Volumetric Evaluation of the Thalamus in Schizophrenic
Male Patients Using Magnetic Resonance Imaging
Chiara M. Portas, Jill M. Goldstein, Martha E. Shenton, Hiroto H. Hokama,
Cynthia G. Wible, Iris Fischer, Ron Kikinis, Robert Donnino, Ferenc A. Jolesz, and
Robert W. McCarley
Background: The thalamus, an important subcortical
brain region connecting limbic and prefrontal cortices,
has a significant role in sensory and cortical processing.
Although inconsistently, previous studies have demon-
strated neuroanatomical abnormalities in the thalamus of
Methods: This structural magnetic resonance imaging
study, based on segmentation of contiguous coronal
1.5-mm images, compared thalamic brain volumes of 15
chronic, male schizophrenic patients with 15 normal
controls matched on age, sex, handedness, and parental
Results: There were no significant differences between
patients and controls in thalamic volumes, right or left,
adjusted for total brain volume; however, there were
significantly different correlations of thalamic volumes
with prefrontal white matter and lateral ventricles among
patients, but not among controls. Thalamic volumes
among patients were also significantly correlated with
bizarre behavior, hallucinations, and thought disorder.
Conclusions: Findings suggest that connectivity between
thalamic nuclei and prefrontal cortical areas are abnor-
mal in chronic male schizophrenic patients. In addition,
ventricular enlargement may be, in part, due to subtle
reduction in thalamic volume and/or in volume of
thalamocortical and corticothalamic fibers secondary to
thalamic abnormalities. Finally, correlations with positive
symptomatology underscore the role of the thalamus in
gating or filtering of sensory information and coordina-
tion of cortical processing.
649–659 © 1998 Society of Biological Psychiatry
Biol Psychiatry 1998;43:
Key Words: Schizophrenia, thalamus, magnetic reso-
dreasen et al 1990; Seidman et al 1996), and limbic and
medial temporal lobe structures (e.g., Bogerts et al 1985,
1990; Shenton et al 1992; Falkai et al 1988). A few recent
in vivo imaging studies have also implicated the thalamus
in schizophrenia (Andreasen et al 1990, 1994; Buschbaum
et al 1996; Flaum et al 1995; Goldstein et al 1996;
Jernigan et al 1991; Seidman et al 1996). The idea that
thalamic abnormalities may contribute to understanding
the pathology in schizophrenia is not new, in particular in
its role in attention and language processing (e.g., Mirsky
1991; Seidman 1983; Crosson and Hughes 1987; Kolb
1977). Moreover, the thalamus is an important part of the
subcortical network, connecting, among other regions,
limbic, basal ganglia, and prefrontal cortical regions (Kolb
1977; Groenewegen et al 1990; Pandya and Yeterian
1990). Since cortical activity is controlled and integrated,
in part, by the thalamic nuclei, abnormalities in these areas
may play a role in the pathogenesis of schizophrenia
(Crosson and Hughes 1987; Gold and Weinberger 1991;
Dolan et al 1993; Silverweig et al 1995).
Early postmortem studies did not show any significant
difference in the total thalamic volume of schizophrenic
patients compared to the thalamic volume of control subjects
(Rosenthal and Bigelow 1972; Lesch and Bogerts 1984;
Kelsoe et al 1988). However, three postmortem studies did
find a significant reduction in the total number of neurons in
the brains of schizophrenic patients compared to normal
control brains, specifically in the mediodorsal thalamic nu-
cleus (Hempel and Treff 1959; Baumer 1954; Pakkenberg
1990, 1992). In contrast, one postmortem study reported no
significant difference in nerve cell density in the mediodorsal
thalamus (Dom et al 1981).
Finally, comparing schizophrenic patients with normal
controls, a number of recent in vivo magnetic resonance
imaging (MRI) studies have reported significant reduc-
tions in thalamic volume (Flaum et al 1995; Goldstein et al
1996), thalamic area (Andreasen et al 1990; Buchsbaum et
he primary brain areas implicated in schizophrenia are
prefrontal cortex (e.g., Weinberger et al 1986; An-
From the Clinical Neuroscience Division, Neuroscience Laboratory, Department of
Psychiatry, Harvard Medical School, (CMP, JMG, MES, HHH, CGW, IF, RD,
RWM); Harvard Medical School, Department of Psychiatry at Massachusetts
Mental Health Center, Harvard Institute of Psychiatric Epidemiology and
Genetics, (JMG); and Surgical Planning Laboratory, MRI Division, Depart-
ment of Radiology, Brigham and Women’s Hospital, Harvard Medical School,
Boston, Massachusetts (RK, FAJ).
Address reprint requests to Jill M. Goldstein, PhD, Harvard Institute of Psychiatric
Epidemiology and Genetics, Massachusetts Mental Health Center, 74 Fenwood
Road, Boston, MA 02115; or Robert W. McCarley, M.D., Brockton-West
Roxbury VAMC, 940 Belmont St., Brockton, MA 02401.
Received July 18, 1996; revised April 15, 1997; accepted June 23, 1997.
© 1998 Society of Biological Psychiatry0006-3223/98/$19.00
al 1996), or in differences between averaged pixel signal
intensities reflecting abnormal increases or decreases in
thalamic parenchyma or in adjacent white matter fiber
tracts (Andreasen et al 1994). Further, a recent study
indicated significant thalamic volume reductions in first-
degree relatives of schizophrenics compared to normal
controls (Seidman et al 1996). Although there have been
seven previous structural MRI studies of the thalamus in
schizophrenia, only a few have produced volumetric data,
and have used slice thicknesses of 3 mm with 1.5-mm gaps
between slices (Flaum et al 1995), 3 mm with no gaps
(Goldstein et al 1996; Seidman et al 1996), and 5 mm with
2.5-mm gaps for diencephalon gray-matter structures
(Jernigan et al 1991). In the present study, we used
contiguous thin slice (1.5 mm) MR high-resolution im-
ages, on which automated and manual segmentation and
three-dimensional (3D) surface rendering techniques were
used to evaluate thalamic volumes in schizophrenic pa-
tients and controls.
Methods and Materials
The sample for this study has been reported in previous publi-
cations (Shenton et al 1992). Briefly, 15 DSM-III-R schizo-
phrenic patients (13 hospitalized) were selected from among
patients at the Brockton Veterans Administration Medical Cen-
ter, according to the following criteria: patients were male,
right-handed, 20–55 years old, had never undergone electrocon-
vulsive shock treatment, had no history of neurological illness,
no major alcohol or drug abuse in the previous 5 years, no history
of alcohol dependence and no secondary diagnosis of DSM-III-R
alcohol abuse, and had not been receiving medications known to
affect the results of MRI of the brain, such as steroids. The
patients had a mean age of 37.6 years and standard deviation (?)
9.3, mean years of education of 11.7, and parental socioeconomic
status (PSES) of lower middle class (Hollingshead classification
of 3.4 ? 0.1). Their mean age of onset was 22.3 ? 2.8 years,
mean duration of illness was 15.7 ? 8.8 years, and mean time
hospitalized was 7.1 ? 4.6 years. Patient diagnoses were made in
accord with DSM-III-R (American Psychiatric Association
1987) on the basis of chart review, and from information
obtained from administration of the Schedule for Affective
Disorders and Schizophrenia (Spitzer and Endicott 1978). All
schizophrenic patients were receiving neuroleptic medication
(mean ? 881 ? 683 mg/day chlorpromazine equivalent).
The normal control group, recruited from newspaper adver-
tisements, consisted of 15 adult males (20–55 years old) previ-
ously screened for neurological or psychiatric histories, and
matched to patients on age, sex, handedness, and PSES. Controls
had no history of electroconvulsive shock treatment, neurologic
illness, or steroid use, no lifetime history of drug/alcohol addic-
tion or DSM-III-R abuse within the last 5 years, nor psychiatric
illness in themselves or their first-degree relatives. There were no
statistically significant differences between patients and controls
in age, height, weight, head circumference, PSES, or in scores on
the WAIS-R information subscale (Wechsler 1981).
Clinical evaluations were reported previously (Shenton et al
1992). Briefly, three instruments were used to assess type and
severity of symptoms: the Scale for the Assessment of Positive
Symptoms (SAPS; Andreasen 1984), the Scale for the Assess-
ment of Negative Symptoms (SANS; Andreasen 1981), and the
Thought Disorder Index (TDI; Johnston and Holzman 1979). The
average score on the TDI was 60 ? 62, and median ? 44.
Normal subjects score ? 5. Using the Andreasen classification,
11 out of the 15 patients had mainly positive symptoms, and 4
were mixed. None of the patients had mainly negative symptoms,
although some negative symptoms were present (mean global
SANS score ? 9.1).
Image Acquisition and Processing
The MRI scans were obtained through the entire brain on a 1.5-T
General Electric SIGNA System (GE Medical System, Milwau-
kee, WI), using a 3D Fourier transform spoiled gradient-recalled
(3DFT SPGR) acquisition in steady state. The SPGR images
were obtained with the following parameters: echo time (TE) ?
5 msec, repetition time (TR) ? 35 msec, one repetition, nutation
angle ? 45 degrees, field of view ? 24 cm, acquisition matrix ?
256 ? 256 ? 124, voxel dimensions ? 0.9375 ? 0.9375 ? 1.5
mm. These data were stored as 124 contiguous 1.5-mm coronal
slices. Intracranial contents were measured by 108 contiguous
double echo spin-echo 3-mm axial slices throughout the brain.
Imaging parameters were: TE ? 30 & 80 msec, TR ? 3000
msec, field of view ? 24 cm, acquisition matrix ? 256 ? 256,
and voxel dimensions ? 0.9375 ? 0.9375 ? 3 mm (see Shenton
et al 1992 for details). Automated and manual segmentation
methods, 3D slice editing techniques that allowed reformatting of
slices in three different planes, and 3D surface rendering tech-
niques were applied to the data collected to create 3D represen-
tations of the thalamus (Kikinis et al 1990; Cline et al 1990).
Thalamic segmentation was performed using coronal slices.
Sagittal and axial reslicing software was utilized to view the
segmentation results in different planes.
Thalamic Boundary Definition
The automated segmentation procedures produced the separation
of gray and white matter based on differences in signal intensity
values (Kikinis et al 1990; Cline et al 1990). Manual segmenta-
tion of the thalamus occurred in 20–21 consecutive slices (out of
an average of 120 slices over the entire brain). To overcome the
problem of partial volume (PV) effects, we decided to include
50% of the PV area. The definitions of the landmarks used for the
thalamus are described as follows. The most anterior boundary
was difficult to resolve. We used the mammillary bodies of the
hypothalamus. The ventralis anterior nucleus is just dorsal to the
hypothalamus, bounded laterally by the internal capsule, dorsally
by the lateral ventricle, and medially by the third ventricle. The
posterior boundary was defined when the thalamus merged under
the crus fornix. The thalamus was medially defined using the
third ventricle. The inferior border was defined when the
thalamus merged with the brain stem and the superior border, by
650 C.M. Portas et al
the main body of the lateral ventricle. (Duvernoy 1991; De
Armond et al 1989; Roberts and Hanaway 1971; Haines 1991
were used as primary anatomical references.) Figure 1 presents
an example of the thalamic segmentation of one coronal slice.
Figure 2 is a 3D representation of the thalamus in relation to the
ventricles. A more detailed description of the thalamic bound-
aries using a case example is presented in the Appendix.
Intrarater and interrater reliability were conducted by three
raters (CMP, IF, RD). Since CMP measured all cases, intrarater
reliability on the thalamic segmentation was assessed 6 months
apart on 3 randomly selected cases. The volume difference
between the first and the second measurement was negligible in
all 3 cases (?1%). Interrater reliability, estimated by intraclass
correlation coefficients, for 3 randomly selected cases across
three raters was: .93 for total thalamic volume, .93 for right
thalamic volume, and .91 for left thalamic volume.
Volumes of the thalamus in the right and left hemispheres were
segmented and analyzed separately and as total thalamic volume.
All volumes were adjusted for intracranial brain volume to adjust
for head size. Volumetric assessments presented below were thus
divided by intracranial volume and multiplied by 100. Analysis
of covariance (ANCOVA) was used to test for the effect of group
(i.e., schizophrenics versus normal controls), controlled for age
and PSES. We controlled for PSES because, even though this
variable was not significantly different between groups, there
were more parents of schizophrenics in the lowest social class
than in the control parental group. The use of general linear
models, such as ANCOVA, was appropriate, since tests of
normality for the total, right, and left thalami in patients and
in controls showed that these data were normally distributed in
We then tested whether there were differences between
schizophrenics and normal controls in the relationships, i.e.,
covariance structure, or correlations, between the thalamus and
some of the other key brain regions that are known to have
specific neural connections to the thalamus and have been found
to be abnormal in schizophrenia, i.e., prefrontal cortex, cingulate
gyrus, striatum, hippocampus, amygdala, and the superior tem-
poral gyrus. Temporal lobe structures and basal ganglia were of
particular interest, since in previous work we found that they
were significantly different in these patients than in the normal
controls (Shenton et al 1992; Hokama et al 1995). We were
specifically interested in a test of the differences in the covari-
ance structure, or correlations, partialled for PSES and age,
among patients versus controls, which consisted of normalizing
the correlations using Fisher’s z transformation (Fisher 1958).
Tests of the differences in the correlational structure between
patients and controls will not be influenced by any potential bias
in correlating adjusted brain regions.
Finally, we were interested in whether, and if so, how,
thalamic volumes among the patients related to their clinical
presentation and severity, particularly positive symptoms. A
recent study argued that thalamic abnormalities are key to
understanding the expression of positive symptoms (Andreasen
et al 1994). Adjusted thalamic volumes (left, right, and total)
among the schizophrenic patients were correlated with global
scores of positive symptoms (i.e., SAPS; Andreasen 1984),
which included auditory hallucinations, delusions, and bizarre
behavior, and total score on the Thought Disorder Index
(Johnston and Holzman 1979). Negative symptoms (i.e., SANS;
Andreasen 1981), which included affective flattening, alogia,
avolition–apathy, asociality, and attentional behavior ratings,
were also examined in relation to thalamic volumes. These
analyses were exploratory, given that the sample in this study
was characterized as “mainly positive.” The nonparametric
Spearman’s correlation coefficient, rho, was applied using two-
tailed significance tests. Spearman’s rho was selected because it
takes into account outliers that can skew findings, particularly in
small sample sizes.
Differences in Thalamic Volumes between Patients
Results of the ANCOVAs, controlling for age and PSES,
showed that there were no significant differences in the
adjusted volumes for the right, left, nor total thalamus
between controls and patients, respectively (right: x ? ?
0.44 ? 0.01 vs. x ? ? 0.45 ? 0.01; left: x ? ? 0.44 ? 0.04 vs.
x ? ? 0.44 ? 0.01; total: x ? ? 0.88 ? 0.02 vs. 0.89 ? 0.02).
Thalamic volumes unadjusted for intracranial brain vol-
ume also showed no significant differences between con-
trols and patients, respectively (right: x ? ? 7.01 ? 0.64 vs.
x ? ? 7.20 ? 0.75; left: x ? ? 6.91 ? 0.81 vs. 7.06 ? 0.63;
total: x ? ? 13.9 ? 1.4 vs. x ? ? 14.26 ? 1.3).
We then tested for differences between patients and
controls in relationships between thalamic volume and
other brain regions of interest (ROIs) found to be impor-
tant in schizophrenia and with which thalamic nuclei are
known to have significant connections. Table 1 presents
the significant correlations between the left and right
adjusted thalamic volumes and the other brain ROIs.
Correlations between left and right thalamic volumes and
ROIs not listed in Table 1, i.e., middle, inferior, and orbital
prefrontal cortex, cingulate gyrus, amygdala, superior
temporal gyrus, and putamen, were not significantly dif-
ferent between patients and controls. In general, as shown
in Table 1, there was a greater likelihood of significant
differences in the correlations between thalamic volume
and ROIs between patients and controls in the left hemi-
sphere than in the right.
The main finding in Table 1 is the significant negative
correlations between the left and right thalamus with left
and right ventricles among the patients, but not among
controls, indicating that the smaller the thalamus, the
larger the ventricles. There was little variability in the size
of the ventricles among controls (SD ? 0.16; x ? ? 0.44),
Thalamus MRI in Schizophrenia651
Figure 1. Example of the segmentation of the thal-
amus on one coronal slice.
Figure 2. Three-dimensional repre-
sentation of the thalamus in relation
to the ventricles.
652 C.M. Portas et al
whereas in patients, the variability was more than twice
the variability among the controls (SD ? 0.37; x ? ? 0.52).
This is compatible with pathological processes operat-
ing on ventricular volume with variable severity in
In addition, the right hippocampus was significantly
correlated with the right thalamus among the controls, but
not among the patients. That is, among the controls, the
larger the right thalamic volume, the larger the right
hippocampus. This relationship did not hold for the left
side among controls or patients. However, a test of the
differences between patients and controls in the correlation
between the right thalamus and hippocampus was not
significant. The left and right caudate were more posi-
tively correlated with the left and right thalamus among
controls than among patients, although the differences
between controls and patients were not significant. Fur-
ther, the left superior frontal cortex was significantly
negatively correlated with left thalamus among the con-
trols but not among the patients. Thus, among the controls,
the larger the left thalamus, the smaller the left superior
frontal cortex. This did not hold for the right hemisphere.
Finally, the left and right frontal white matter correlated
differently with the left and right thalamus among patients
and controls. Frontal white matter was positively corre-
lated with the size of the thalamus, bilaterally, among
patients. However, frontal white matter was negatively
correlated with the size of the left thalamus among
controls, and uncorrelated with the size of the right
thalamus among controls. Total frontal gray matter did not
significantly correlate with the size of the thalamus among
patients or controls. Although the correlations with total
white matter on the left did not reach significance, the
difference in the correlations on the left between patients
and controls was close to significant (p ? .086).
Correlations of Thalamic Volumes with Clinical
Symptoms among Patients
Correlations between right, left, and total adjusted tha-
lamic volumes with four of the global negative symptom
ratings (affective blunting, alogia, apathy, asociality) and
total negative symptom score were not significant among
patients. However, there was some suggestion that atten-
tion was related to thalamic volume (total thalamic vol-
ume: rho ? ?.51, p ? .06). As one would predict, the
Table 1. Significant Correlations of Right and Left Volumes
of Brain Regions of Interest with Right and Left Thalamic
Volumes between Schizophrenics and Controls
(n ? 14)
(n ? 15) p-valuea
Left posterior temporal
Left superior frontal
Left frontal white matter (total)
Left frontal gray matter (total)
Right posterior temporal
Right superior frontal
Right frontal white matter
Right frontal gray matter (total)
aCorrelations were standardized using Fisher’s Z transformation to test for
whether the correlations between left and right thalamic volumes with ROIs were
different between groups, i.e., for patients versus controls.
bp ? .05,cp ? .01. These are tests of whether the correlations of right and left
thalamic volumes with ROIs are significant within group.
Table 2. Effect Sizesain Previous MR Imaging Studies Comparing Schizophrenic Male Patients with Normal Male Controls on
Total Thalamic Size
AuthorsMale sample sizeSlice size
Male effect sizes
Unadjusted volumesAdjusted volumes
Jernigan et al (1991)28sz
5 mm w/2.5-mm gaps
Andreasen et al (1994)
Signal intensity differences.75–1.0b
Flaum et al (1995)
3 mm w/1.5-mm gaps
7.5 mm; area assessed
Values not reported
Buchsbaum et al (1996)
Goldstein et al (1996)
3 mm, no gaps .35.80b
sz, schizophrenic subjects; nc, normal controls.
aEffect size was calculated according to Cohen (1977): mean volume of the controls minus the mean volume of the patients divided by the pooled standard deviation.
bSignificant difference in thalamic size at p ? .05.
Thalamus MRI in Schizophrenia653
direction of the effects suggests that the more severe the
attentional problems, the smaller the thalamic volume.
In contrast, there was a stronger relationship between
positive symptoms and thalamic volume. That is, the
correlation of adjusted total thalamic volume with total
positive symptom rating was statistically significant
(rho ? ?.69, p ? .04), reflecting mainly a correlation with
the left (rho ? ?.64, p ? .06) and to a lesser extent, the
right (rho ? ?.44, p ? .23). Results suggest that the
smaller the thalamic volume, the higher the positive
symptom severity. This correlation was primarily due to
two global scales, i.e., hallucinations and bizarre behavior.
The global hallucinations score significantly correlated
with left and right thalamic volumes (respectively, rho ?
?.52, p ? .057; and rho ? ?.60, p ? .02), and with total
thalamic volume (rho ? ?.59, p ? .02). We specifically
looked at the relationship of auditory hallucinations to
thalamic volumes, since they have been related to superior
temporal gyrus abnormalities (Barta et al 1990), and our
sample of patients showed significant structural abnormal-
ities in the superior temporal gyrus (Shenton et al 1992).
However, auditory hallucinations were not significantly
correlated to left or right thalamic volumes (respectively,
rho ? ?.38, p ? 22; and rho ? ?.46, p ? .13). The SAPS
bizarre behavior scale also significantly correlated with
total thalamic volume (rho ? ?.54, p ? .05), with a rho ?
?.50, p ? .07 for left and a rho ? ?.51, p ? .06 for the
right. In addition, left and right thalamic volumes were not
significantly correlated with the global delusions score nor
formal thought disorder, as measured by the SAPS or the
TDI in the total sample.
Male schizophrenic patients in this study did not have a
significant difference in thalamic volume (total, right, nor
left) compared to matched normal controls. Most of the
previous work on the thalamus in schizophrenia was
conducted in postmortem studies, some of which demon-
strated significant differences (Hempel and Treff 1959;
Baumer 1954; Dom et al 1981; Pakkenberg 1990),
whereas others did not (Rosenthal and Bigelow 1972;
Lesch and Bogerts 1984; Kelsoe et al 1988). It is difficult
to compare findings in this study with postmortem work,
given differences in diagnostic criteria, areas assessed, the
nature of the variables measured, and small sample sizes in
postmortem work. However, we were initially surprised
that there were no significant differences in thalamic
volumes in our sample of chronic male patients, given
recent MRI studies showing significant thalamic area and
volumetric reductions and image intensity abnormalities
(Andreasen et al 1990, 1994; Flaum et al 1995; Goldstein
et al 1996), especially in male schizophrenic patients
(Andreasen et al 1990; Goldstein et al 1996).
Although this study included a small sample size, thus
necessitating replication, we would argue that limitations
of sample size do not wholly account for the lack of
significance in this study. We calculated the approximate
effect sizes for males (ES ? mean volume of the con-
trols ? mean volume of the patients/pooled standard
deviation; Cohen 1977) in a number of the previous
imaging studies listed in Table 2: Andreasen et al 1994
(ES already provided); Flaum et al 1995; Buchsbaum et al
1996; Jernigan et al 1991; and Goldstein et al 1996. The
approximate ESs for each of these studies were, respec-
tively, .75, .55 (among male subjects), .10 (for adjusted
and unadjusted volumes), .11, and .35 (for unadjusted
volumes)/.80 (for adjusted volumes among male subjects).
The ES for our study, which was .27 for unadjusted
volumes and .50 for adjusted volumes, is in line with the
previous work. Thus, although the Andreasen and Flaum
studies, which were significant, had the highest ESs and
largest samples, sample size alone did not wholly account
for the range of effect sizes found across studies.
We understand that a larger sample size is important to
replicate findings. However, we would argue that the
differences in boundary definitions most likely contributes
to explaining the inconsistencies of findings reported in
the literature. For example, in our study, the lateral and
medial geniculates were included in the measurement of
the thalamus. However, in Andreasen et al 1990 and
Goldstein et al 1996, studies in which significant thalamic
volumetric differences among male subjects emerged, the
geniculates were not included in the thalamic definition,
given less ability to reliably measure the geniculates using
10-mm (Andreasen et al 1990) or 3-mm (Goldstein et al
1996) slices rather than the 1.5-mm slices used in this
study. We would argue that significant differences be-
tween schizophrenics and controls will emerge when only
specific nuclei are considered. This was true for previous
imaging studies, some of which had sample sizes similar
to ours. For example, Buchsbaum et al (1996) reported no
significant differences in total thalamic area (n ? 19 male
subjects; 1 female subject), but did report a significant
reduction in patients in the right posterior and left anterior
thalamic regions. Jernigan et al (1991) found a statistically
nonsignificant volumetric reduction among schizophrenic
patients compared to normals in a measure of total
diencephalon, including the thalamus. However, there was
a significant reduction in the anterior thalamus in patients
when considered apart from the total thalamus. Andreasen
et al (1994) also identified thalamic abnormalities in
particular thalamic regions, i.e., lateral and medial and the
right posterior and left anterior regions.
These studies are consistent with postmortem work, in
654C.M. Portas et al
which, specifically, the dorsomedial and anterior thalamic
nuclei were found to be abnormal in schizophrenia. The
dorsomedial and anterior nuclei are the primary thalamic
nuclei that connect with prefrontal cortex, a key region
found to be abnormal in schizophrenia (Weinberger et al
1986). In particular, the dorsomedial thalamic nucleus
constitutes the most prominent subcortical afferent to the
prefrontal cortex (Fuster 1989). Further, studies have
identified the pulvinar (Posner and Peterson 1990; Mesu-
lam 1990) and the reticular formation and intralaminar
nuclei (Kinomura et al 1996) as related to attentional
processing, a key deficit in schizophrenia (Mirsky et al
1991; Seidman 1983), which we found to be correlated
with thalamic size in patients. We would therefore argue
that when specific nuclei are considered, e.g., dorsome-
dial, anterior, pulvinar, reticular formation, and intralami-
nar nuclei, significant volumetric differences between
schizophrenic patients and controls will emerge.
Given the extensive connections between thalamic nu-
clei and other brain regions found to be abnormal in
schizophrenia, we tested for correlational differences
among patients versus among controls between thalamic
volume and specific prefrontal, temporal, cingulate, and
basal ganglia brain areas. Although we chose a limited
number of predicted ROIs, we had a small sample to test
for numerous correlational differences, and thus, the re-
sults should be considered exploratory. However, we
found that there was a greater number of “different”
significant correlations among patients compared to
among controls, in the left hemisphere, particularly re-
garding correlations of left thalamic volume with left
prefrontal white matter, than in the right hemisphere. This
suggests that, although there were no significant overall
volumetric differences between groups, there may be
subtle volumetric abnormalities and abnormal connectiv-
ity between specific cortical regions and thalamic nuclei,
particularly in the left hemisphere. This is consistent with
previous work showing abnormalities in the left hemi-
sphere among primarily male schizophrenic patients
(Bogerts et al 1990; Friston et al 1992). In fact, male
patients in this study demonstrated significantly more
abnormalities in left medial and superior temporal regions
(Shenton et al 1992) and left basal ganglia (Hokama et al
1995) than on the right. Although our sample of patients
exhibited increased left hemisphere abnormalities, tha-
lamic abnormalities in schizophrenic patients in general
may be bilateral, given recent work by Andreasen et al
(1994), reporting increased thalamic abnormalities espe-
cially in the right hemisphere.
We also found significantly different correlations
among patients than among controls between lateral ven-
tricular volume and thalamic volumes in the right and left
hemispheres. In patients only, the smaller the thalamic
volumes bilaterally, the greater was the ventricular cere-
brospinal fluid, bilaterally. This is consistent with an
earlier study reporting lateral ventricular enlargement
among primarily male schizophrenic patients, particularly
at the level of the anterior thalamus (Kelsoe et al 1988).
The relationship between ventricular and thalamic vol-
umes is interesting and may reflect decreased volume of
the interconnections of the thalamocortical fiber tracts, due
to thalamic abnormalities reflected in subtle volume alter-
ations. In fact, recent work by Andreasen et al (1994)
reported abnormalities in schizophrenic patients compared
to normal controls in the thalamus and its adjacent white
Thalamic volumes in these patients were also signifi-
cantly related to the severity of positive symptomatology,
i.e., bizarre behavior and hallucinations. Given the central
role of the thalamus in filtering or gating sensory infor-
mation to cortical brain regions (Jones 1985), and the
integration of cortical processing and behavior (Crosson
and Hughes 1987), it is not surprising that we found
significant relationships of thalamic volumes with bizarre
behavior, hallucinations, and thought disorder. The signif-
icant relationship of thalamic volume with bizarre behav-
ior is consistent with findings in these patients regarding
abnormalities in the basal ganglia, which were also related
to severity of bizarre behavior (Hokama et al 1995). This
fits nicely with animal studies that have demonstrated the
connectivity between caudate, putamen, and dorsomedial
thalamic nucleus (e.g., Kolb 1977). Further, one of the
major efferent pathways of the globus pallidus, found to
have the largest basal ganglia abnormality in these patients
(Hokama et al 1995), is to the ventral anterior thalamic
nucleus. In another previous analysis of these subjects,
right middle frontal gyrus was significantly correlated
with severity of delusions (Wible et al 1995). In conjunc-
tion with thalamic correlations with positive symptoms,
this suggests that perhaps thalamic connections with pre-
frontal regions,i.e., dorsolateral
(DLPFC), are affected in these schizophrenic patients.
This is consistent with a recent positron emission tomog-
raphy study suggesting a corticosubcortical imbalance in
schizophrenic patients that included DLPFC and anterior
thalamic regions (Friston et al 1992).
Variability in thalamic volumes among patients was
also significantly related to hallucinations. Again, the
thalamic role in gating sensory information seems to be
important for the production of positive symptoms in-
volved with sensory processing (i.e., hallucinations). Al-
though in previous work we found significant abnormali-
ties in the superior temporal gyrus (STG) that were related
to formal thought disorder (Shenton et al 1992), STG
volume was not significantly related to thalamic volumes
in this study. Thalamic volume was also not significantly
Thalamus MRI in Schizophrenia655
correlated with negative symptomatology. However, since
our sample of male patients was primarily characterized
by positive symptoms, the small variability in negative
symptom ratings may have resulted in decreased statistical
power to find significant correlations between thalamic
volume and negative symptom expression.
Findings on the relationships between symptomatology
and thalamic volumes must be viewed as exploratory,
given the small sample size in this study. Further, the
relationships between current symptomatology, perhaps
reflecting “state” characteristics, with structural brain
volume, reflecting “trait” characteristics, should be inter-
preted with caution. However, our findings are consistent
with previous work suggesting a primary role for the
thalamus in the production of positive symptoms (e.g.,
Crosson and Hughes 1987; Andreasen et al 1994) and
raise hypotheses for future studies.
Although the study presented here produced a number
of nonsignificant results regarding overall thalamic vol-
ume differences in patients compared to normal controls,
it raises some important issues for structural MRI studies
that examine the role of the thalamus in schizophrenia.
First of all, the thalamus, which is a composite of several
functionally different nuclei, is a difficult structure to
segment. One must make decisions about which nuclei
have to be included in the definition of the thalamus,
which may have implications for finding significant dif-
ferences between patients and controls. Further, the deci-
sion regarding which specific nuclei may be included in
the definition may be as dependent on practical concerns
as on theory, such as the resolution of the MR images and
the ability of one’s software to segment in alternative MR
planar views. Finally, the ability to segment specific areas
of the thalamus, rather than the thalamus as a whole, may
be more fruitful in identifying significant thalamic differ-
ences between schizophrenic patients and controls and
may provide a better understanding of the role of the
thalamus in schizophrenia.
This work was supported by grants from the National Institute of Mental
Health (NIMH) [MH K02 MH 01110 (MES), MH 1R29 50747 (MES),
MH R01 40,799 (RWM)], and by MERIT and Schizophrenia Center
Awards from the Department of Veterans Affairs (RWM), and NARSAD
Senior Investigator Award (RWM), and was initiated by JMG with
support from an NIMH Clinical Research Training Fellowship in
Biological Psychiatry, Department of Psychiatry, Harvard Medical
School at Massachusetts Mental Health Center (1991) and a Scientist
Development Award to JMG, K21 MH00976 (1992–1994).
Andreasen NC (1981): Scale for the Assessment of Negative
Symptoms (SANS). Iowa City, IA: Dept. of Psychiatry, Uni-
versity of Iowa College of Medicine.
Andreasen NC (1984): Scale for the Assessment of Positive
Symptoms (SAPS). Iowa City, IA: Dept. of Psychiatry, Uni-
versity of Iowa College of Medicine.
Andreasen NC, Ehrhardt JC, Swayze VW II et al (1990):
Magnetic resonance imaging of the brain in schizophrenia.
Arch Gen Psychiatry 47:35–44.
Andreasen NC, Arndt S, Swayze V II, et al (1994): Thalamic
abnormalities in schizophrenia visualized through magnetic
resonance image averaging. Science 266:294–298.
Barta PE, Pearlson GD, Powers RE, Richards SS, Tune LE
(1990): Auditory hallucinations and smaller superior tempo-
ral gyral volume in schizophrenia. Am J Psychiatry 147:
Baumer H (1954): Veranderungen des Thalamus bei Schizophre-
nie. J Hirnforsch 1:157–172.
Bogerts B, Meetz E, Schonfeld-Bausch R (1985): Basal ganglia
and limbic system pathology in schizophrenia: A morphomet-
ric study of brain volume and shrinkage. Arch Gen Psychiatry
Bogerts B, Ashtari M, Degreef G, Alvir JM, Bilder RM,
Lieberman JA (1990): Reduced temporal limbic structure
volumes on magnetic resonance images in first episode
schizophrenia. Psychiatry Res 35:1–13.
Buchsbaum MS, Someya T, Teng CY, et al (1996): PET and
MRI of the thalamus in never-medicated patients with schizo-
phrenia. Am J Psychiatry 153:191–199.
Cline HE, Lorensen WR, Kikinis R, Jolesz AF (1990): Three-
dimensional segmentation of MR images of the head using
probability and connectivity. J Comput Assist Tomogr 14:
Cohen J (1977): Statistical Power Analysis for the Behavioral
Sciences, rev. New York: Academic Press.
Crosson B, Hughes CW (1987): Role of the thalamus in
language: Is it related to schizophrenic thought disorder.
Schizophr Bull 13:605–621.
De Armond SJ, Fusco MM, Dewey MM (1989): Structure of the
Human Brain. A Photographic Atlas. New York: Oxford
Dolan RJ, Bench CJ, Liddle PF, et al (1993): Dorsolateral
prefrontal cortex dysfunction in the major psychoses: Symp-
tom or disease specificity? J Neurol Neurosurg Psychiatry
Dom R, De Saedelar J, Bogerts B, Hopf A (1981): Quantitative
cytometric analysis of basal ganglia in catatonic schizophren-
ics. In: Perris C, Struwe G, Jansson B (eds). Biological
Psychiatry. Amsterdam: Elsevier, pp 723–726.
Duvernoy HM (1991): The Human Brain: Surface, Three-
Dimensional Sectional Anatomy and MRI. New York: Spring-
Falkai P, Bogerts B, Rozumek M (1988): Limbic pathology in
schizophrenia: The entorhinal region—A morphometric
study. Biol Psychiatry 24:515–521.
Fisher RA (1958): Statistical methods for research workers.
Edinburgh: Oliver & Boyd.
Flaum M, Swayze VW II, O’Leary DS, et al (1995): Effects on
diagnosis, laterality, and gender on brain morphology in
schizophrenia. Am J Psychiatry 152:704–714.
Friston KJ, Liddle PF, Frith CD, Hirsch SR, Frackowiak RSJ
656 C.M. Portas et al
(1992): The left medial temporal region and schizophrenia: A
PET study. Brain 115:367–382.
Fuster JM (1989): The Prefrontal Cortex: Anatomy, Physiology,
and Neuropsychology of the Frontal Lobe, 2nd ed. New
York: Raven Press.
Gold JM, Weinberger DR (1991): Frontal lobe structure function
and connectivity in schizophrenia. In: Kerwin R, Dawbarn,
McCulloch, Tamminga, CA, editors. Neurobiology and Psy-
chiatry. Cambridge: Cambridge University Press.
Goldstein JM, Seidman LJ, Goodman JM, et al (1996): The
impact of sex on cognition and structural brain abnormalities
in schizophrenia. Biol Psychiatry 39:575.
Groenewegen HJ, Berendse HW, Wolters JG, Lohman AHM
(1990): The anatomical relationship of the prefrontal cortex
with the striatopallidal system, the thalamus and the amyg-
dala: Evidence for a parallel organization. In: Uylings HBM,
Van Eden CG, De Bruin JPC, Corner MA, Feenstra MGP,
editors. Progress in Brain Research: Vol 85. The Prefrontal
Cortex. Amsterdam: Elsevier, pp 105–118.
Haines DE (1991): Neuroanatomy: An Atlas of Structures Sec-
tions and Systems. Baltimore: Urban & Schwarzenberg.
Hempel KJ, Treff WM (1959): Quantitative untersuchungen uber
relative zell- und Zellkernvolumina bei klinisch gesunden and
schizophrenia. J Hirnforsch 4:412–454.
Hokama H, Shenton ME, Nestor PG, et al (1995): Basal ganglia
volume abnormalities in schizophrenia: An MRI study. Psy-
chiatry Res Neuroimaging 61:209–229.
Jernigan TL, Zisook S, Heaton RK, Moranville JT, Hesselink JR,
Braff DL (1991): Magnetic resonance imaging abnormalities
in lenticular nuclei and cerebral cortex in schizophrenics.
Arch Gen Psychiatry 48:881–890.
Johnston MH, Holzman PS (1979): Assessing Schizophrenic
Thinking. San Francisco, CA: Jossey-Bass.
Jones ED (1985): The Thalamus. New York: Plenum Press.
Kelsoe JR, Cadet JL, Pickar D, Weinberger DR (1988): Quan-
titative neuroanatomy in schizophrenia. Arch Gen Psychiatry
Kikinis R, Jolesz FA, Gerig G, et al (1990): 3D morphometric
and morphometric information derived from clinical brain
MR images. In: Hohne KH, Fuchs H, Pizer SM, et al, editors.
3D Imaging in Medicine. Berlin: Springer-Verlag, pp 441–
Kinomura S, Larsson J, Gulyas B, Roland PE (1996): Activation
by attention of the human reticular formation and thalamic
intralaminar nuclei. Science 271:512–515.
Kolb B (1977): Studies on the caudate-putamen and the dorso-
medial thalamic nucleus of the rat: Implications for mamma-
lian frontal-lobe functions. Physiol Behav 18:237–244.
Lesch A, Bogerts B (1984): The diencephalon in schizophrenia:
Evidence for reduced thickness of the periventricular grey
matter. Eur Arch Psychiatry Neurol Sci 234:212–219.
Mesulam MM (1990): Large-scale neurocognitive networks and
distributed processing for attention, language, and memory.
Ann Neurol 28:597–613.
Mirsky AF, Anthony BJ, Duncan CC, Ahern MB, Kellam SG
(1991): Analysis of the elements of attention: A neuropsy-
chological approach. Neuropsychol Rev 2:109–145.
Pakkenberg B (1990): Pronounced reduction of total neuron
number in mediodorsal thalamic nucleus and nucleus accum-
bens in schizophrenics. Arch Gen Psychiatry 47:1023–1028.
Pakkenberg B (1992): The volume of the mediodorsal thalamic
nucleus in treated and untreated schizophrenics. Schizophr
Pandya DN, Yeterian EH (1990): Prefrontal cortex in relation to
other cortical areas in the rhesus monkey: Architecture and
connections. In: Uylings HBM, Van Eden CG, De Bruin JPC,
Corner MA, Feenstra MGP, editors. Progress in Brain
Research: Vol 85. The Prefrontal Cortex. Amsterdam:
Elsevier, pp 63–94.
Posner MI, Peterson SE (1990): The attention system of the
human brain. Ann Rev Neurosci 13:25–42.
Roberts MP, Hanaway J (1971): Atlas of the Human Brain in
Section. Philadelphia: Lea & Fibiger.
Rosenthal R, Bigelow LB (1972): Quantitative brain measure-
ments in chronic schizophrenia. Br J Psychiatry 121:259–
Seidman LJ (1983): Schizophrenia and brain dysfunctions: An
integration of recent neurodiagnostic findings. Psychol Bull
Seidman LJ, Faraone SV, Goldstein JM, et al (1996): Reduced
subcortical brain volumes in nonpsychotic siblings of schizo-
phrenic patients. Biol Psychiatry 39:602.
Shenton ME, Kikinis R, Ferenc JA, et al (1992): Abnormalities
of the left temporal lobe and thought disorder in schizophre-
nia. N Engl J Med 327:604–612.
Silverweig DA, Stern E, Frith C, et al (1995): A functional
neuroanatomy of hallucinations in schizophrenia. Nature
Spitzer RL, Endicott J (1978): Schedule for Affective Disorders
and Schizophrenia, 3rd ed. New York: New York State
Psychiatric Institute, Biometrics Research.
Wechsler D (1981): WAIS-R Manual: Wechsler Adult Intelli-
gence Scale, Revised. New York: Harcourt Brace Jovanovich.
Weinberger DR, Berman KF, Zec RF (1986): Physiologic
dysfunction of dorsolateral prefrontal cortex in schizophrenia.
Arch Gen Psychiatry 43:114–124.
Wible CG, Shenton ME, Hokama H, et al (1995): Prefrontal
cortex: A quantitative magnetic resonance imaging study.
Arch Gen Psychiatry 52:279–288.
The thalamus (left and right thalami) was defined on
20–21 consecutive slices rostrally to caudally. Since it is
crucial to obtain a reliable segmentation of the ROI, we
produced, in each case, a “slice by slice” illustration of
ROI definition, along with the anatomical landmarks.
Figure 3 is a sample case (1 control subject, coronal plane
only) in which the anatomical landmarks are displayed.
Note that some structures are described only as they
appear in the first of the sequentially presented images and
may not be reported in the following section (to avoid
constant repetition of anatomical landmarks).
Thalamus MRI in Schizophrenia657
Description of ROI on Coronal Slices, with
Most anterior slice #1: The most anterior part of the
thalamus is shown here. Note the onset of the ventralis
anterior nucleus just dorsal to the hypothalamus. Bound-
aries are: laterally, the internal capsule; dorsally, the main
body of the lateral ventricle; and medially, the third
Slice #2: Note the anterior nuclei (reference atlas:
Duvernoy 1991 pages 116, 118; Roberts and Hanaway
1971 Table 17).
Slice #3: The mammillary bodies are clearly visible
(reference atlas: Roberts and Hanaway 1971 Table 17).
Note also the size of the caudate nucleus. The thalamus is
Slice #4: Note the merger of the mammillary bodies of
the hypothalamus (reference atlas: Roberts and Hanaway
1971 Table 19), the globus pallidus, and the putamen. The
lateral profile of the thalamus presents a small partial
Slice #5: Note the mammillothalamic tract (reference
atlas: Roberts and Hanaway 1971 Table 19; Duvernoy
1991 page 125).
Slice #6: Note the interpeduncular fossa and the appear-
ance of the mammillothalamic tract (reference atlas: Rob-
erts and Hanaway 1971 Table 19; Duvernoy 1991 page
125. See also Duvernoy 1991 page 126 and Roberts and
Hanaway 1971 Table 21).
Slice #7: On slice #14, note the globus pallidus on left
subject (reference atlas: Roberts and Hanaway 1971 Table
Slice #8: On slice #13, note the extension of the
Slice #9: The medial lemniscus (ML) is clearly evident
(reference atlas: Duvernoy 1991 pages 133, 131; Roberts
and Hanaway 1971 Table 23), as are the zona incerta
hypothalami (reference atlas: Duvernoy 1991 pages 123,
125, 127, 129), the subthalamic nucleus (Duvernoy 1991
pages 123, 128), the crus cerebri (Duvernoy 1991 page
127), and the putamen (reference atlas: Roberts and
Hanaway 1971 Table 23; Duvernoy 1991 page 128). The
contour of the thalamus appears well resolved; there is a
modest partial volume effect. The lateral geniculate nuclei
(LGN) is not visible anymore; the optic tract appears
(reference atlas: Roberts and Hanaway 1971 Tables 23,
21; Duvernoy 1991 pages 121, 129). Note the volume of
the massa intermedia.
Slice #10: The brain stem is clearly evident, as is the
ML (reference atlas: Duvernoy 1991 pages 133, 131;
Roberts and Hanaway 1971 Table 23). Note also the
substantia nigra lateral to the red nucleus (reference atlas:
Roberts and Hanaway 1971 Tables 23, 21), the interpe-
duncular fossa (Roberts and Hanaway 1971 Table 28), and
the putamen (reference atlas: Roberts and Hanaway 1971
Table 23; Duvernoy 1991 page 128).
Slice #11: Note the massa intermedia of the thalamus
(“interthalamic adhesion”) (reference atlas: Roberts and
Hanaway 1971 Table 21). The interpeduncular fossa is just
visible (reference atlas: Roberts and Hanaway 1971 Tables
23, 21). Within the thalamus, the dorsomedial nucleus
(DMN) appears; the white matter can be seen that sepa-
rates the main body of the thalamus from the LGN. Note
also the red nucleus (reference atlas: Duvernoy 1991 pages
132, 130; Roberts and Hanaway 1971 Table 28), the
Figure 3. A case example of the segmentation of thalamic
boundaries on 21 coronal slices.
658 C.M. Portas et al
myelinated medial longitudinal bundle in the brain stem Download full-text
(MLB) (reference atlas: Roberts and Hanaway 1971 Ta-
bles 23, 25), the ML (reference atlas: Roberts and Han-
away 1971 Tables 23, 25; Duvernoy 1991 pages 133, 131),
and the substantia nigra (Roberts and Hanaway 1971
Table 23). See also De Armond et al 1989 page 110.
Slice #12: Note the DMN, the stria medullaris of the
thalamus, and the putamen (reference atlas: Duvernoy
1991 pages 130, 132; Haines 1991 page 129). The lateral
and inferior limit of the thalamus can be seen. In this slice,
MLB emerges bilaterally in the brain stem (reference
atlas: Roberts and Hanaway 1971 Table 25). The profile of
the medial geniculate nuclei (MGN) in the background of
the cerebrospinal fluid (CSF) is not visible anymore
(reference atlas: Duvernoy 1991 pages 132, 236). It is
important to try to locate the subthalamic nucleus (refer-
ence atlas: Roberts and Hanaway 1971 Table 23).
Slice #13: Note the silhouette of the third ventricle, the
stria medullaris of the thalamus, the cerebral peduncles,
and the hippocampus (reference atlas: Duvernoy 1991
page 236). The DMN is now clearly evident; the putamen
is evident on the left subject. To delineate the brain stem
from the thalamus (when the habenula is not visible) it is
necessary to draw a line from the hypothalamic sulci of the
third ventricle to the deepest indentation of CSF, ventrally
and medially to the MGN (reference atlas: Duvernoy 1991
Slice #14: To delineate the thalamus from white matter,
we decided to include arbitrarily, but systematically 50%
of the partial volume visible on the lateral contour of the
thalamus. Moreover, several anatomical atlases were com-
paratively used as a reference for the segmentation of this
area (atlases: Duvernoy 1991 pages 141, 148, 240; Roberts
and Hanaway 1971 Table 25). Note the DMN and the
brain stem closing on the gap between the two thalami to
form the third ventricle.
Slice #15: Note the LGN, the MGN, the DMN, the
habenula, and the posterior commissure (reference atlas:
Duvernoy 1991 pages 140, 142, 240). To resolve the brain
stem from the thalamus at this level it is necessary to draw
a line dorsally from the habenula to the deepest indenta-
tion of the CSF, ventrally and medially to MGN. The
lateral limit is marked by the internal capsule.
Slice #16: Note the aqueduct and the periaqueductal
gray matter, and the clear presence of LGN and MGB as
ventral bumps on the pulvinar (reference atlas: Duvernoy
1991 page 243; Haines 1991 page 125; Roberts and
Hanaway 1971 Table 27).
Slice #17: The brain stem is clearly delineated from the
pulvinar. Note the fourth ventricle, the aqueduct, and the
hippocampal sulcus. Boundaries were: laterally, temporal
stem; ventrally, CSF.
Slice #18: Note the fourth ventricle and the pineal gland
(slightly visible in the background of the CSF of the
cistern of the great cerebral vein). The pulvinar appears
ball-shaped (reference atlas: Roberts and Hanaway 1971
Slice #19: Note that the superior and inferior colliculi
appear as bilateral globes. The pulvinar does not appear
clearly along with the tail of the caudate nucleus (refer-
ence atlas: Duvernoy 1991 page 146; Roberts and Han-
away 1971 Table 29).
Slice #20: Boundaries are: dorsally, the lateral ventri-
cles; ventrally, the cistern of the great cerebral vein;
laterally, temporal stem; medially, CSF (reference atlas:
Duvernoy 1991 page 146). Note also the superior colliculi.
The pulvinar again, as in the previous slice, appears now
as a rounded structure separated from the parahippocam-
pal gyrus by the CSF of the cistern of the great cerebral
Slice #21: A small portion of the pulvinar appears
between the crus fornix and the CSF of the cistern of the
great cerebral vein (also called superior cistern) (reference
atlas: Duvernoy 1991 pages 146, 148, 244). The crus
fornix and the splenium of the corpus callosum cross
diagonally. Note also the lateral ventricles, the cistern of
the great cerebral vein, the hippocampus, the parahip-
pocampal gyrus, and the temporal stem.
Thalamus MRI in Schizophrenia 659