, then we might imagine differences in the number of ‘significant’ voxels related simply to the precision with which the anatomical data have been co-registered. One distinctive aspect of our approach is that we have ascertained the null distributions for statistical testing by repeated resampling of the data (Bullmore et al., 1999); others (Thompson et al., 2001) have also adopted resampling strategies for control of family-wise type 1 error in structural MRI data analysis. A key advantage of data resampling is that it allows us to test geometric properties of suprathreshold voxel clusters, which do not generally have theoretically tractable distributions under the null hypothesis (Ashburner and Friston, 2001), thereby increasing the sensitivity of the analysis to effects distributed over a local neighbourhood of voxels and also substantially reducing the search volume or total number of tests conducted. Moreover, as noted previously (Bookstein, 2001), the use of resampling means that critical values for hypothesis testing will be adaptive to the precision of anatomical registration rather than theoretically fixed and, therefore, between-group differences will be tested correctly for significance, albeit at the cost of efficiency in the context of very imperfect registration.

Clinical Implications

Grey Matter

Considering first the schizophrenia versus controls comparison, the results confirm recent meta-analyses and ROI work showing significant temporal lobe grey-matter abnormalities (Nelson et al., 1998; Sullivan et al., 1998; McCarley et al., 1999; Gur et al., 2000a; Wright et al., 2000; Hulshoff-Pol et al., 2001) in schizophrenia and less prominent, but also highly significant, medial frontal deficits (Sullivan et al., 1998; McCarley et al., 1999; Gur et al., 2000b). The results also suggest further extensive deficits in the insula, a region not widely studied. The majority of studies using voxel-based morphometry, whether single or dual-echo, show insula deficits either left-sided (Paillère-Martinot et al., 2001) or bilaterally (Wright et al., 1995, 1999; Chua et al., 1997; Sigmundsson et al., 2001). Hence, the current study has succeeded in its aim of establishing the broad generalizability of the work of Sigmundsson et al. (Sigmundsson et al., 2001) using similar analytic techniques. Studies using ROI methods to measure insula volumes (Goldstein et al., 1999; Crespo-Facorro, et al. 2000) have also found significant reductions and these correlate with positive symptoms (see below).

Increases in grey-matter volume have been reported almost exclusively in the context of the basal ganglia, namely the caudate and putamen (Chakos et al., 1994; Keshavan et al., 1994; Hulshoff-Pol et al., 2001) and have been attributed to dopamine antagonist-induced hypertrophy. Volume reductions have also been noted in subcortical grey structures (Andreasen et al., 1994a). Increases in grey-matter tissue proportion were detected in the study of Sigmundsson et al. (Sigmundsson et al., 2001), but not in the current study using the same methods — perhaps because lifetime exposure to neuroleptics was not notably high, nearly 20% were being treated with atypical agents and the patients were neither treatment resistant nor chronic. However, voxel-based morphometry uncovered a significant increase in grey-matter proportional tissue volume in patients compared to controls in the right inferior parietal lobe. Though highly significant, this finding must be interpreted with caution since it was not predicted and is rather novel. Determination of whether such an increase represents some kind of compensatory growth or aberrant neuronal migration would require microscopic validation. Functionally, this region plays a role in visual spatial attention and possibly somatic awareness (Harris et al., 2000; Landis, 2000). This, along with the deficits found in the precuneus and lingual gyrus, regions thought to be involved in imagery and mnemonic processes, may be relevant to schizophrenic psychopathology such as perceptual abnormalities. Indeed, others (Gaser et al., 1999) have found abnormalities in this part of the cortex using a deformation-based automated analysis technique.

White Matter

Cerebral white matter has been the object of quantitative study in MRI much less frequently than grey matter. The average white-matter volume deficit in schizophrenia from one metaanalysis was estimated as 2% [95% confidence intervals (CI) 0–5] compared to global grey-matter deficits of ∼4% (95% CI 1–6) (Wright et al., 2000), although Lawrie and Abukmeil found a possible average white-matter excess in their review (Lawrie and Abukmeil, 1998). The current study showed a greater generalized whitethan grey-matter proportional volume decrease overall, in comparison with controls. Specific white-matter abnormalities in schizophrenia have been reported (Sullivan et al., 1998), especially pre-frontally (Buchanan et al., 1998), but appear less extensive than grey (Cannon et al., 1998; Sanfilipo et al., 2000) — as in the current study. Deficits were detected in the region of the large antero-posterior tract known as the superior longitudinal fasciculus and may contribute to disordered functional temporal–frontal connectivity postulated to underlie certain features of schizophrenia (McGuire and Frith, 1996). It would be of value to determine whether the white-matter changes were primary, or secondary to grey-matter changes. This will depend on further histopathological work (Akbarian et al., 1996).

One reason why white-matter volumes have been studied less frequently is the difficulty in defining particular structures or tracts, with the exception of the corpus callosum (Woodruff et al., 1995b) and also the lack of sensitivity of imaging methods. Voxel-based methods offer distinct advantages in this respect. Most report white-matter changes (Wright et al., 1995; Wolkin et al., 1998; Paillère-Martinot et al., 2001). Sigmundsson et al. (Sigmundsson et al., 2001) showed left frontal and temporal white-matter deficits in his negative symptom group (Sanfilipo et al., 2000). Newer imaging modalities such as diffusion tensor (Buchsbaum et al., 1998; Lim et al., 1999) and magnetization transfer imaging (Foong et al., 2000) have revealed white-matter pathology in schizophrenia.

Hallucinations

The other aim of the current study was to examine the possible structural neural correlates of hallucinations by contrasting hallucinators and non-hallucinators, separately, with controls and finally with each other. In general, the non-hallucinators showed less extensive abnormalities than the hallucinators, who were more similar to the schizophrenia group as a whole. In other words, the hallucinator sub-group appeared to contribute the bulk of the schizophrenia versus controls differences. One qualitative difference was the more right unilateral distribution of grey-matter deficits (and absence of excesses) in the nonhallucinator group. This cannot be accounted for by differences in handedness, which were controlled for in the analysis. The implication might be that hallucinations depend more on left-sided fronto-temporal structures. This was borne out in the more specific hallucinator versus non-hallucinator comparison, which revealed a single deficit cluster in the left insula cortex extending to the uncus and most medial part of the superior temporal gyrus. The size of the left insula cluster in individual subjects correlated significantly with clinical ratings of hallucinations. Such areas have been implicated in functional MRI of hallucinations (Woodruff et al., 1995a, 1997; David et al., 1996; Dierks et al., 1999).

The combination of grey-matter ‘deficits’ and white-matter ‘excess’ in the same temporal lobe somewhat more posteriorly may point to anatomical and, hence, functional dysconnectivity. Indeed, it has been proposed that excessive connections to aberrant grey matter as a result of a failure of pruning may lead to symptoms such as hallucinations due to ‘cross-talk’ between inner speech and auditory processing modules (Nasrallah, 1985; David, 1994a,b) which are presumed to reside in the left hemisphere (McGuire et al., 1996). Further, the white-matter excess stretched to the claustrum, which has been identified as a cross-modal integration centre (Calvert et al., 1999) and abnormalities in this realm have also been postulated to underlie psychotic symptoms (Surguladze et al., 2001). In non-hallucinators, greyand white-matter changes aggregated in the right insula and temporal lobe. However, it should be conceded that no white-matter abnormalities emerged in the hallucinator versus non-hallucinator comparison, on either side, suggesting that such changes are rather subtle and occur to a variable extent across the patients, only standing out ‘in relief’ against healthy controls.

Insula and Schizophrenia

As noted above, the insula emerges as a commonly affected region in MRI studies of schizophrenia. Functionally, the anterior insula has been implicated in visceral perception, especially taste and the emotion of disgust (Phillips et al., 1997) and pain (Ploghaus et al., 1999), while the left insula has a critical role in speech production (Dronkers, 1996). The insula is a large triangular structure which extends from ∼y = +20 to –20 in the anterior–posterior plane and from ∼z = –4 to +16 in the inferior–superior plane from the Talairach and Tournoux atlas. It has a complex gross anatomy (Augustine, 1985, 1996; Türe et al., 1999; Crespo-Facorro et al., 2000) and is covered by the frontal and parietal opercula superiorly and the temporal opercula inferiorly. It is divided by the central insula sulcus which runs diagonally downwards and forwards, into the larger anterior portion which itself comprises three short gyri, while the smaller posterior portion comprises mainly the anterior and posterior long gyri. The area of proportional volume deficit in the hallucinators seems to involve the inferior parts of the anterior and middle short gyri plus, perhaps, the anterior long gyrus, whereas those in the schizophrenia group as a whole extend more posteriorly.

The pattern of deficits in the schizophrenia group is very similar to that uncovered by Goldstein et al. (Goldstein et al., 1999). These authors parcellated standardized images into 48 topographically defined brain regions. They showed greatest reductions in the middle frontal gyrus and so-called ‘paralimbic’ (anterior cingulate and paralimbic gyrus) brain regions and also in the supramarginal gyrus and insula. The insula forms a key central network in the paralimbic system (Mega et al., 1997) which has extensive reciprocal connections to inferior and orbital frontal regions, medial temporal (limbic) structures, especially amygdala, hippocampus, entorhinal cortex and superior temporal sulcus and pole, as well as parietal cortex and medially to the thalamus and basal nuclei (Augustine, 1985; 1996). Mesulam (Mesulam, 1985) proposed that there were two paralimbic ‘bands’ or streams: a more anterior orbito-frontal division involving frontal lobes, insula, amygdala, anterior parahippocampus, temporal pole and infra-callosal cingulate, and a more posterior/medial division involving the posterior hippocampus, retrosplenium, precuneus and supra-callosal cingulate. The former, it is argued plays an important role in visceral inputs and drives, while the latter is more important for memory and attention. Clearly, the regions highlighted in this study overlap considerably with the paralimbic system as so described and abnormalities in visceral drives and cognition could be said to characterize schizophrenia.

Conclusions

We have demonstrated, in a group of schizophrenia patients, a number of regional grey-matter deficits which centre primarily on the medial temporal lobes, the insulae and the medial and orbito-frontal cortex. These are the same regions that emerge time and again in ROI morphometry, in narrative and systematic meta-analytic reviews of cortical abnormalities in schizophrenia, as well as in voxel-based approaches. Future studies, particularly using computational morphometry and newer imaging modalities, will probably add consistent white-matter deficits (as well as, perhaps, relative excesses) to this overall picture. Clearly, different studies emphasize one region over another and find differences in lateral distribution, but this appears to be the result of heterogeneity of course and symptom type. Future work based on sub-groups selected on the basis of the presence or absence of such features will likely prove to be worthwhile in the understanding of such heterogeneity. Using such an approach, we can report a specific left medial temporal/insula deficit in people with prominent auditory–verbal hallucinations.

The authors wish to thank the Wellcome Trust who funded this project.

Table 1

Demographic and clinical characteristics of the subjects

Variable Controls (n = 32) Schizophrenics Significance 
  Total group (n = 72) Hallucinators (n = 41) Non-hallucinators (n = 31)  
aOne-way ANOVA between three subject groups, d.f. (2,102). 
bBetween two patient groups, d.f. (1,71). 
cχ2 distribution between the three subject groups. See text for key to abbreviations. 
Age (years)  33.3 (8.7)  34.1 (8.5)  35.3 (8.9)  32.4 (7.8) 1.12a n.s. 
Education (years)  14.7 (3.3)  13.9 (2.4)  13.9 (2.3)  14.0 (2.5) 0.82a n.s. 
National Adult Reading Test 115.5 (8.6) 113.3 (10.1) 112.4 (10.7) 114.5 (9.2) 1.01a n.s. 
Handedness  25R 7L  54R 18L  33R 8L  21R 10L 1.81b n.s. 
Length of illness (months)  – 138.1 (94.0) 149.3 (97.9) 120.3 (87.3) 0.83c n.s. 
Years on medication  –  10.9 (13.6)  13.3 (17.1)  8.2 (7.2) 1.64 n.s. 
Mean SAPS  –  1.24 (0.90)  1.60 (0.87)  0.78 (0.74) 4.23bP < 0.001 
Mean SANS  –  1.76 (1.01)  2.25 (0.93)  1.15 (0.77) 5.39bP < 0.001 
Variable Controls (n = 32) Schizophrenics Significance 
  Total group (n = 72) Hallucinators (n = 41) Non-hallucinators (n = 31)  
aOne-way ANOVA between three subject groups, d.f. (2,102). 
bBetween two patient groups, d.f. (1,71). 
cχ2 distribution between the three subject groups. See text for key to abbreviations. 
Age (years)  33.3 (8.7)  34.1 (8.5)  35.3 (8.9)  32.4 (7.8) 1.12a n.s. 
Education (years)  14.7 (3.3)  13.9 (2.4)  13.9 (2.3)  14.0 (2.5) 0.82a n.s. 
National Adult Reading Test 115.5 (8.6) 113.3 (10.1) 112.4 (10.7) 114.5 (9.2) 1.01a n.s. 
Handedness  25R 7L  54R 18L  33R 8L  21R 10L 1.81b n.s. 
Length of illness (months)  – 138.1 (94.0) 149.3 (97.9) 120.3 (87.3) 0.83c n.s. 
Years on medication  –  10.9 (13.6)  13.3 (17.1)  8.2 (7.2) 1.64 n.s. 
Mean SAPS  –  1.24 (0.90)  1.60 (0.87)  0.78 (0.74) 4.23bP < 0.001 
Mean SANS  –  1.76 (1.01)  2.25 (0.93)  1.15 (0.77) 5.39bP < 0.001 
Table 2

Overall intracranial and tissue volumes in schizophrenia patients (with and without hallucinations) and normal controls

Volume (ml) Controls [n = 32; mean (SD)] Schizophrenics Statistical comparisons 
  Total group [n = 72; mean (SD)] Hallucinators [n = 41; mean (SD)] Non-hallucinators [n = 31; mean (SD)] ANOVAa Significant contrastsb Percentage tissue lossc 
aANOVA between three subject groups, d.f. (2,101). 
bPost hoc contrasts: controls (C) versus total patient group (Sz); non-hallucinators (NH) versus C; hallucinators (H) versus C; H versus NH. Bonferroni corrected. Sz < C = schizophrenia group volume less than controls. 
cC versus Sz. 
Intracranial 1301 (97) 1236 (110) 1241 (102) 1228 (121) F = 4.19, P = 0.02 Sz < C; NH < C 
Grey matter 590 (70)  570 (71)  566 (79)  574 (61) F = 1.06, n.s. n.s. 
White matter 544 (76)  506 (73)  514 (69)  496 (79) F = 3.36, P = 0.04 Sz < C; NH < C 
CSF – sulcal and ventricular 168 (54)  160 (50)  166 (53)  152 (45) F = 0.82, n.s. n.s. – 
Volume (ml) Controls [n = 32; mean (SD)] Schizophrenics Statistical comparisons 
  Total group [n = 72; mean (SD)] Hallucinators [n = 41; mean (SD)] Non-hallucinators [n = 31; mean (SD)] ANOVAa Significant contrastsb Percentage tissue lossc 
aANOVA between three subject groups, d.f. (2,101). 
bPost hoc contrasts: controls (C) versus total patient group (Sz); non-hallucinators (NH) versus C; hallucinators (H) versus C; H versus NH. Bonferroni corrected. Sz < C = schizophrenia group volume less than controls. 
cC versus Sz. 
Intracranial 1301 (97) 1236 (110) 1241 (102) 1228 (121) F = 4.19, P = 0.02 Sz < C; NH < C 
Grey matter 590 (70)  570 (71)  566 (79)  574 (61) F = 1.06, n.s. n.s. 
White matter 544 (76)  506 (73)  514 (69)  496 (79) F = 3.36, P = 0.04 Sz < C; NH < C 
CSF – sulcal and ventricular 168 (54)  160 (50)  166 (53)  152 (45) F = 0.82, n.s. n.s. – 
Table 3

Regional proportional tissue volume differences between schizophrenia patients (combined) versus controls: grey and white matter

Cluster location Talairach coordinates of centroid (x, y,zCluster size (No. of voxels; % volume reduction) 
BA, Brodmann area. 
Grey matter deficits   
    1. Orbitofrontal cluster, extending posteriorly from the gyrus rectus into inferior portions of the anterior cingulate (probable BAs 10, 11, 32)  –2, 28, –15 231 (15%) 
    2. Right medial temporal lobe cluster, extending posteriorly from uncus, including body and tail of hippocampus and parahippocampal gyrus (BAs 28, 34)  17, –2, –25 645 (17%) 
    3. Left temporal cluster, extending posteriorly through insula and into left superior temporal gyrus (BA 22) –42, –3, 8 581 (12%) 
    4. Subgenual portion of the right anterior cingulate (BAs 24, 25), extending into anterior thalamus  3, –4, 4 161 (14%) 
    5. Left medial temporal lobe cluster, almost identical to deficit cluster 2 (BA 28) –20, –8, –15 332 (19%) 
    6. Right insula  41, –8, 3 620 (13%) 
    7. Left precuneus (BA 31)  –4, –66, 32 271 (10%) 
    8. Right lingual gyrus, extending into right precuneus (BAs 17, 31)  3, –67, 24 307 (16%) 
    9. Left lingual gyrus, extending into left precuneus (BAs 17, 31)  –3, –67, 23 222 (14%) 
Grey matter excess   
    Right inferior parietal lobe (BA 40)  61, –24, 34 362 (25%) 
White-matter deficit   
    Right frontal lobe, in the vicinity of the superior longitudinal fasciculus  24, 12, 26 189 (8%) 
Cluster location Talairach coordinates of centroid (x, y,zCluster size (No. of voxels; % volume reduction) 
BA, Brodmann area. 
Grey matter deficits   
    1. Orbitofrontal cluster, extending posteriorly from the gyrus rectus into inferior portions of the anterior cingulate (probable BAs 10, 11, 32)  –2, 28, –15 231 (15%) 
    2. Right medial temporal lobe cluster, extending posteriorly from uncus, including body and tail of hippocampus and parahippocampal gyrus (BAs 28, 34)  17, –2, –25 645 (17%) 
    3. Left temporal cluster, extending posteriorly through insula and into left superior temporal gyrus (BA 22) –42, –3, 8 581 (12%) 
    4. Subgenual portion of the right anterior cingulate (BAs 24, 25), extending into anterior thalamus  3, –4, 4 161 (14%) 
    5. Left medial temporal lobe cluster, almost identical to deficit cluster 2 (BA 28) –20, –8, –15 332 (19%) 
    6. Right insula  41, –8, 3 620 (13%) 
    7. Left precuneus (BA 31)  –4, –66, 32 271 (10%) 
    8. Right lingual gyrus, extending into right precuneus (BAs 17, 31)  3, –67, 24 307 (16%) 
    9. Left lingual gyrus, extending into left precuneus (BAs 17, 31)  –3, –67, 23 222 (14%) 
Grey matter excess   
    Right inferior parietal lobe (BA 40)  61, –24, 34 362 (25%) 
White-matter deficit   
    Right frontal lobe, in the vicinity of the superior longitudinal fasciculus  24, 12, 26 189 (8%) 
Table 4

Regional proportional tissue volume differences in hallucinators, non-hallucinators and controls: grey and white matter

Cluster location Talairach coordinates of centroid (x, y, zCluster size (No. of voxels; % volume reduction) 
Hallucinators versus controls   
    Grey matter deficits   
        1. Right medial temporal lobe cluster, extending posteriorly from uncus, predominantly through the parahippocampal gyrus, into posterior portion of the hippocampus (BAs 28, 34)  22, –1, –23 457 (18%) 
        2. Subgenual portion of the anterior cingulate (bilaterally), extending posteriorly to near anterior portion of the right thalamus (BAs 24, 25)  6, 0, –2 278 (15%) 
        3. Left insula –42, –7, 9 623 (12%) 
        4. Right insula  40, –8, 6 560 (14%) 
        5. Left precuneus, extending into lingual gyrus (BAs 7, 18, 31)  –4, –68, 16 626 (15%) 
        6. Right precuneus, extending into lingual gyrus (BAs 7, 18, 31)  5, –67, 18 431 (17%) 
    Grey matter excess   
        Right inferior parietal lobe (BA 40)  59, 26, 27 353 (25%) 
    White-matter deficit   
        Right frontal lobe, in the vicinity of the superior longitudinal fasciculus  28, 16, 26 498 (8%) 
    White-matter excess   
        Left hemisphere cluster, extending posteriorly from between hippocampus/putamen to claustrum, adjacent to insula –37, –11, –8  74 (29%) 
Non-hallucinators versus controls   
    Grey-matter deficits   
        1. Right medial temporal lobe cluster, extending posteriorly from the uncus, including the body and tail of the hippocampus and the parahippocampal gyrus (BAs 28, 34)  19, –3, –25 660 (20%) 
        2. Right temporal lobe cluster, extending posteriorly from anterior portions of the STG into the insula (BA 22)  40, –5, –8 392 (17%) 
    White-matter deficit   
        Right frontal lobe cluster  27, 20, 13 306 (15%) 
    White-matter excess   
        Right medial temporal cluster, extending length adjacent to the hippocampus  38, –25, –11 668 (21%) 
Hallucinators versus non-hallucinators   
    Grey-matter deficit   
        Left insular cortex –37, 7, –4 275 (18%) 
Cluster location Talairach coordinates of centroid (x, y, zCluster size (No. of voxels; % volume reduction) 
Hallucinators versus controls   
    Grey matter deficits   
        1. Right medial temporal lobe cluster, extending posteriorly from uncus, predominantly through the parahippocampal gyrus, into posterior portion of the hippocampus (BAs 28, 34)  22, –1, –23 457 (18%) 
        2. Subgenual portion of the anterior cingulate (bilaterally), extending posteriorly to near anterior portion of the right thalamus (BAs 24, 25)  6, 0, –2 278 (15%) 
        3. Left insula –42, –7, 9 623 (12%) 
        4. Right insula  40, –8, 6 560 (14%) 
        5. Left precuneus, extending into lingual gyrus (BAs 7, 18, 31)  –4, –68, 16 626 (15%) 
        6. Right precuneus, extending into lingual gyrus (BAs 7, 18, 31)  5, –67, 18 431 (17%) 
    Grey matter excess   
        Right inferior parietal lobe (BA 40)  59, 26, 27 353 (25%) 
    White-matter deficit   
        Right frontal lobe, in the vicinity of the superior longitudinal fasciculus  28, 16, 26 498 (8%) 
    White-matter excess   
        Left hemisphere cluster, extending posteriorly from between hippocampus/putamen to claustrum, adjacent to insula –37, –11, –8  74 (29%) 
Non-hallucinators versus controls   
    Grey-matter deficits   
        1. Right medial temporal lobe cluster, extending posteriorly from the uncus, including the body and tail of the hippocampus and the parahippocampal gyrus (BAs 28, 34)  19, –3, –25 660 (20%) 
        2. Right temporal lobe cluster, extending posteriorly from anterior portions of the STG into the insula (BA 22)  40, –5, –8 392 (17%) 
    White-matter deficit   
        Right frontal lobe cluster  27, 20, 13 306 (15%) 
    White-matter excess   
        Right medial temporal cluster, extending length adjacent to the hippocampus  38, –25, –11 668 (21%) 
Hallucinators versus non-hallucinators   
    Grey-matter deficit   
        Left insular cortex –37, 7, –4 275 (18%) 
Figure 1

Brain changes in patients with schizophrenia in comparison with normal controls. Red denotes regions of tissue loss and green denotes regions of tissue excess in patients. Results are displayed on averaged grey and white matter maps. Numbers refer to approximate y coordinates in the standard space of Talairach and Tournoux. Note that the particular slices displayed are those that showed the most significant differences and are not the same for each comparison. (A) Grey (upper) and white (lower) matter changes in all patients with schizophrenia compared to controls. (B) Grey (upper) and white (lower) matter changes in hallucinators compared to controls. (C) Grey (upper) and white (lower) matter changes in non-hallucinators compared to controls. (D) Grey matter changes in hallucinators compared to non-hallucinators. The left side of the image corresponds to the right side of the brain.

Figure 1

Brain changes in patients with schizophrenia in comparison with normal controls. Red denotes regions of tissue loss and green denotes regions of tissue excess in patients. Results are displayed on averaged grey and white matter maps. Numbers refer to approximate y coordinates in the standard space of Talairach and Tournoux. Note that the particular slices displayed are those that showed the most significant differences and are not the same for each comparison. (A) Grey (upper) and white (lower) matter changes in all patients with schizophrenia compared to controls. (B) Grey (upper) and white (lower) matter changes in hallucinators compared to controls. (C) Grey (upper) and white (lower) matter changes in non-hallucinators compared to controls. (D) Grey matter changes in hallucinators compared to non-hallucinators. The left side of the image corresponds to the right side of the brain.

References

Akbarian S, Kim JJ, Potkin SG, Hetrick WP, Bunney WE Jr, Jones EG (
1996
) Maldistribution of interstitial neurons in prefrontal white matter of the brains of schizophrenic patients.
Arch Gen Psychiatry
 
53
:
425
–436.
Andreasen NC, Olsen S (
1982
) Negative vs positive schizophrenia: definition and validation.
Arch Gen Psychiatry
 
39
:
789
–794.
Andreasen NC, Arndt S, Swayze V, Cizadlo T, Flaum M, O’Leary D, Ehrhardt JC, Yuh WTC (
1994
) Thalamic abnormalities in schizophrenia visualized through magnetic resonance image averaging.
Science
 
266
:
294
–298.
Andreasen NC, Flashman L, Flaum M, Arndt S, Swayze V, O’Leary D, Ehrhardt JC, Yuh WTC (
1994
) Regional brain abnormalities in schizophrenia measured with magnetic resonance imaging.
J Am Med Assoc
 
272
:
1763
–1769.
van Amelsvoort T, Daly E, Robertson D, Suckling J, Ng V, Critchley H, Owen MJ, Henry J, Murphy KC, Murphy DG (
2001
) Structural brain abnormalities associated with deletion at chromosome 22q11: quantitative neuroimaging study of adults with velo-cardio-facial syndrome.
Br J Psychiatry
 
178
:
412
–419.
Ashburner J, Friston K (
1997
) Multimodal image coregistration and partitioning-a unified framework.
Neuroimage
 
6
:
209
–217.
Ashburner J, Friston KJ (
2000
) Voxel-based morphometry — the methods.
Neuroimage
 
11
:
805
–821.
Ashburner J, Friston KJ (
2001
) Why voxel-based morphometry should be used.
Neuroimage
 
14
:
1238
–1243.
Augustine JR (
1985
) The insular lobe in primates including humans.
Neurol Res
 
7
:
2
–9.
Augustine JR (
1996
) Circuitry and functional aspects of the insular lobe in primates including humans.
Brain Res Rev
 
22
:
229
–244.
Barta PE, Pearlson GD, Powers RE, Richards SS, Tune LE (
1990
) Auditory hallucinations and smaller superior temporal gyral volume in schizophrenia.
Am J Psychiatry
 
147
:
457
–462.
Bookstein FL (
2001
) ‘Voxel-based morphometry’ should not be used with imperfectly registered images.
Neuroimage
 
14
:
1454
–1462.
Brammer MJ, Bullmore ET, Simmons A, Williams SC, Grasby PM, Howard RJ, Woodruff PW, Rabe-Hesketh S (
1997
) Generic brain activation mapping in functional magnetic resonance imaging: a nonparametric approach.
Magn Reson Imaging
 
15
:
763
–770.
Buchsbaum MS, Tang CY, Peled S, Gudbjartsson H, Lu D, Hazlett EA, Downhill J, Haznedar M, Fallon JH, Atlas SW (
1998
) MRI white matter diffusion anisotropy and PET metabolic rate in schizophrenia.
Neuroreport
 
9
:
425
–430.
Buchanan RW, Vladar K, Barta PE, Pearlson GD (
1998
) Structural evaluation of the prefrontal cortex in schizophrenia.
Am J Psychiatry
 
155
:
1049
–1055.
Bullmore E, Brammer M, Rouleau G, Everitt B, Simmons A, Sharma T, Frangou S, Murray RM, Dunn G (
1995
) Computerized brain tissue classification of magnetic resonance images: a new approach to the problem of partial volume artifact.
Neuroimage
 
2
:
133
–147.
Bullmore ET, Suckling OS, Rabe-Hesketh S, Taylor E, Brammer MJ (
1999
) Global, voxel and cluster tests, by theory and permutation for a difference between two groups of structural MR images of the brain.
IEEE Trans Med Imag
 
18
:
32
–42.
Calvert GA, Brammer MJ, Bullmore ET, Campbell R, Iversen SD, David AS (
1999
) Response amplification in sensory-specific cortices during crossmodal binding.
Neuroreport
 
10
:
2619
–2623.
Cannon TD, van Erp TG, Huttunen M, Lonnqvist J, Salonen O, Valanne L, Poutanen VP, Standeriskjold-Nordenstam CG, Gur RE, Yan M (
1998
) Regional gray matter, white matter, and cerebrospinal fluid distributions in schizophrenic patients, their siblings, and controls.
Arch Gen Psychiatry
 
55
:
1084
–1091.
Chakos MH, Lieberman JA, Bilder RM, Borenstein M, Lerner G, Bogerts B, Wu H, Kinon B, Ashtari M (
1994
) Increase in caudate nuclei volumes of first-episode schizophrenic patients taking antipsychotic drugs.
Am J Psychiatry
 
151
:
1430
–1436.
Chua S, Wright I, Murray RM, Friston KJ, Liddle PF, McGuire PK (
1997
) Grey matter correlates of syndromes in schizophrenia: a semiautomated analysis of structural magnetic resonance images.
Br J Psychiatry
 
170
:
406
–410.
Cox RW (1994) Analysis of functional neuroimages, v. 1.01. Medical College of Wisconsin.
Crespo-Facorro B, Kim JJ, Andreasen NC, O’Leary DS, Bockholt J, Magnotta V (
2000
) Insular cortex abnormalities in schizophrenia: a structural magnetic resonance imaging study of first-episode patients.
Schizophr Res
 
46
:
35
–43.
David AS (1994a) The neuropsychology of auditory–verbal hallucinations. In: The neuropsychology of schizophrenia (David A, Cutting J, eds), pp. 269–312. Hove: Lawrence Erlbaum.
David AS (
1994
) Dysmodularity: a neurocognitive model for schizophrenia.
Schizophr Bull
 
120
:
249
–255.
David A, Woodruff PWR, Howard R, Mellers JDC, Brammer M, Bullmore E, Wright I, Andrew C, Williams SCR (
1996
) Auditory hallucinations inhibit exogenous activation of auditory association cortex.
Neuroreport
 
7
:
932
–936.
DeQuardo JR, Keshavan MS, Bookstein FL, Bagwell WW, Green WD, Sweeney JA, Haas GL, Tandon R, Schooler NR, Pettegrew JW (
1999
) Landmark-based morphometric analysis of first-episode schizophrenia.
Biol Psychiatry
 
45
:
1321
–1328.
Dierks T, Linden DEJ, Jandl M, Formisano E, Goebel R, Lanfermann H, Singer W (
1999
) Activation of Heschl’s gyrus during auditory hallucinations.
Neuron
 
22
:
615
–621.
Dronkers NF (
1996
) A new brain region for coordinating speech articulation.
Nature
 
384
:
159
–161.
Ellis CM, Suckling J, Amaro E Jr, Bullmore ET, Simmons A, Williams SC, Leigh PN (
2001
) Volumetric analysis reveals corticospinal tract degeneration and extramotor involvement in ALS.
Neurology
 
57
:
1571
–1578.
Flaum M, O’Leary DS, Swayze VW, Miller DD, Arndt S, Andreasen N (
1995
) Symptom dimensions and brain morphology in schizophrenia and related psychotic disorders.
J Psychiatr Res
 
29
:
261
–276.
Foong J, Maier M, Barker GJ, Brocklehurst S, Miller DH, Ron MA (
2000
) In vivo investigation of white matter pathology in schizophrenia with magnetisation transfer imaging.
J Neurol Neurosurg Psychiatry
 
68
:
70
–74.
Friston KJ, Jezzard P, Turner R (
1994
) Analysis of functional MRI time-series.
Hum Brain Mapp
 
1
:
153
–171.
Gaser C, Volz HP, Kiebel S, Riehemann S, Sauer H (
1999
) Detecting structural changes in whole brain based on nonlinear deformations — application to schizophrenia research.
Neuroimage
 
10
:
107
–113.
Goldstein JM, Goodman JM, Seidman LJ, Kennedy DN, Makris N, Lee H, Tourville J, Caviness VS, Faraone SV, Tsuang MT (
1999
) Cortical abnormalities in schizophrenia identified by structural magnetic resonance imaging.
Arch Gen Psychiatry
 
56
:
537
–547.
Good CD, Johnsrude IS, Ashburner J, Henson RN, Friston KJ, Frackowiak RS (
2001
) A voxel-based morphometric study of ageing in 465 normal adult human brains.
Neuroimage
 
14
:
21
–36.
Good CD, Johnsrude I, Ashburner J, Henson RN, Friston KJ, Frackowiak RS (
2001
) Cerebral asymmetry and the effects of sex and handedness on brain structure: a voxel-based morphometric analysis of 465 normal adult human brains.
Neuroimage
 
14
:
685
–700.
Gitelman DR, Ashburner J, Friston KJ, Tyler LK, Price CJ (
2001
) Voxel-based morphometry of herpes simplex encephalitis.
Neuroimage
 
13
:
623
–31.
Günther W, Petsch R, Steinberg R, Moser E, Streck P, Heller H, Kurtz G, Hippins H (
1991
) Brain dysfunction during motor activation and corpus callosum alterations in schizophrenia measured by cerebral blood flow and magnetic resonance imaging.
Biol Psychiatry
 
29
:
535
–555.
Gur RE, Cowell PE, Latshaw A, Turetsky BI, Grossman RI, Arnold SE, Bilker WB, Gur RC (
2000
) Reduced dorsal and orbital prefrontal gray matter volumes in schizophrenia.
Arch Gen Psychiatry
 
57
:
761
–768.
Gur RE, Turetsky BI, Cowell PE, Finkelman C, Maany V, Grossman RI, Arnold SE, Bilker WB, Gur RC (
2000
) Temporolimbic volume reductions in schizophrenia.
Arch Gen Psychiatry
 
57
:
769
–775.
Harris IM, Egan GF, Sonkkila C, Tochon-Danguy HJ, Paxinos G, Watson JD (
2000
) Selective right parietal lobe activation during mental rotation: a parametric PET study.
Brain
 
123
:
65
–73.
Hulshoff-Pol HE, Schnack HG, Mandl RCW, van Haren NEM, Koning H, Collins L, Evans AC, Kahn RS (
2001
) Focal gray matter density changes in schizophrenia.
Arch Gen Psychiatry
 
58
:
1118
–1125.
Keller SS, Mackay CE, Barrick TR, Wieshmann UC, Howard MA, Roberts N (
2002
) Voxel-based morphometric comparison of hippocampal and extrahippocampal abnormalities in patients with left and right hippocampal atrophy
Neuroimage
 
16
:
23
–31.
Keshavan MS, Bagwell WW, Haas GL, Sweeney JA, Schooler NR, Pettegrew JW (
1994
) Changes in caudate volume with neuroleptic treatment.
Lancet
 
344
:
1434
.
Landis T (
2000
) Disruption of space perception due to cortical lesions.
Spatial Vision
 
13
:
179
–191.
Lawrie SM, Abukmeil SS (
1998
) Brain abnormality in schizophrenia. A systematic and quantitative review of volumetric magnetic resonance imaging studies.
Br J Psychiatry
 
172
:
110
–120.
Levitan C, Ward PB, Catts SV (
1999
) Superior temporal gyral volumes and laterality correlates of auditory hallucinations in schizophrenia.
Biol Psychiatry
 
46
:
955
–962.
Liddle PF, Friston KJ, Frith CD, Hirsch SR, Jones T, Frackowiak RSJ (
1992
) Patterns of cerebral blood flow in schizophrenia.
Br J Psychiatry
 
160
:
179
–186.
Lim KO, Hedehus M, Moseley M, de Crespigny A, Sullivan EV, Pfefferbaum A (
1999
) Compromised white matter tract integrity in schizophrenia inferred from diffusion tensor imaging.
Arch Gen Psychiatry
 
56
:
367
–374.
McCarley RW, Wible CG, Frumin M, Hirayasu Y, Levitt JJ, Fischer IA, Shenton ME (
1999
) MRI anatomy of schizophrenia.
Biol Psychiatry
 
45
:
1099
–1119.
McGuire PK, Frith CD (
1996
) Disordered functional connectivity in schizophrenia.
Psychol Med
 
26
:
663
–667.
McGuire PK, Silbersweig DA, Murray RM, David AS, Frackowiak RSJ, Frith CD (
1996
) The functional anatomy of inner speech and auditory imagery.
Psychol Med
 
26
:
29
–38.
Mega MS, Cummings JL, Salloway S, Malloy P (
1997
) The limbic system: an anatomic phylogenetic, and clinical perspective.
J Neuropsychiatry Clin Neurosci
 
9
:
315
–330.
Mesulam M-M (1985) Patterns in behavioural neuroanatomy: association areas, the limbic system, and hemispheric specialization. In: Behavioral neurology (Mesulam M-M, ed.), pp. 1–70. Philadelphia, PA: FA Davis.
Mummery C, Patterson K, Price C, Ashburner J, Frackowiak R, Hodges J (
2000
) A voxel-based morphometry study of semantic dementia: relationship between temporal lobe atrophy and semantic memory.
Ann Neurol
 
47
:
36
–45.
Nasrallah HA (
1985
) The integrated right cerebral hemispheric consciousness as alien intruder.
Comp Psychiatry
 
26
:
273
–282.
Nelson MD, Saykin AJ, Flashman LA, Riordan HJ (
1998
) Hippocampal volume reduction in schizophrenia as assessed by magnetic resonance imaging: a meta-analytic study.
Arch Gen Psychiatry
 
55
:
433
–440.
Overmeyer S, Bullmore ET, Suckling J, Simmons A, Williams SC, Santosh PJ, Taylor E (
2001
) Distributed grey and white matter deficits in hyperkinetic disorder: MRI evidence for anatomical abnormality in an attentional network.
Psychol Med
 
31
:
1425
–1435.
Paillère-Martinot ML, Caclin A, Artiges E, Poline JB, Joliot M, Mallet L, Recasens C, Attar-Lévy D, Martinot J-L (
2001
) Cerebral gray and white matter reductions and clinical correlates in patients with early onset schizophrenia.
Schizophr Res
 
50
:
19
–26.
Phillips ML, Young AW, Senior C, Brammer M, Andrew C, Calder AJ, Bullmore ET, Perrett DI, Rowland D, Williams SCR, Gray JA, David AS (
1997
) A specific neural substrate for perceiving facial expressions of disgust.
Nature
 
389
:
495
–498.
Ploghaus A, Tracey I, Gati JS, Clare S, Menon RS, Matthews PM, Rawlins JNP (
1999
) Dissociating pain from its anticipation in the human brain.
Science
 
284
:
1979
–1981.
Poline J (
1997
) Combining spatial extent and peak intensity to test for activation in functional imaging.
Neuroimage
 
5
:
83
–96.
Poline JB, Mazoyer BM (
1993
) Analysis of individual positron emission tomography activation maps by detection of high signal-to-noise pixel clusters.
J Cereb Blood Flow Metab
 
13
:
425
–437.
Press WH, Teukolsky SA, Vetterling WT, Flannery BP (1992) Numerical recipes in C: the art of scientific computing. Cambridge: Cambridge University Press.
Rabe-Hesketh S, Bullmore ET, Brammer MJ (
1997
) Analysis of functional magnetic resonance images.
Statist Meth Med Res
 
6
:
215
–237.
Rajarethinam RP, DeQuardo JR, Napela R, Tandon R (
2000
) Superior temporal gyrus in schizophrenia: a volumetric magnetic resonance imaging study.
Schizophr Res
 
4
:
303
–312.
Rossell SL, Shapleske J, Fukuda R, Woodruff PWR, Simmons A, David AS (
2001
) Corpus callosum area and functioning in schizophrenic patients with auditory–verbal hallucinations.
Schizophr Res
 
50
:
9
–17.
Ridler K, Bullmore ET, De Vries PJ, Suckling J, Barker GJ, Meara SJ, Williams SC, Bolton PF (
2001
) Widespread anatomical abnormalities of grey and white matter structure in tuberous sclerosis.
Psychol Med
 
31
:
1437
–1446.
Sanfilipo M, Lafargue T, Rusinek H, Arena L, Loneragan C, Lautin A, Feiner D, Rotrosen J, Wolkin A (
2000
) Volumetric measure of the frontal and temporal lobe regions in schizophrenia: relationship to negative symptoms.
Arch Gen Psychiatry
 
57
:
471
–480.
Sauer H, Volz H-P, Gaser C (
2000
) Whole brain analysis of structural changes in schizophrenia.
Schizophr Res
 
41
:
5
.
Shapleske J, Rossell SL, Woodruff PWR, David AS (
1999
) The planum temporale: a systematic, quantitative review of its structural, functional and clinical significance.
Brain Res Rev
 
29
:
26
–49.
Shapleske J, Rossell SL, Simmons A, David AS, Woodruff PWR (
2001
) Are auditory hallucinations the consequence of abnormal cerebral lateralisation? A morphometric MRI study of the Sylvian fissure and planum temporale.
Biol Psychiatry
 
49
:
685
–693.
Shenton ME, Kikinis R, Jolesz FA, Pollack SD, LeMay M, Wible CG, Hokama H, Martin J, Metcalf D, Coleman M, McCarley RW (
1992
) Abnormalities of the left temporal lobe and thought disorder in schizophrenia: a quantitative magnetic resonance imaging study.
N Engl J Med
 
327
:
604
–612.
Sigmundsson T, Suckling J, Maier M, Williams SCR, Bullmore ET, Greenwood K, Fukuda R, Ron MA, Toone BK (
2001
) Structural abnormalities in frontal, temporal and limbic regions and interconnecting white matter tracts in schizophrenia.
Am J Psychiatry
 
158
:
234
–243.
Simmons A, Arridge SR, Barker GJ, Williams SCR (
1996
) Simulation of MRI cluster plots and application to neurological segmentation.
Magn Reson Imaging
 
14
:
73
–92.
Sowell ER, Levitt J, Thompson PM, Holmes CJ, Blanton RE, Kornsand DS, Caplan R, McCracken J, Asarnow R, Toga AW (
2000
) Brain abnormalities in early-onset schizophrenia spectrum disorder observed with statistical parametric mapping of structural magnetic resonance images.
Am J Psychiatry
 
157
:
1475
–1484.
Stephane M, Barton S, Boutros NN (
2001
) Auditory verbal hallucinations and dysfunction of the neural substrates of speech.
Schizophr Res
 
50
:
61
–78.
Suckling J, Brammer MJ, Lingford Hughes A, Bullmore ET (
1999
) Removal of extra cerebral tissues in dual-echo magnetic resonance images via linear scale-space features.
Magn Reson Imaging
 
17
:
247
–256.
Suckling J, Sigmundsson T, Greenwood K, Bullmore ET (
1999
) A modified fuzzy clustering algorithm for operator independent brain tissue classification of dual echo MR images.
Magn Reson Imaging
 
17
:
1065
–1682.
Sullivan EV, Lim KO, Mathalon D, Marsh L, Beal DM, Harris D, Hoff AL, Faustman WO, Pfefferbaum A (
1998
) A profile of cortical gray matter volume deficits characteristic of schizophrenia.
Cereb Cortex
 
8
:
117
–124.
Surguladze SA, Calvert GA, Brammer MJ, Campbell R, Bullmore ET, Giampietro V, David AS (
2001
) Audio-visual speech perception in schizophrenia: an fMRI study.
Psychiatry Res Neuroimaging
 
106
:
1
–14.
Thompson PM, Cannon TD, Narr KL, van Erp T, Poutanen V-P, Huttunen M, Lönnquist J, Standertskjöld-Nordenstam C-G, Kaprio J, Khaledy M, Dail R, Zoumalan CI, Toga AW (
2001
) Genetic influences on brain structure.
Nat Neurosci
 
4
:
1253
–1258.
Talairach J, Tournoux P (1988) Co-planar stereotaxic atlas of the human brain. Three-dimensional proportional system. New York: Thieme Medical.
Türe U, Yasargil DCH, Al-Mefty O, Yasargil MG (
1999
) Topographic anatomy of the insular region.
J Neurosurg
 
90
:
720
–733.
Watkins KE, Vargha-Khadem F, Ashburner J, Passingham RE, Connelly A, Friston KJ, Frackowiak RS, Mishkin M, Gadian DG (
2002
) MRI analysis of an inherited speech and language disorder: structural brain abnormalities
Brain
 
125
:
465
–478.
Wilke M, Kaufmann C, Grabner A, Pütz B, Wetter TC, Auer DP (
2001
) Gray matter-changes and correlates of disease severity in schizophrenia: a statistical parametric mapping study.
Neuroimage
 
13
:
814
–824.
Wolkin A, Rusinek H, Vaid G, Arena L, Lafargue T, Sanfilipo M, Loneragan C, Lautin A, Rotrosen J (
1998
) Structural magnetic resonance image averaging in schizophrenia.
Am J Psychiatry
 
155
:
1064
–1073.
Woodruff P, Brammer M, Mellers J, Wright I, Bullmore E, Williams S (
1995
) Auditory hallucinations and perception of external speech.
Lancet
 
346
:
1035
.
Woodruff PWR, McManus C, David A (
1995
) A meta-analysis of corpus callosum size in schizophrenia.
J Neurol Neurosurg Psychiatry
 
58
:
457
–461.
Woodruff PWR, Wright IC, Bullmore ET, Brammer M, Howard RJ, Williams SCR, Shapleske J, Rossell S, David AS, McGuire PK, Murray RM (
1997
) Auditory hallucinations and the temporal cortical response to speech in schizophrenia: a functional magnetic resonance imaging study.
Am J Psychiatry
 
154
:
1676
–1682.
Wright IC, McGuire PK, Poline J-B, Travere JM, Murray RM, Frith CD, Frackowiak RSJ, Friston KJ (
1995
) A voxel-based method for the statistical analysis of grey and white matter density applied to schizophrenia.
Neuroimage
 
2
:
244
–252.
Wright I, Ellison Z, Sharma T, Friston K, Murray RM, McGuire PK (
1999
) Mapping of grey matter changes in schizophrenia.
Schizophr Res
 
35
:
1
–14.
Wright IC, Rabe-Hesketh S, Woodruff PWR, David AS, Murray RM, Bullmore ET (
2000
) Meta-analysis of regional brain volumes in schizophrenia.
Am J Psychiatry
 
157
:
16
–25.