Disrupted Axonal Fiber Connectivity in Schizophrenia · 2011. 5. 23. · studies, and theoretical neurobiology (9–20) further suggest that abnormal function in these brain structures

DAG

Bbc

Mnpatw

Rpnc

Cpa

Kw

Acmt(sadCdadc

ldmbg

F

A

R

0d

isrupted Axonal Fiber Connectivity in Schizophreniandrew Zalesky, Alex Fornito, Marc L. Seal, Luca Cocchi, Carl-Fredrik Westin, Edward T. Bullmore,ary F. Egan, and Christos Pantelis

ackground: Schizophrenia is believed to result from abnormal functional integration of neural processes thought to arise from aberrantrain connectivity. However, evidence for anatomical dysconnectivity has been equivocal, and few studies have examined axonal fiberonnectivity in schizophrenia at the level of whole-brain networks.

ethods: Cortico-cortical anatomical connectivity at the scale of axonal fiber bundles was modeled as a network. Eighty-two networkodes demarcated functionally specific cortical regions. Sixty-four direction diffusion tensor-imaging coupled with whole-brain tractogra-hy was performed to map the architecture via which network nodes were interconnected in each of 74 patients with schizophrenia and 32ge- and gender-matched control subjects. Testing was performed to identify pairs of nodes between which connectivity was impaired inhe patient group. The connectional architecture of patients was tested for changes in five network attributes: nodal degree, small-orldness, efficiency, path length, and clustering.

esults: Impaired connectivity in the patient group was found to involve a distributed network of nodes comprising medial frontal,arietal/occipital, and the left temporal lobe. Although small-world attributes were conserved in schizophrenia, the cortex was intercon-ected more sparsely and up to 20% less efficiently in patients. Intellectual performance was found to be associated with brain efficiency inontrol subjects but not in patients.

onclusions: This study presents evidence of widespread dysconnectivity in white-matter connectional architecture in a large sample ofatients with schizophrenia. When considered from the perspective of recent evidence for impaired synaptic plasticity, this study points to

ey Words: Axonal connectivity, diffusion tensor imaging, net-ork, schizophrenia, tractography, white-matter

bnormal brain function in schizophrenia involves a distrib-uted network of spatially separated but functionally relatedbrain structures, including the frontal, temporal and parietal

ortices, the basal ganglia, cerebellum, hippocampus, and thala-us (1). Studies modeling the brain as a network (2–5), investiga-

ions of the default-mode network in patients with schizophrenia6 – 8), together with functional neuroimaging, electrophysiologicaltudies, and theoretical neurobiology (9 –20) further suggest thatbnormal function in these brain structures might be coupled toisturbances within the functional circuits via which they interact.urrent theories therefore posit that the cognitive and behavioralisturbances of schizophrenia are an expression of the brain’s in-bility to properly integrate neural processes segregated acrossistributed regions, a state which has been termed functional dys-onnectivity (18,21).

A shortcoming of functional neuroimaging and electrophysio-ogical studies is that they cannot determine whether functionalysconnectivity stems from aberrant axonal connectivity or abnor-al neuronal function within the cortical regions interconnected

y a functional circuit. The former refers to axonal pathophysiolo-ies like demyelinative processes and diminished oligodendrocyte

rom the Melbourne Neuropsychiatry Centre (AZ, AF, MLS, LC, CP); FloreyNeuroscience Institutes (GFE, CP); Centre for Neuroscience (GFE), TheUniversity of Melbourne, Melbourne, Australia; Behavioural and ClinicalNeuroscience Institute (AF, ETB), University of Cambridge, Cambridge,United Kingdom; and the Department of Radiology (C-FW), Brigham andWomen’s Hospital, Harvard Medical School, Boston, Massachusetts.

ddress correspondence to Andrew Zalesky, Ph.D., The University of Mel-bourne and Melbourne Health, Melbourne Neuropsychiatry Centre,Level 3, Alan Gilbert Building, 161 Barry Street, Melbourne, Victoria VIC,Australia; E-mail: [email protected].

eceived May 10, 2010; revised Aug 17, 2010; accepted Aug 18, 2010.

006-3223/$36.00oi:10.1016/j.biopsych.2010.08.022

function (22,23) as well as the possible “miswiring” of axonal fiberbundles during brain maturation, whereas the latter specificallyrefers to aberrant synaptic transmission and plasticity (13,18,24,35).

The former can be readily assessed with modern, noninvasiveimaging techniques; namely, diffusion-weighted imaging (26) andmagnetization transfer imaging (27). However, findings yielded bythe application of these techniques to schizophrenia have beenequivocal, with a recent meta-analysis indicating that, for any re-gion where a positive finding was reported, at least as many studiesfailed to find one (1,28,29). This has led some theorists to posit thatwhite-matter connectional architecture might be relatively pre-served in schizophrenia, and the primary pathophysiology might liein impaired synaptic plasticity (18,21,25,30).

One difficulty with studies in this field is that most of them havenot directly investigated axonal connectivity per se. Instead, re-gions have been searched for which myelin, axonal, or other patho-physiological abnormalities are suspected, with the assumptionthat axonal fibers traversing the vicinity of these regions are im-paired, thus suggesting a “dysconnection” between the gray-mat-ter regions that they interconnect (31).

The principal aim of this study was to test whether white-matterconnectional architecture demonstrates dysconnectivity in schizo-phrenia at a macroscale (i.e., at the scale of axonal fiber bundles) bymapping a network model of cortico-cortical connectivity. In par-ticular, diffusion-weighted imaging followed by whole-brain trac-tography was performed to derive a network (graph) model ofmacroscale cortico-cortical connectivity for each of 74 patients withchronic schizophrenia and 32 control subjects. Each network nodeencapsulated a distinct gray-matter region, and pairs of nodes werelinked if they were interconnected via an axonal fiber (3,32–38). Anetwork-based statistic (NBS) (Methods and Materials) was thenused to identify impaired network connections in the patientgroup. Attributes (39) of the resulting network models were alsoinvestigated, enabling determination of whether disordered net-work organization was evident in schizophrenia. With these tech-

BIOL PSYCHIATRY 2011;69:80–89© 2011 Society of Biological Psychiatry

itt

M

S

pspCiwcggaops

gicyc.1teyrfdwBwpocNwr

AI

bAtm

lumTmdtmwt

A. Zalesky et al. BIOL PSYCHIATRY 2011;69:80–89 81

ty in schizophrenia is not exclusive to abnormal synapticransmission and plasticity but also involves altered axonal connec-ivity as a determinant.

ethods and Materials

ample CharacteristicsSeventy-four patients with schizophrenia were recruited as

art of the Australian Schizophrenia Research Bank (http://www.chizophreniaresearch.org.au). The study was approved by an inde-endent ethics committee (Melbourne Health Research Ethicsommittee, 2006: 061), and all participants were able to provide

nformed consent. A diagnosis of schizophrenia was confirmedith the diagnostic interview for psychoses (40). Thirty-two healthy

ontrol subjects, statistically matched (� � .05) for average age,ender, and handedness were also recruited from similar geo-raphic and demographic regions [age of patients: 38 � 11 years,ge of control subjects: 33 � 13 years, t (49.9) � 1.9, p � .06; genderf patients: 47 men, gender of control subjects: 15 men, �2(1) � 2.5,� .1; handedness of patients: 62 right, handedness of control

ubjects: 26 right, �2(1) � 1.0, p � .4].The patient sample demonstrated a lower premorbid intelli-

ence quotient (IQ), estimated with the Wechsler test of adult read-ng (41), and fewer years of education (IQ of patients: 102 � 13, IQ ofontrol subjects: 110 � 9, p � .01; education of patients: 13 � 13ears, education of control subjects: 16 � 3 years, p � .01). Otherharacteristics of the patient sample were age of illness onset (23 �

3 years) and global assessment of functioning (GAF) score (61% �1%); 90% of patients were on a course of antipsychotic medica-ion, predominantly olanzapine or quetiapine. Participants werexcluded if they had: a history of organic brain disorder; wereounger than 16 or older than 65 years; had serious brain injuryesulting in posttraumatic amnesia for more than 24 h; or hadull-scale IQ � 70, movement disorder, or a current diagnosis ofrug or alcohol dependence. Furthermore, control participantsere excluded if they had a relative with a history of psychosis oripolar I Disorder or a personal history of any mental illness forhich they had been hospitalized or received treatment. Partici-ants were not excluded if they had a concurrent non-neurologicalr nonpsychiatric diagnosis. Participants were recruited from theommunity through television, newspaper, and radio advertising.o incidental findings (including white-matter hyperintensities)ere noted by the radiologist in any of the structural magnetic

esonance imaging scans acquired in each subject.

cquisition and Preprocessing of Magnetic Resonancemaging Data

A series of diffusion-weighted magnetic resonance images ofrain anatomy were acquired in each subject with a Siemensvanto 1.5-Tesla system (Siemens, Erlangen, Germany) located at

he Royal Children’s Hospital, Melbourne, Australia (see Supple-ent 1 for acquisition parameters).

In brief, the overall preprocessing pipeline comprised the fol-owing steps: rigid-body registration of diffusion-weighted vol-mes to baseline T2-weighted volume (to correct for slight headotion); stripping of skull and other noncerebral material from

2-weighted volume; tissue classification of gray-matter, white-atter, and cerebrospinal fluid; fitting and eigen-decomposition of

iffusion tensor with weighted linear least squares (42); computa-ion of fractional anisotropy (FA) volume (43); and nonlinear nor-

alization of FA volume to Montreal Neurological Institute spaceith the FMRIB58 FA image (http://www.fmrib.ox.ac.uk/fsl/) as the

arget. This preprocessing pipeline was repeated for each subject

identically. Rigid-body registration, skull stripping, and nonlinearwarping was performed with FSL software (http://www.fmrib.ox-.ac.uk/fsl/).

Fiber Tract Network ModelTractography (44 – 47) was performed in each subject to gener-

ate three-dimensional curves characterizing neural fiber tract con-nectivity. Tens of thousands of such curves, or streamlines, weregenerated to etch out all major white-matter tracts (34 –36,38).Details of the tractography algorithm are presented in Supplement1. Tractographic maps are shown in Figure 1 for the patient andcontrol group.

Corticocortical fiber tract connectivity was modeled as a net-work, or graph, comprising a total of 82 nodes. Each node encapsu-lated a distinct gray-matter region, and pairs of nodes were joinedby a link if they were interconnected via a sufficient number ofstreamlines (see Supplement 1 for details). Note that this kind ofgraph modeling of whole-brain connectivity derived from diffu-sion-weighted imaging has been previously demonstrated inhealthy control subjects (32–38,48 –50). Corresponding graphmodels of anatomical connectivity derived from cortical thicknessand volumetric measurements have been studied in patients withschizophrenia (2), Alzheimer’s disease (51), and multiple sclero-sis (52).

Network OrganizationTo characterize white-matter connectional architecture, five key

measures describing specific attributes of network organizationwere considered: nodal degree, efficiency, path length, clusteringcoefficient, and small-worldness (53). In brief, the degree of a node isthe total number of connections it forms with other nodes of thenetwork, the path length between two nodes is the minimum num-ber of links that must be traversed to establish a connection, effi-ciency is the inverse of path length and is associated with how wella network supports parallel information transfer, clustering is a mea-sure of the efficiency of local information transfer between neigh-boring nodes, and finally, a network is said to be small-world if itjointly offers long-range and local efficiency (i.e., efficient corticalintegration as well as segregation). The extent to which a network issmall-world was quantified with the �-ratio. These five measuresare discussed in detail in Supplement 1.

Between-group differences in each measure were investigatedfor a range of binarizing thresholds. A binarizing threshold of Tindicates that links were only modeled between pairs of nodesinterconnected by more than T streamlines.

The significance of any difference was determined with permu-tation testing (54) with randomized groupings of control subjectsand patients. A one-tailed test of the null hypothesis was performedwith the empirical null distribution.

Spearman’s rank correlation was computed to assess the signif-icance of any linear associations between variables describing thesample characteristics (i.e., age, intelligence quotient, years of edu-cation and GAF score) and measures of network organization (i.e.,network efficiency, path length, clustering coefficient, �-ratio andnodal degree).

While both groups were statistically matched for age, the pa-tient group was on average 5 years older. Therefore, to investigateany potential age-related effects, the entire analysis was repeatedwith age included as a nuisance linear covariate.

Impaired ConnectionsStatistical testing was performed to identify any pairs of nodes

(cortical regions) that were connected in control subjects but dem-

www.sobp.org/journal

obscfsr

niaal

jactfi

I

cispbemm

Fsship emisb ection

82 BIOL PSYCHIATRY 2011;69:80–89 A. Zalesky et al.

w

nstrated impaired connectivity in patients. The connectivityetween a pair of nodes was quantified by the total number oftreamlines that interconnected them. Furthermore, a particularonnection was said to be impaired if it comprised significantlyewer streamlines in patients than in control subjects (see Discus-ion for suggestions on the pathophysiological significance of aeduced streamline count).

Statistical testing was performed on the subset of all pairs ofodes that were connected on average by at least Kavg streamlines

n the control group. This constraint ensured a connection was notssumed between a pair of nodes that happened to be connecteds a matter of chance by one or two potentially spurious stream-

ines.The null hypothesis of equal connectivity between control sub-

ects and patients (i.e., equal streamline count) was considered forll node pairs satisfying the Kavg threshold. A one-tailed p value wasalculated independently for each hypothesis with permutationesting (5000 permutations), as described in the preceding text. Thealse discovery rate (FDR) was used to correct for multiple compar-sons (55).

mpaired NetworksThe NBS (56) was used to identify and ascribe significance to any

onnected subnetworks evident in the set of impaired links foundn the Impaired Connections section of this text. The NBS is de-cribed in Supplement 1. The NBS was used with the followingarameter settings: Kavg � 2, the primary threshold for each link-ased t statistic was set to 1.5, and only cortico-cortical streamlinesxceeding 20 mm in length were used to generate the connectivityatrix for each subject. Figure 1 shows an overview of the complete

0 20 40 60 80

Schizophrenia

Controls

I. Tractography II. Connectivity matrix I

igure 1. Overview of methodology. (I) The tractographic maps show the sepace). II. Connectivity matrices were populated and (III) binarized, wheretreamlines) between node pair (i,j). The rows/columns of the connectivitemisphere nodes occupied the first 41 rows/columns. Hence, the two str

ntrahemispheric connections, whereas the two off-diagonal sub-blocks inrofile indicated by the connections along the main diagonal of each interhased statistic (Methods and Materials) was used to identify impaired conn

ethodology used in this study.

ww.sobp.org/journal

Results

Anomalous Network OrganizationAnomalous network organization was investigated for graph

binarizing thresholds in the range T � 0,1, . . . . 8. Thresholds exceed-ing T � 8 resulted in highly fragmented graphs and were thereforenot considered.

Average nodal degree and network efficiency were significantlyreduced in patients for all thresholds considered (Figure 2). Themaximum decrease in nodal degree was 10% and found at T � 8,whereas the maximum difference in efficiency was 22% and alsofound at T � 8.

Note that efficiency and nodal degree are not independentmeasures; graphs of higher degree typically offer greater efficiency,simply because they comprise more links. Therefore, a likely con-tributing factor to the reduction in efficiency found in patients wasthe fact that their cortices were more sparsely interconnected (i.e.,lower nodal degree).

To attribute the observed change in global efficiency to partic-ular cortical regions, regional network efficiency (57,58) was calcu-lated individually for each node. Although several parietal, occipi-tal, and frontal nodes were found to show a trend towarddiminished efficiency, no such difference survived FDR correction.

Both patients and control subjects were found to demonstratesmall-world topologies. In particular, both groups demonstratedclustering coefficients that were considerably larger than degree-matched random graphs (� 8� greater for T � 4) (Figure 2),whereas path lengths were only marginally longer (� 1.5� longerfor T � 4). Therefore, the optimal balance between local special-ization and global integration (59) arising from the small-worldnature of anatomical connectivity seems to be conserved in

1 0

Impaired connectionsp = 0.021, corrected

narized IV. Network-based statistic

l cortico-cortical streamlines (normalized to Montreal Neurological Instituteix element (i,j) quantifies the connectivity (i.e., number of interconnectingtrix and the binarized connectivity matrix were ordered such that all left

interconnected sub-blocks along the matrix diagonal exclusively involveinterhemispheric connections. Note the strong contralateral connection

pheric sub-block. All matrices represent group averages. (IV) The network-s.

II. Bi

t of almatry maonglyvolve

schizophrenia.

wm

tocg

sri

I

tp

Fbatf(ewg

A. Zalesky et al. BIOL PSYCHIATRY 2011;69:80–89 83

A trend toward increased path length, clustering, and �-ratioas found in patients, but these differences were not significant forost thresholds (Figure 2).

A strong linear association was found in the control group be-ween intelligence quotient and three key attributes of networkrganization: global efficiency, path length, and clustering coeffi-ient (Figure 3). No such associations were found in the patientroup (Supplement 1).

Significant associations with age, years of education, and GAFcore were not found. The significance of all the reported resultsemained unchanged when the analysis was repeated with agencluded as a nuisance linear covariate.

mpaired ConnectionsHaving found anomalous network organization at a global level,

he next step was to introduce localizing power and attempt to

igure 2. Graph measures describing topological attributes of corticocortiinarizing threshold (horizontal axis) determined the minimum number ofssumed. Data points marked with a star indicate a significant difference (pested at each threshold, correction for multiple comparisons was not perfoor a link to be drawn. (I) Nodal degree and (II) network efficiency were reducIII) clustering, (IV) path length, and (V) small-worldness was demonstratedxceeding T � 4, the graph for at least one subject broke apart into two or mhen considering path length, clustering, and small-worldness. The clusterraphs.

inpoint particular cortical pairs that were abnormally connected in

patients. Impaired connectivity was tested in all links satisfying athreshold of Kavg � 4, meaning that, for a node pair to be tested, ithad to be connected on average by at least four streamlines in thecontrol group.

For an FDR of 5%, three node pairs showed impaired connec-tivity (i.e., reduced streamline count) in patients: 1) right angularand right middle temporal; 2) right superior frontal and leftsuperior frontal; and 3) left inferior frontal triangularis and leftinsula.

An additional two pairs showed a trend toward impaired con-nectivity (i.e., FDR � 10%): 1) left cuneus and right superior occipi-tal; and 2) left cuneus and right cuneus.

The anatomical name by which each node is labeled was takendirectly from the Anatomical Automatic Labeling Atlas (60). Stream-lines interconnecting each of these five node pairs were visualizedto gain insight into the trajectory and morphology of the fiber tracts

nnectivity were quantified in a sample of patients with schizophrenia. Themlines that needed to interconnect a pair of nodes for a connection to be) between patients and control subjects. Because the same hypothesis was. Note that T � 0 indicates at least one or more streamlines must be presentschizophrenia for all binarizing thresholds considered. A trend for increased:tients but was not significant at most thresholds. For binarizing thresholdssconnected subgraphs. The binarizing threshold was therefore constrainedoefficient and path length were normalized with degree-matched random

cal costrea� .05rmeded inin pa

ore diing c

they represented (Figure 4).

www.sobp.org/journal

nte

Faw ) wase for th

Fso

84 BIOL PSYCHIATRY 2011;69:80–89 A. Zalesky et al.

w

An increased streamline count in patients was also tested butot found significant for any node pairs. The results presented in

his section were stable to variations in the threshold Kavg notxceeding �1.5.

igure 3. A linear association was found in the control group between intellttributes of network organization: namely, (I) global efficiency, (II) characteere found in the patient group. Spearman’s rank correlation coefficient (rs

xcluded from this correlation analysis, because IQ data were not available

JQ

I. l. Cuneus – r. Sup. Occipital

D

R

III. r. Angular – r. Mid. Temporal

ST

V.

igure 4. Impaired connections in patients with schizophrenia satisfyinguperior-inferior: blue. Node abbreviations: (D) right angular; (J) left cuneus; (

ww.sobp.org/journal

Impaired NetworkThe NBS was used to identify any connected networks that were

evident in control subjects but not in patients. A single network wasfound to be significantly impaired in patients (p � .05, corrected). The

e quotient (IQ), estimated with the Wechsler test of adult reading, and threepath length, and (III) average clustering coefficient. No associations with IQused to assess the significance of an association. Two control subjects wereem.

O

P

.

J Kl Cuneus – r. Cuneus

false discovery rate. Anterior-posterior fibers: green; left–right: red; andt cuneus; (O) right superior frontal; (P) left superior frontal; (Q) right superior

igencristic

IV

II.

a 10%K) righ

t insula.

nc1vbA5aTp

D

msest

FnApocc

A. Zalesky et al. BIOL PSYCHIATRY 2011;69:80–89 85

etwork interconnected medial frontal regions and several nodesomprising the parietal/occipital lobes. A total of 14 nodes (of 82) and5 links were involved. Streamlines interconnecting this network wereisualized as three-dimensional curves, and the volume encapsulatedy each node was rendered atop the streamline trajectories ( Figure 5).schematic representation of the network was also depicted (Figure

). The significance of the network remained unchanged when subjectge was included as a nuisance linear covariate for each link-based test.he sensitivity of the results were found to be robust to alternative NBSarameter settings (Supplement 1).

iscussion

The present study found evidence of dysconnection in white-atter connectional architecture in a large sample of patients with

chizophrenia. When considered from the perspective of mountingvidence for impaired synaptic plasticity (13,18,24,25), this resultuggests that functional dysconnectivity in schizophrenia might be

igure 5. Schematic of the frontal-parietal/occipital network that was impairetwork was impaired in patients but not in control subjects. Left: uniquelynterior-posterior fibers: green; left–right: red; and superior-inferior: blueositioned at its node’s center of gravity. Note that the positioning of somverlapping. Top: sagittal, left hemisphere. Bottom: axial. Node abbreviatioingulate, (D) right angular, (E) right superior occipital, (F) left precuneus, (G)uneus, (L) left middle occipital, (M) left lingual, and (N) left fusiform.

he result of a multifaceted pathophysiology involving both axonal

in addition to putative synaptic mechanisms. Indeed, both mecha-nisms can coexist (21).

Impaired connectivity in schizophrenia was found to involvethree regionally distinct groups of nodes: medial frontal, parietal/occipital, and left temporal. In patients, the frontal group was lesswell-connected to the parietal/occipital group via the cingulumbundle and left posterior cingulate, whereas multiple transcallosalpathways interconnecting contralateral nodes within the frontaland parietal/occipital groups also demonstrated disruption. Boththe cingulum (61– 64) and the corpus callosum (65– 69) have beenimplicated in several previous voxelwise and morphological inves-tigations of schizophrenia. Cingulate dysfunction is widely impli-cated in the pathophysiology of schizophrenia (70 –73), and thisstudy provides new evidence suggesting that its connections withan extended network of cortical regions is altered in the disorder.

The nodes found to demonstrate the greatest number of disrup-tions were the right superior occipital node (degree of 2 in Figure 5,

schizophrenia (p � .021 � .004, corrected). Each connection comprising thisred nodes and streamline representation of interconnecting fiber bundles.t: planar graph representation, where each node is depicted as a circlesterior nodes was slightly shifted from the true center of gravity to avert) right medial orbital frontal, (B) left superior medial frontal, (C) left anterioruperior temporal, (H) right calcarine, (I) left calcarine, (J) left cuneus, (K) right

ed incolo

. Righe po

ns: (Aright s

degree of 5 in Figure S3 in Supplement 1), left middle occipital node

www.sobp.org/journal

(iewanhvccs

tr(idrn

lnron

sepntpfsasioarcd

ccsi

P

oinsofiapdidr

86 BIOL PSYCHIATRY 2011;69:80–89 A. Zalesky et al.

w

degree 3 and 4), and the left precuneus (degree 5 and 2). It ismportant to note that the null hypothesis cannot be rejected forach of these nodes individually but only at the level of the subnet-ork they comprise. Other implicated nodes included the antero-

nd postero-medial components of the so-called default modeetwork, which have been identified as important connectivityubs in functional brain networks (74,75). These findings also con-erge with structural imaging evidence suggesting these medialomponents of the default mode network are among the mostommonly reported regions showing gray-matter changes inchizophrenia (31,72,76,77).

In addition, Figure 4 implicated connections originating fromhe left insula, superior frontal gyrus bilaterally, and a midtemporalegion; however, these connections were not detected by the NBSFigure 5). Note that the NBS does not have the ability to detectmpaired connections that exist in isolation (i.e., connections thato not form an interconnected subnetwork)—and hence, the likely

eason for some of the impaired connections identified in Figure 4ot featuring in Figure 5 and Figure S3 in Supplement 1.

It should be noted that some regions, such as the medial andateral temporal areas, were not identified as showing strong con-ectivity differences in our study, despite often showing volumetric

eductions in patients. This discrepancy might reflect the fact thatur analysis is indexing a distinct aspect of neuropathology—amely, connectivity as opposed to volume.

By localizing connectivity impairments in schizophrenia to apecific subset of cortical regions, the present study provides freshvidence for the macro-circuit theory of schizophrenia (1,78), whichosits only specific white-matter fibers are disrupted in schizophre-ia, either as a secondary effect owing to neuronal dysfunction in

he gray-matter regions they interconnect or as a primary patho-hysiology in schizophrenia. Although regional white-matter dif-

erences reported in previous studies also support this theory, thesetudies have not been able to model connectivity disturbancescross the entire brain. In contrast, our approach represents a firsttep toward a whole-brain description of connectivity disturbancesn schizophrenia. More specifically, our findings point to a networkf anterior and posterior medial regions as a key macro-circuitffected in schizophrenia. Interestingly, this network bears closeesemblance to a previously identified connectivity core in humanortical networks (35), which might support efficient functionalynamics (48).

Intellectual performance in the control group was strongly asso-iated with network efficiency, path length, and clustering, which isonsistent with previous studies (3,79) and supports the notion thatpecific properties of brain network organization have functionalmportance. No such associations were evident in patients.

athophysiological SignificanceIn this study, the differences found in patients can all have their

rigins traced back to a reduction in the total number of streamlinesnterconnecting particular pairs of gray-matter nodes. It is currentlyot understood whether a reduction in the number of streamlinesupporting a fiber tract can be ascribed to demyelinative processesr to a reduction in the number, density, or coherence of axonbers. The speed at which information can be transferred betweenpair of cortical regions connected via an impaired fiber tract is

ossibly diminished; in particular, a reduction in axonal number orensity might imply fewer nerve pulses are transferred, hence less

nformation is carried (1), whereas demyelination results in a slow-own of the conduction velocity of nerve pulses (80), thereby also

educing information transfer.

ww.sobp.org/journal

Volumetric reductions in white-matter and ventricular increaseshave been reported in some structural studies in schizophrenia (81).Volumetric changes were a potential confound in this study, be-cause a separate streamline was initialized from every voxel classi-fied as white-matter. Therefore, a significant volumetric white-mat-ter reduction might have resulted in a systematic reduction in thenumber of streamlines initialized in patients. However, a significantdifference between groups with respect to the total number ofstreamlines (i.e., inclusive of cortico-cortical, corticospinal, subcor-tical, and cerebellum and not length thresholded) was not evident.Note that streamlines were significantly longer in control subjects(length in patients: 4.7 � .3 cm, length in control subjects: 4.9 � .3cm, t � 2.7, p � .009), and the total number of streamlines exceed-ing 20 mm in length was also significantly greater in control sub-jects (total patients: 15,380 � 3651; total control subjects: 17,230 �3251, t � 2.4, p � .02). Furthermore, in a supplementary analysis, itwas found that the number of voxels classified as white-matter wasnot significantly different between both groups (Figure S1 in Sup-plement 1). Although these findings cannot rule out localized atro-phies, they suggest that global differences in white-matter volumewere unlikely to be the predominant factor underlying the effectsreported in this study. Any potential differences in gray-mattervolume are unlikely to be the source of these effects, becausetractography was explicitly constrained to white-matter and allconnectivity maps were normalized to a standard space where acommon gray-matter parcellation was used.

Methodological ConsiderationsPrevious studies have searched for regions at which myelin,

axonal, or other pathophysiological changes are suspected, withsingle-valued measures like FA. If abnormalities are detected, theassumption is that any fiber tracts traversing or encompassed bythese anomalous regions are damaged, thus suggesting a dyscon-nection between the gray-matter regions that they interconnect.However, due to the diffuse nature of localized findings, it can bedifficult to implicate specific fiber bundles affected by a localizedchange in white-matter. For example, as Kanaan et al. (28) pointedout, the study of Ardekani et al. (82) was only able to report anisot-ropy reductions “in the vicinity” of the cingulum, because the local-ized difference found also implicated the corpus callosum andother distinct fasciculi. Although recent studies (83– 85) have delin-eated fiber tracts of interest with tractography and then parameter-ized anisotropy along the delineated tract, these studies are disad-vantaged by the need for tract-specific hypotheses.

In the present study, whole-brain tractography was performedto derive a cortico-cortical network model of white-matter connec-tional architecture. Although this approach is more sensitive tonetwork disruptions, the results are contingent on the accuracy ofthe tractography algorithm.

The fiber assignment by continuous tracking (FACT) streamlinetracking algorithm (86) was used in this study. The FACT algorithmis simple, computationally inexpensive, and has been demon-strated to yield robust tractographic maps. However, streamlinetracking algorithms in general are known to have difficulty in fol-lowing long distance fibers, either due to partial volume effects, apoor fit of the diffusion tensor, or simply due to noise (37). There-fore, the ability to resolve long association fibers, such as the arcu-ate fasciculus, was limited in this study. Although advanced tracto-graphic algorithms (e.g., 36,37,87–90) alleviate some of the limitationsinherent to streamline tracking, they might not be computationallyfeasible for studies involving hundreds of subjects.

Furthermore, complex tract geometries (i.e., crossing fibers) re-

ctspcibtc

ipssomtdnhsymatcha

BCCswmbPBmptbppts6(nNGi

A. Zalesky et al. BIOL PSYCHIATRY 2011;69:80–89 87

ould have potentially resulted in a failure to detect some connec-ions. If crossing fibers were modeled, tractographic maps demon-trating a more extensive lateral and interhemispheric connectivityattern could have been expected (38). Moreover, although theortical nodes between which connectivity was impaired were

dentified, due to limitations of the tractography algorithm andecause the crossing fibers were not modeled, it might be difficult

o pinpoint the precise axonal bundles responsible for the loss ofonnectivity between each cortical pair identified.

Finally, the findings reported in this study might potentially benterpreted as effects arising from the significantly lower IQ of theatient group; however, disassociating low IQ from the disorder pere is generally difficult, given the pervasive cognitive difficultieshown by most patients (91). Furthermore, although the influencef antipsychotic medication and duration of illness on white-matterorphology cannot be ruled out in this study due to the nature of

he patient sample, recent studies (92) suggest no white-matterifferences between chronically and briefly medicated patients ando correlation with illness duration. Finally, although white-matteryperintensities were not noted by the radiologist, the presence ofubtle anomalies was possible in the individuals over the age of 50ears (93,94), which might have affected connectivity measure-ents. Note that 11 patients and 4 control subjects were over the

ge of 50 years, whereas 1 patient and 1 control subject were overhe age of 60 years. Future work will investigate any potentialoncordance with impaired functional connectivity, an avenue thatas received recent attention in Honey et al. (48) and Skudlarski etl. (95).

This study was supported by the Australian Schizophrenia Researchank, which is supported by the National Health and Medical Researchouncil (NHMRC) of Australia, the Pratt Foundation, Ramsay Healthare, the Viertel Charitable Foundation, and the Schizophrenia Re-

earch Institute. The computing resources used to undertake this studyere provided by the Florey Neuroscience Institutes and the Depart-ent of Electrical and Electronic Engineering at the University of Mel-

ourne. Software development was supported by a Human Brainroject grant from the National Institute of Biomedical Imaging andioengineering and the National Institute of Mental Health. Grapheasures reported in this article were computed with the MatlabBGL

ackage written by D. Gleich. The author AZ is supported by the Aus-ralian Research Council (ID: DP0986320). The author AF is supportedy an NHMRC CJ Martin Fellowship (ID: 454797). The author MLS is sup-orted by a Ronald Phillip Griffith Fellowship. The author LC is sup-orted by the Swiss Academy of Medical Science and the Swiss Na-

ional Science Foundation (ID: PASMP3 129357/1). The author CP isupported by an NHMRC Senior Principal Research Fellowship (ID:28386). The author CFW is supported by National Institutes of Health

NIH) grants: Novel DT-MRI Analyses of White Matter in Schizophre-ia (ID: NIH R01MH074794) and Neuroimage Analysis Center (ID:IH P41RR013218). The author EB is a half-time employee oflaxoSmithKline. All other authors reported no biomedical financial

nterests or potential conflicts of interest.

Supplementary material cited in this article is available online.

1. Konrad A, Winterer G (2008): Disturbed structural connectivity in schizo-phrenia–primary factor in pathology or epiphenomenon? Schizophr Bull34:72–92.

2. Bassett DS, Bullmore E, Verchinski BA, Mattay VS, Weinberger DR, Meyer-Lindenberg A (2008): Hierarchical organization of human cortical net-works in health and schizophrenia. J Neurosci 28:9239 –9248.

3. Li Y, Liu Y, Li J, Qin W, Kuncheng L, Yu C, Jiang T (2009): Brain anatomical

4. Liu Y, Liang M, Zhou Y, He Y, Hao Y, Song M, et al. (2008): Disruptedsmall-world networks in schizophrenia. Brain 131:945–961.

5. Rubinov M, Knock SA, Stam CJ, Micheloyannis S, Harris AWF, WilliamsLM, Breakspear M (2009): Small-world properties of nonlinear brainactivity in schizophrenia. Hum Brain Mapp 30:403– 416.

6. Bluhm RL, Miller J, Lanius RA, Osuch EA, Boksman K, Neufeld RW, et al.(2007): Spontaneous low-frequency fluctuations in the BOLD signal inschizophrenic patients: anomalies in the default network. Schizophr Bull33:1004 –1012.

7. Camchong J, Macdonald AW, Bell C, Mueller BA, Lim KO (2009): Alteredfunctional and anatomical connectivity in schizophrenia [published on-line ahead of print November 17]. Schizophr Bull. doi:10.1093/schbul/sbp131.

8. Garrity AG, Pearlson GD, McKiernan K, Lloyd D, Kiehl KA, Calhoun VD(2007): Aberrant “default mode” functional connectivity in schizophre-nia. Am J Psychiatry 164:450 – 457.

9. Andreasen NC, Paradiso S, O’Leary DS (1998): “Cognitive dysmetria” asan integrative theory of schizophrenia: A dysfunction in cortical-subcor-tical-cerebellar circuitry? Schizophr Bull 24:203–218.

10. Bullmore ET, Frangou S, Murray RM (1997): The dysplastic net hypothe-sis: An integration of developmental and disconnectivity theories ofschizophrenia. Schizophr Res 28:143–156.

11. Fletcher P, McKenna PJ, Friston KJ, Frith CD, Dolan RJ (1999): Abnormalcingulate modulation of fronto-temporal connectivity in schizophrenia.Neuroimage 9:337–342.

12. Friston KJ, Frith CD (1995): Schizophrenia: A disconnection syndrome?Clin Neurosci 3:89 –97.

13. Friston KJ (1998): The disconnection hypothesis. Schizophr Res 30:115–125.

14. Friston KJ (2002): Dysfunctional connectivity in schizophrenia. WorldPsychiatry 1:66 –71.

15. Hoffman RE, Buchsbaum MS, Escobar MD, Makuch RW, Nuechterlein KH,Guich SM (1991): EEG coherence of prefrontal areas in normal andschizophrenic males during perceptual activation. J Neuropsychiatr ClinNeurosci 3:169 –175.

16. Meyer-Lindenberg A, Poline JP, Kohn PD, Holt MS, Egan MF, WeinbergerDR, Berman KF (2001): Evidence for abnormal cortical functional con-nectivity during working memory in schizophrenia. Am J Psychiatry 158:1809 –1817.

17. Pantelis C, Barnes TR, Nelson HE (1992): Is the concept of frontal-subcor-tical dementia relevant to schizophrenia? Br J Psychiatry 160:442– 460.

18. Stephan KE, Baldeweg T, Friston KJ (2006): Synaptic plasticity and dys-connection in schizophrenia. Biol Psychiatry 59:929 –939.

19. Weinberger DR (1993): A connectionist approach to the prefrontal cor-tex. J Neuropsychiatr Clin Neurosci 5:241–253.

20. Winterer G, Coppola R, Egan MF, Goldberg TE, Weinberger DR (2003):Functional and effective frontotemporal connectivity and genetic riskfor schizophrenia. Biol Psychiatry 54:1181–1192.

21. Stephan KE, Friston KJ, Frith CD (2009): Dysconnection in schizophrenia:From abnormal synaptic plasticity to failures of self-monitoring. Schizo-phr Bull 35:509 –527.

22. Hakak Y, Walker JR, Li C, Wong WH, Davis KL, Buxbaum JD, et al. (2001):Genome-wide expression analysis reveals dysregulation of myelina-tion-related genes in chronic schizophrenia. Proc Natl Acad Sci U S A98:4746 – 4751.

23. Uranova NA, Casanova MF, DeVaughn NM, Orlovskaya DD, Denisov DV(1996): Ultrastructural pathology of neuronal connectivity in postmor-tem brains of schizophrenic patients. Schizophr Res 22:81– 83.

24. Daskalakis ZJ, Christensen BK, Fitzgerald PB, Chen R (2008): Dysfunc-tional and neural plasticity in patients with schizophrenia. Arch GenPsychiatry 65:378 –385.

25. McGlashan TH, Hoffman RE (2000): Schizophrenia as a disorder of devel-opmentally reduced synaptic connectivity. Arch Gen Psychiatry 57:637–648.

26. Basser PJ, Mattiello J, LeBihan D (1996): Diffusion tensor MR imaging ofthe human brain. Radiology 201:637– 648.

27. Foong J, Symms MR, Barker GJ, Maier M, Woermann FG, Miller DH, RonMA (2001): Neuropathological abnormalities in schizophrenia: Evidencefrom magnetization transfer imaging. Brain 124:882– 892.

28. Kanaan RAA, Kim JS, Kaufmann WE, Pearlson GD, Barker GJ, McGuire PK(2005): Diffusion tensor imaging in schizophrenia. Biol Psychiatry 58:

www.sobp.org/journal

2

3

3

3

3

3

3

3

3

3

3

4

4

4

4

4

4

4

4

4

4

5

5

5

5

88 BIOL PSYCHIATRY 2011;69:80–89 A. Zalesky et al.

w

9. Kubicki M, McCarley RW, Shenton ME (2005): Evidence for white matterabnormalities in schizophrenia. Curr Opin Psychiatry 18:121–134.

0. Feinberg I (1982): Schizophrenia: Caused by a fault in programmedsynaptic elimination during adolescence? J Psychiatr Res 17:319 –334.

1. Ellison-Wright I, Glahn DC, Laird AR, Thelen SM, Bullmore E (2008): Theanatomy of first-episode and chronic schizophrenia: An anatomical like-lihood estimation meta-analysis. Am J Psychiatry 165:1015–1023.

2. Gong G, He Y, Concha L, Lebel C, Gross DW, Evans AC, Beaulieu C (2009):Mapping anatomical connectivity patterns of human cerebral cortexusing in vivo diffusion tensor imaging tractography. Cereb Cortex9:524 –536.

3. Gong G, Rosa-Neto P, Carbonell F, Chen ZJ, He Y, Evans AC (2009): Age-and gender-related differences in the cortical anatomical network.J Neurosci 29:15684 –15693.

4. Hagmann P, Kurant M, Gigandet X, Thiran P, Wedeen VJ, Meuli R, ThiranJ-P (2007): Mapping human whole-brain structural networks with diffu-sion MRI. PLoS ONE 2:e597.

5. Hagmann P, Cammoun L, Gigandent X, Meuli R, Honey CJ, Wedeen VJ,Sporns O (2008): Mapping the structural core of the human cerebralcortex. PLoS Biol 6:e159.

6. Iturria-Medina Y, Sotero RC, Canales-Rodríguez EJ, Alemán-Gómez Y,Melie-García L (2008): Studying the human brain anatomical networkvia diffusion-weighted MRI and graph theory. Neuroimage 40:1064 –1076.

7. Zalesky A, Fornito A (2009): A DTI-derived measure of cortico-corticalconnectivity. IEEE Trans Med Imaging 28:1023–1036.

8. Zalesky A, Fornito A, Harding IH, Cocchi L, Yücel M, Pantelis C, BullmoreET (2010): Whole-brain anatomical networks: Does the choice of nodesmatter? Neuroimage 50:970 – 83.

9. Bullmore E, Sporns O (2009): Complex brain networks: Graph theoreticalanalysis of structural and functional systems. Nat Rev Neurosci 10:186 –198.

0. Castle DJ, Jablensky A, McGrath JJ, Carr V, Morgan V, Waterreus A, et al.(2006): The diagnostic interview for psychoses (DIP): Development, re-liability and applications. Psychol Med 36:69 – 80.

1. Wechsler D (1981): Manual for the Wechsler Adult Intelligence Scale(WAIS-R). San Antonio, Texas: Psychological Corporation.

2. Salvador R, P̃ena A, Menon DK, Adrian Carpenter T, Pickard JD, BullmoreE (2005): Formal characterization and extension of the linearized diffu-sion tensor model. Hum Brain Mapp 24:144 –155.

3. Basser PJ (1995): Inferring microstructural features and the physiologi-cal state of tissues from diffusion-weighted images. NMR Biomed 8:333–344.

4. Basser PJ, Pajevic S, Pierpaoli C, Duda J, Aldroubi A (2000): In vivo fibertractography using DT-MRI data. Magn Reson Med 44:625– 632.

5. Catani M, Howard RJ, Pajevic S, Jones DK (2002): Virtual in vivo interac-tive dissection of white matter fasciculi in the human brain. Neuroimage17:77–94.

6. Catani M (2006): Diffusion tensor magnetic resonance imaging tractog-raphy in cognitive disorders. Curr Opin Neurol 19:599 – 606.

7. Conturo TE, Lori NF, Cull TS, Akbudak E, Snyder AZ, Shimony JS, et al.(1999): Tracking neuronal fiber pathways in the living human brain. ProcNatl Acad Sci U S A 96:10422–10427.

8. Honey CJ, Sporns O, Cammoun L, Gigandet X, Thiran J-P, Meuli R, Hag-mann P (2009): Predicting human resting-state functional connectivityfrom structural connectivity. Proc Natl Acad Sci U S A 106:2035–2040.

9. Park C-H, Kim SY, Kim Y-H, Kim K (2008): Comparison of the small-worldtopology between anatomical and functional connectivity in the hu-man brain. Phys A 387:5958 –5962.

0. Skudlarski P, Jagnnathan K, Calhoun VD, Hampson M, Skudlarska BA,Pearlson G (2008): Measuring brain connectivity: Diffusion tensor imag-ing validates resting state temporal correlations. Neuroimage 43:554 –561.

1. He Y, Chen ZJ, Evans AC (2008): Structural insights into aberrant topo-logical patterns of large-scale cortical networks in Alzheimer’s disease.J Neurosci 28:5756 – 4766.

2. He Y, Dagher A, Chen Z, Charil A, Zijdenbos A, Worsley K, Evans A (2009):Impaired small-world efficiency in structural cortical networks in multi-ple sclerosis associated with white matter lesion load. Brain 132:3366 –3379.

3. Rubinov M, Sporns O (2010): Complex network measures of brain con-

ww.sobp.org/journal

54. Nichols TE, Holmes AP (2001): Nonparametric permutation tests forfunctional neuroimaging: A primer with examples. Hum Brain Mapp15:1–25.

55. Genovese CR, Lazar NA, Nichols T (2002): Thresholding of statisticalmaps in functional neuroimaging using the false discovery rate. Neuro-image 15:870 – 878.

56. Zalesky A, Fornito A, Bullmore ET (2010): Network-based statistic: Iden-tifying differences in brain networks. Neuroimage 53:1197-1207.

57. Achard S, Bullmore E (2007): Efficiency and cost of economical brainfunctional networks. PLoS Comput Biol 3:e17.

58. Latora V, Marchiori M (2001): Efficient behavior of small-world networks.Phys Rev Lett 87:198 –701.

59. Sporns O, Tononi G, Edelman GM (2000): Theoretical neuroanatomy:Relating anatomical and functional connectivity in graphs and corticalconnection matrices. Cereb Cortex 10:127–141.

60. Tzourio-Mazoyer N, Landeau B, Papathanassiou D, Crivello F, Etard O,Delcroix N, et al. (2002): Automated anatomical labeling of activations inSPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage 15:273–289.

61. Kubicki M, Westin CF, Nestor PG, Wible CG, Frumin M, Maier SE, et al.(2003): Cingulate fasciculus integrity disruption in schizophrenia: Amagnetic resonance diffusion tensor imaging study. Biol Psychiatry 54:1171–1180.

62. Seal ML, Yucel M, Fornito A, Wood SJ, Harrison BJ, Walterfang M, et al.(2008): Abnormal white matter microstructure in schizophrenia: A vox-elwise analysis of axial and radial diffusivity. Schizophr Res 101:106 –110.

63. Sun Z, Wang F, Cui L, Breeze J, Du X, Wang X, et al. (2003): Abnormalanterior cingulum in patients with schizophrenia: A diffusion tensorimaging study. Neuroreport 14:1833–1836.

64. Wang F, Sun Z, Cui L, Du X, Wang X, Zhang H, et al. (2004): Anteriorcingulum abnormalities in male patients with schizophrenia deter-mined through diffusion tensor imaging. Am J Psychiatry 161:573–575.

65. Cheung V, Cheung C, McAlonan GM, Deng Y, Wong JG, Yip L, et al.(2008): A diffusion tensor imaging study of structural disconnectivity innever-medicated, first-episode schizophrenia. Psychol Med 38:877– 885.

66. Gasparotti R, Valsecchi P, Carletti F, Galluzzo A, Liserre R, Cesana B,Sacchetti E (2009): Reduced fractional anisotropy of corpus callosum infirst-contact, antipsychotic drug-naive patients with schizophrenia.Schizophr Res 108:41– 48.

67. Kubicki M, Styner M, Bouix S, Gerig G, Markant D, Smith K, et al. (2008):Reduced interhemispheric connectivity in schizophrenia-tractographybased segmentation of the corpus callosum. Schizophr Res 106:125–131.

68. Walterfang M, Yung A, Wood AG, Reutens DC, Phillips L, Wood SJ, et al.(2008): Corpus callosum shape alterations in individuals prior to theonset of psychosis. Schizophr Res 103:1–10.

69. Walterfang M, Wood AG, Reutens DC, Wood SJ, Chen J, Velakoulis D, etal. (2008): Morphology of the corpus callosum at different stages ofschizophrenia: Cross-sectional study in first-episode and chronic illness.Br J Psychiatry 192:429 – 434.

70. Benes FM (1993): Neurobiological investigations in cingulate cortex ofschizophrenic brain. Schizophr Bull 19:537–549.

71. Fornito A, Yucel M, Dean B, Wood SJ, Pantelis C (2009): Anatomicalabnormalities of the anterior cingulate cortex in schizophrenia: Bridg-ing the gap between neuroimaging and neuropathology. Schizophr Bull35:973–993.

72. Glahn DC, Laird AR, Ellison-Wright I, Thelen SM, Robinson JL, LancasterJL, et al. (2008): Meta-analysis of gray matter anomalies in schizophrenia:Application of anatomic likelihood estimation and network analysis.Biol Psychiatry 64:774 –781.

73. Snitz BE, MacDonald A, Cohen JD, Cho RY, Becker T, Carter CS (2005):Lateral and medial hypofrontality in first-episode schizophrenia: Func-tional activity in a medication-naive state and effects of short-termatypical antipsychotic treatment. Am J Psychiatry 162:2322–2329.

74. Buckner RL, Sepulcre J, Talukdar T, Krienen FM, Liu H, Hedden T, et al.(2009): Cortical hubs revealed by intrinsic functional connectivity: Map-ping, assessment of stability, and relation to Alzheimer’s disease. J Neu-rosci 29:1860 –1873.

75. Gusnard DA, Raichle ME (2001): Searching for a baseline: Functionalimaging and the resting human brain. Nat Rev Neurosci 2:685– 694.

76. Fornito A, Yucel M, Patti J, Wood SJ, Pantelis C (2009): Mapping grey

7

7

7

8

8

8

8

8

8

8

A. Zalesky et al. BIOL PSYCHIATRY 2011;69:80–89 89

tion analysis of voxel-based morphometry studies. Schizophr Res 108:104 –113.

7. Fornito A, Yucel M, Wood SJ, Adamson C, Velakoulis D, SalingMM, et al. (2008): Surface-based morphometry of the anterior cingulatecortex in first episode schizophrenia. Hum Brain Mapp 29:478–489.

8. Ellison-Wright I, Bullmore E (2009): Meta-analysis of diffusion tensorimaging studies in schizophrenia. Schizophr Res 108:3–10.

9. van den Heuvel MP, Stam CJ, Kahn RS, Hulshoff Pol HE (2009): Efficiencyof functional brain networks and intellectual performance. J Neurosci29:7619 –7624.

0. Rushton WAH (1951): Theory of the effects of fibre size in medulatednerve. J Physiol 115:101–122.

1. Shenton ME, Dickey CC, Frumin M, McCarley RW (2001): A review of MRIfindings in schizophrenia. Schizophr Res 49:1–52.

2. Ardekani BA, Nierenberg J, Hoptman MJ, Javitt DC, Lim KO (2003): MRIstudy of white matter diffusion anisotropy in schizophrenia. Neurore-port 14:2025–2029.

3. Kanaan RA, Shergill SS, Barker GJ, Catani M, Ng VW, Howard R, et al.(2006): Tract-specific anisotropy measurements in diffusion tensor im-aging. Psychiatry Res 146:73– 82.

4. O’Donnell LJ, Westin C-F, Golby AJ (2009): Tract-based morphometry forwhite matter group analysis. Neuroimage 45:832– 844.

5. Oh JS, Kubicki M, Rosenberger G, Bouix S, Levitt JJ, McCarley RW, et al.(2009): Thalmo-frontal white matter alterations in chronic schizophre-nia: A quantitative diffusion tractography study. Hum Brain Mapp 30:3812–3825.

6. Mori S, Crain BJ, Chacko VP, Van Zijl PC (1999): Three-dimensional track-

87. Behrens TE, Berg HJ, Jbabdi S, Rushworth MF, Woolrich MW (2007):Probabilistic diffusion tractography with multiple fibre orientations:What can we gain? Neuroimage 34:144 –155.

88. Behrens TE, Woolrich MW, Jenkinson M, Johansen-Berg H, Nunes RG,Clare S, et al. (2003): Characterization and propagation of un-certainty in diffusion-weighted MR imaging. Magn Reson Med 50:1077–1088.

89. Jbabdi S, Woolrich MW, Anderson JLR, Behrens TEJ (2007): A Bayesianframework for global tractography. Neuroimage 37:116 –129.

90. Zalesky A (2008): Diffusion tensor magnetic resonance imaging (DT-MRI) fiber tracking: A shortest paths approach. IEEE Trans Med Imaging27:1458 –1471.

91. Heinrichs RW, Zakzanis KK (1998): Neurocognitive deficit in schizophrenia:A quantitative review of the evidence. Neuropsychology 12:426–445.

92. Kanaan R, Barker G, Brammer M, Giampietro V, Shergill S, Woolley J, et al.(2009): White matter microstructure in schizophrenia: Effects of disor-der, duration and medication. Br J Psychiatry 194:236 –242.

93. Wen W, Sachdev PS, Li JJ, Chen X, Anstey KJ (2009): White matterhyperintensities in the forties: Their prevalence and topography inan epidemiological sample aged 44 – 88. Hum Brain Mapp 30:1155–1167.

94. Schmidt R, Enzinger C, Ropele S, Schmidt H, Fazekas F (2003): Progres-sion of cerebral white matter lesions; 6-year results of the AustrianStroke Prevention Study. Lancet 361:2046 –2048.

95. Skudlarski P, Jagannathan K, Anderson K, Stevens MC, Calhoun VD,Skudlarska BA, Pearlson G (2010): Brain connectivity is not only lowerbut different in schizophrenia: A combined anatomical and functional

www.sobp.org/journal