Schizophrenia Research, 5 (1991) 1033113

Elsevier 103

SCHIZO 00172

Application of automated MRI volumetric measurement techniques to the ventricular system in schizophrenics and

normal controls*

Martha E. Shenton’, Ron Kikinis’, Robert W. McCarley’, David Metcalf2, James Tieman and Ferenc A. Jolesz2

Harvard Medical School, ‘Deparrment oj” Psychiatric (Brockron VAMC and Massachusetts Mental Health CenierJ and ‘Department

of Radiology (Brigham and Women’s Hospital), Brockton. MA, U.S.A.

(Received 4 September 1990. revised received 13 December 1990. accepted 20 December 1990)

As an initial approach to computer-automated segmentation of cerebral spinal fluid (CSF) vs. brain parenchyma in MR scans, and the transformation of these data sets into volumetric information and 3D display, we examined the ventricular system in a sample of ten chronic schizophrenics with primarily positive symptoms and 12 normal subjects. While no significant differences were noted between groups on volumetric measures of ventricular brain ratio or lateral ventricle size, normals showed a pattern of left > right lateral ventricular volume asymmetry not present in the schizophrenics. Within the schizophrenic group, departure from the normal left > right pattern was highly correlated with thought disorder.

Key words: Magnetic resonance imaging; Ventricular volume; Magnetic resonance imaging volumetric measurement

INTRODUCTION

With the advent of new in vivo brain imaging techniques such as computed axial tomography (CT) and magnetic resonance imaging (MRI), it has now become possible to test longstanding hypotheses of abnormal morphometric features in the schizophrenic brain (e.g., Bleuler I9 I l/ 1950; Kraepelin, 1919). Many CT studies in schizophre- nia have reported enlargement of the lateral and third ventricles and of the sulci, findings which, though nonspecific, may indicate tissue loss or

Correspondence to: R.W. McCarley. Department of Psychia-

try-l I6A. 940 Belmont Street. Brockton. MA 02401, U.S.A. (Tel.: (508) 583-4500 X367).

*Supported by NIMH Research Scientist Development

Award KOI-MH00746602 (M.E.S.), the Department of Veter- ans Affairs (R.W.M.). NIMH 40,799 (R.W.M.). the Common-

wealth of Massachusetts Research Center (R.W.M.), NIH

Research Career Development Award K04-NS01083 (F.A.J.), and the Swiss National Foundation (R.K.).

failure of development (e.g., Johnstone et al., 1976; Weinberger et al., 1979; Nasrallah et al., 1982; see also discussion in McCarley et al., 1989, and review in Shelton and Weinberger, 1986). Similar CSF

space abnormalities have been reported in MRI studies in schizophrenia (e.g., Besson et al., 1987; Kelsoe et al., 1988; Andreasen et al., 1990; Suddath et al., 1990). That negative findings have also been reported (e.g., Smith et al., 1984, 1987; Mathew et al., 1985; Rossi et al., 1990) suggests that particular features of abnormal brain morphology may be present in only a subgroup(s) of schizophrenic patients and/or may reflect methodological differences.

A clear desiderata in MRI ventricular studies would be to have automated segmentation tech- niques that fully exploit volumetric information, thus maximizing sensitivity and reducing potential inconsistencies inherent in manually performed linear/area measurements. We here summarize the methodology, and present initial results, of what is to our knowledge the first use of automated,

0920-9964!91/$03.50 ~‘1 1991 Elsevier Science Publishers B.V.

104

computerized MR image processing techniques to make volumetric measurements and three dimen- sional displays of brain and CSF spaces in a sample of chronic schizophrenics and normal con- trols. Our goals were thus two-fold: (1) to describe specific methodological advances in extracting and displaying information from MR scans, and (2) to illustrate the application of these new image pro- cessing techniques in a small subject sample.

METHODS

Subjects

The schizophrenic group was comprised of ten chronic, male, right-handed, neuroleptic medicated schizophrenics (median daily dose equivalent to 860 mg of chlorpromazine), recruited from in- patient wards at the Brockton Veterans Admin- istration Medical Center. Information from the Schedule for Affective Disorders and Schizophre- nia (SADS) (Spitzer et al., 1978) and from chart reviews was used to make DSM-III-R (APA, 1987) diagnoses. Further criteria for subject selection were: between the ages of 20 and 55, right-handed, no history of ECT (electroconvulsive shock treat- ment), no neurological illness, no significant alco- hol or drug abuse in the last 5 years (using DSM- III-R criteria and determined from a series of interview questions regarding the amount of alco- hol consumption over various time periods), and no medications with known effects on brain MRI. The descriptive data for this sample are presented in Table 1 which shows that the patients were chronic and had a preponderance of positive symptoms.

12 normal control subjects, carefully screened for disease factors that could affect brain function, including alcohol abuse, were recruited from among hospital staff and were matched for sex, handedness and age. They were receiving no medi- cation, and had no history of psychiatric or neuro- logical illness in themselves or in their first degree relatives. There were no significant differences be- tween the two groups in height, weight, head circumference, or social-economic status of family of origin (see Table 1). Differences were present in education (p < 0.002) and there was a non-signifi- cant trend for normal controls to have a higher estimated verbal IQ (based on Information Sub-

scale; p lO.058) (see Table 1). All subjects gave informed consent.

MR imaging protocol

Instrumentation. The MRI data were acquired using conventional long repetition rate double echo spin echo sequences on a GE Signa 1.5 Tesla MRI System (General Electric Co., Milwaukee, WI) with a standard 30 cm head coil. Multi-axial, long TR, double echo series were performed with either TR = 2800 TE 30/80 ms, slice thickness 5 mm, and 2.5 mm gap (50%), for the patient group, or TR= 2000 TE 30/80 ms, interleaving 5 mm, no gap, for the normal control group. In these double echo images the second echo provides high contrast for CSF while the combination of first and second echoes provides the best contrast between gray and white matter. The protocols were different because of upgrades in the hardware and software of the MR scanner which followed the data collec- tion on the schizophrenic patients. This upgrade allowed the acquisition of more data, and hence better spatial resolution, without sacrificing signal- to-noise ratio. Careful visual examination of the images from the two protocols revealed no visually apparent differences in the contrast between CSF and brain in the two sets of images. However, since the two protocols were different, we thought it important to test for any measurable effect of parameters by comparing them in the same subject. Thus for control purposes we acquired both proto- cols in one subject and then followed the same procedure of segmenting the images into tissue classes (e.g., brain, CSF, connective tissue/vessel ~ see below) for both acquisitions. The differences between these two protocols on all of the absolute and relative volumetric measures (i.e., % of in- tracranial volume), for all tissue classes, were less than 1%. Further independent data showing the comparability of the two protocols is presented in Kikinis et al. (1990). We emphasize nevertheless that these differences were taken into account in making calculations regarding the number of slices needed to obtain the 120 mm slab that contained the ventricular system. For the no gap protocol, the distance between slice centers was 5 mm and we used 24 slices to comprise the 120 mm slab (5 x 24= 120); for the 2.5 mm gap protocol, the distance between slice centers was 7.5 mm so we used 16 slices to comprise the 120 mm slab

105

TABLE I

Descriptive dam (mean ~standard deviutionj ,for schizophrenic and normal control subjects

Schizophrenics Normal controls

n

Age Height (inches)

Weight (Ibs.)

Head circumference” (cm)

Educationi’

Social class family of origin’

Wais-Verbal IQd

Age of onset (years)

Duration of illness (months)

“/o time in hospital’

Mostly positive symptoms’

Mostly mixed symptoms

Mostly negative symptomsg

Total thought disorder”

10 12

40.9OkO6.15 40.3OkO9.27

70.50 i 02.73 70.21 kO3.32

179.40 * 30.90 180.6Ok26.76

5s.75*01.59 57.79k02.23

11.75~01.18 16.58kO4.14

3.20~01.03 2.82*01.33(n= 11)

100.60& 18.14 119.20i22.76(n= 10)

2O.lOkO5.38

23 1 .OO + 74.26

63.10k02.26

7

3

0

90.02 k 64.25

* = Head circumference assessed using nasion. inion, and preauricular landmarks.

“/= 3.56, df=20. p<O.O02. ‘t = ~ 0.73. df = 19, p < 0.474 (SES-based on the Hollingshead Index (Hollingshead, 1965) where I = businessiprofessional, 2 =

medium business. 3 = skilled, 4 = semiskilled, and 5 = unskilled.

dl = 2.02, df= 18. pi 0.058.

‘Actual time in months spent in the hospital divided by duration of illness defined by the time span between first hospitalization

and the present.

‘Ratings done using the Scale for Assessment of Positive Symptoms (SAPS, Andreasen, 1984).

eRatings done using the Scale for the Assessment of Negative Symptoms (SANS, Andreasen. 1981).

hAssessment of total thought disorder using the Thought Disorder Index (TDI) (Solovay et al., 1986).

(7.5 x 16 = 120; interpolations were made for data between slices). We emphasize also that all volumes computed on the gap protocol were adjusted by extrapolation to reflect true brain volume of the tissue class under consideration, as is standard using a single pass Spin Echo technique (Pavlicek et al., 1984). The no gap is simply a spin echo sequence with two passes, with the second pass covering the area not included in the first. In a single pass spin echo, the values in the gap are based on interpolations from information from the slice below and above the gap. The start and end point for the 120 mm used the same landmarks, as described below.

Landmurks and normalization. As an initial approach to brain volumetric analyses we focused on a brain slab selected to include the entire ventricular system, including the third and fourth ventricles. This slab was defined as beginning 15 mm below the internal acoustic canal (IAC), a convenient and objective reference point, and ex- tending 120 mm rostrally. Measurements were

made on all intracranial contents within this slab and all were corrected for head size.* For conve- nience in reference we will henceforth refer to this 120 mm slab as the intracranial cavity (ICC).

Computer assisted volumetric measures. Axial images were processed on a DEC Micro Vax (Digital Equipment Corporation, Westborough,

*The formula for normalization for head size was:

[a:(3 - $)/a(3 -a*)]; where a, = the length of the structure (from the IAC to the top of the 120 mm mark) divided by the

radius of the sphere (from IAC to vertex) for each subject. The

mean was then derived for all ‘ai’s (there were no differences

between normal controls and schizophrenics and so the num-

bers were pooled, n = 22). The mean for all subjects was then

used as the term in the denominator. All anatomical measures

for each subject were then multiplied by this individually

specific scale factor. When calculations were performed for the non-normalized data, the results were virtually the same, i.e.,

the same measures discriminated between groups. Further.

when volume was calculated using relative volume. i.e., O/o of

intracranial volume, again, the same measures discriminated

between groups. For the sake of brevity we present only the normalized data.

106

MA) with ITEX image processing hardware using: (1) a semi-automated algorithm to identify the intracranial cavity (ICC), (2) a fully automated separation of its contents into brain, CSF, and vessels/tissues, and (3) a semi-automated separa- tion of the CSF into subarachnoidal space and ventricular space (entire system contained in 120 mm slab). The 3D representation of the ven- tricular system was carried out using a connectivity program (see below) in conjunction with a TAAC board on a SUN workstation for visual display and operator interaction.

STEP 1 An automated edge detection algorithm was used to identify the intracranial cavity (ICC) for each subject. This generated short line segments which were connected automatically under the assumption that the ICC has a curvilinear shape. The outline of the ICC was then used as a mask for the remaining steps.

STEP 2 The contents of the ICC were divided into brain, CSF, and connective tissue/vessels based on a multivariate look-up table (Udupa et al., 1982). This procedure uses information from both the first echo of the spin echo sequence (see Fig. I B, column 1) and the second echo of the spin echo sequence (see Fig. 1 B, column 2) in making tissue character- izations. For example, part of the algorithm used the rule that what is bright on both the first and

second echo is initially classified as CSF. Then, in addition, a local histogram technique was used to compensate for fluctuations in signal intensity (see Sandor et al., 1988). The absolute volumes of each pixel (voxels) were then used to calculate volumes, automatically, for all slices and for each tissue class (for more details see Kikinis et al., 1990; Sandor et al., 1990). Thus at the end of the analysis, a tabulation of absolute and relative volumes was provided for all tissue classes and this information was available for every slice and also for specified 3D brain tissue/structures.

STEP 3 CSF was further differentiated into sub- arachnoidal and intraventricular regions by placing the cursor manually on intraventricular spaces and using a 2D region growing algorithm to change all of the contiguous CSF labeled pixels in these regions into a different label and color code on the display (see Fig. 1A). This process was com- pletely automated. The only manual steps per- formed involved placing the cursor on the left and right lateral ventricles (manually separated by a

midline division), and on the subarachnoidal space. The automation of the region growing algorithm is desirable because it is: (1) repeatable, and (2) the computer is ‘blind’ to subject group and to left vs. right lateral ventricles. The color coded segmen- tation of tissue can be seen in Fig. 1 B (see column 3) where the final tissue characterization for three different axial slices is presented. This figure serves as an illustrative example only since the final product is the volume measures based on all slices that comprise the ventricular system, as well as the 3D display of the ventricular system (see below; see also Fig. IA). Ventricular brain ratio (VBR) was computed by dividing the volume of the lateral ventricles by the volume of the ICC contained in the 120 mm slab. Brain/CSF ratio was also com- puted (atrophy measure).

STEP 4 3D reconstructions were generated from the labeled maps that were used for volumetric analysis; this step involved using a connectivity program that marked the surface voxels and used the dividing cubes algorithm of Cline et al. (1987, 1988) to then generate a model of the surface. This procedure is very different from previous techniques for extracting information from MR data sets in that it is able to use data from an arbitrary number of 2D slices to create a 3D representation. The process, and its end result, is analogous to tightly fitting a rubber sheet over the surface of a structure.

To correlate the extent of agreement of results from this new approach with previous measures, we also calculated area measurements for the lateral ventricles using the axial slice which showed the largest body of the lateral ventricles, as done for our CT work (McCarley et al., 1989). The outline of the intracranial cavity, and the body of the lateral ventricle, was then traced and digitized (wire-grid electromagnetic digitizer, SPD-Series Graphics Tablet Scriptel Corporation, 1985). In- terrater reliability for this procedure for two raters was r=0.929 (n= 20, p<O.OOl). All MR images were evaluated by a clinical neuroradiologist who reported no gross abnormalities for any of the subjects.

RESULTS

Fig. 1A illustrates the results of the image process- ing and three-dimensional ventricular system re-

107

construction possible with this technique. The three different perspectives on the ventricular system illustrate the capability for rotations in 3-space. Of note is the detail of the rendering of the island in the third ventricle produced by the thalamic massa intermedia and the relatively modest in vivo size of the temporal horn portion of the lateral ventricle. Fig. 1B illustrates the information gleaned from T2 weighted and proton density images when they are segmented using the special- ized image processing hardware and software that we described in the Methods section.

As shown in Table 2, there were no significant differences between schizophrenic and control groups for: (1) CSF volume (total, subarachnoidal or ventricular CSF), (2) brain volume, or (3) volume of connective tissue/vessels. There were also no significant differences noted for brain/CSF ratio or VBR (ventricles/ICC x 100). Similarly, the manual linear VBR measurements showed no group differences and there was a high correlation between values obtained using the automated volumetric and the manual linear measurements (r=0.91, p<O.OOl).

A comparison of left vs. right lateral ventricle volumes did however reveal that the normal sub- jects showed a left > right asymmetry (1 l/12 sub- jects, ~~0.05, .Mann Whitney U test; and t= - 4.40, df= 11, p_<O.OOl), whereas the schizo- phrenics did not show a statistically significant left > right asymmetry and, in fact, 3/10 schizo- phrenics showed a right > left pattern (see Table 3). The schizophrenics also showed a statistically sig- nificantly larger variance on the left-right measure of asymmetry (F,,, = 7.08, df=9, 11, p<O.Ol).

Our previous CT study in schizophrenia (McCarley et al., 1989) found a significant associa-

tion within the schizophrenic group between ab-

normalities of left temporal lobe CSF spaces and the Thought Disorder Index (TDI) (Solovay et al., 1986). We therefore thought it important in the present study to examine the relationship, within the schizophrenic group, between the trend toward an abnormal left-right lateral ventricular system asymmetry and TDI scores. These two variables were highly and statistically significantly correlated

(see Fig. 2; Spearman’s rho = - 0.93, p I 0.01, two- tailed test); schizophrenics who showed more left >right asymmetry (the pattern shown by nor- mals) had lower thought disorder while those with right >left asymmetry (a pattern unlike the nor- mals) showed higher thought disorder.

DISCUSSION

The goals of this study were two-fold: (1) to present automated brain-CSF voxel segmentation techniques and their use in volumetric measure- ments and 3D reconstructions of MR images, and (2) to present illustrative, initial data from the application of these techniques to a small sample of schizophrenic patients and normal control sub- jects. We believe that these techniques represent an important new approach to quantifying infor- mation from MR images, while three dimensional reconstruction allows the visual depiction of com- plex structures. Previously applied to patients with neurological disorders (Kikinis et al., 1990), these new techniques also appear to be useful in studies of schizophrenics and, as an initial illustrative application, we have examined the ventricular sys-

Fig. I. A: 3D reconstruction of the ventricular system based on 2 mm axial slices from a 35year-old normal control subject using

the image processing technology described in this paper and illustrating the capacity for rotations in space. Note in particular the

clear depiction of the third ventricle space occupied by the massa intermedia and the relatively small volume of the in vivo temporal

horns of the lateral ventricles. In post-mortem casts (Last and Tompsett, 1953) the temporal horns often appear enlarged relative

to the in vivo young normal control subjects presented here. This enlargement may result from the fact that: (1) casts are made from formalin-fixed brains, a process that leads to shrinkage of the brain tissue (Filipek et al., 1989); (2) forcing the casting material

into the ventricles may artifactually enlarge the temporal horns; and (3) casts are often done on brains from elderly or diseased

subjects. LL = body of the left lateral ventricle; RL = body of the right lateral ventricle; FH = frontal horn of lateral ventricle; TH = temporal horn of lateral ventricle; 3 = third ventricle; 4 = fourth ventricle; D = dorsal; V = ventral; A = anterior; and P = posterior

aspects of ventricular system. B: segmentation of MRI images into brain, CSF (intraventricular and subarachnoidal), and vessels, The color codes are: navy blue = subarachnoidal CSF; green = 4th ventricle; the lighter blue on viewer’s left = right lateral ventricle;

the slightly darker blue on the viewer’s right =left lateral ventricle; brown=gray matter; tan= white matter; and red=vessels.

Anatomical features to be noted are: the fourth ventricle/pans (top row); aqueduct/midbrain and temporal horn (middle row); and

body of lateral ventricles (bottom row).

Le\

4th

I v

Tem

pora

l &

O

ccip

ital

Hor

ns

of L

ater

al

Ven

tric

les

Slic

e S

how

ing

Larg

est

Are

a of

La

tera

l V

entr

icle

110

TABLE 2

Absolute volume (cm31 and relative volume of brain and CSF spaces in schizophrenics and normal control subjects (120 mmJ

Group Anatomical

region

NCL

SZ

NCL

SZ

NCL

SZ

NCL

SZ

Total ICC-120 mm slab

(Intracranial cavity)

Brain

CSF (Subarachnoidal)

CSF

(Intraventricular)

Absolute volume”

mean (SD)

p value Relative volume (%)

1,505.03 + 119.82 0.383 100.00

1,545.35+084.88 100.00

1,212.23f092.49 0.148 80.62

1,267.95+078.53 82.05

237.02+041.38 0.665 15.72

229.98kO31.69 14.88

30.06f014.98 0.147 1.96

22.26kOO7.12 I .44

Subcomponents of intraventricular CSF (CSFVENT) Relative volume of CSFVENT

NCL

sz

NCL

sz

NCL

sz

NCL

sz

(a) Lateral ventricles

(b) Left ventricle

(c) Right ventricle

(d) 3rd and 4th ventricles

24.21 kO12.72

17.06,006.42

13.02+006.85

9.03 kOO3.64

11.19+005.92

8.03+003.17

5.76kOO2.29

5.2OkOOl.05

0.123 [I.581

[l.ll]

0.114 [0.85]

[0.58]

0.146 [0.73]

[0.52]

0.486 [0.38]

[0.34]

NCL

sz

Connective tissue/

vessels

25.72kO06.24 0.813 1.71

25.16kOO4.40 1.63

“Note that the values here are larger than in many other studies although our VBR measure is similar to that noted in a previous

CT study done in our laboratory (McCarley et al., 1989) and these values are within the range of 9-35 cm3 for lateral ventricles reported in the literature (e.g., Grant et al., 1988 (13.7k5.2 ml); Wyper et al., 1979 (37 ml)).

TABLE 3

Comparison between left and right lateral ventricles,Jor schizophrenics and normal control subjects

Anatomic region Group Mean (SD) Paired t test df P

Left lateral ventricle

Right lateral ventricle

Left lateral ventricle

Right lateral ventricle

Schizophrenic 9.03k3.64 - 1.35 9 0.210

8.03k3.17

Controls 13.02k6.85 - 4.40 11 0.001

11.1915.92

tern in a sample of schizophrenics and normal control subjects.

In this initial small sample, no differences be- tween groups were noted on several over-all volu- metric measurements: (1) CSF volume (total, intraventricular, or subarachnoidal), (2) brain vol- ume, (3) brain/CSF ratio (atrophy measure), (4) connective tissue/vessels, (5) VBR, or, (6) third and fourth ventricle volume. This lack of schizo- phrenic-normal differences on the CSF measure is

in contrast to many previous reports (reviewed in Shelton and Weinberger, 1986) although there have also been reports of negative MR findings (e.g., Smith et al., 1984, 1987; Mathew et al., 1985; Rossi et al., 1990). While it is difficult to interpret negative findings, which are not unexpected using small sample sizes, we think it equally important to point out some possible differences in our subject populations which may also contribute to the lack of CSF differences.

111

0 60 160 240

TOTAL TDI SCORE

Fig. 2. Relationship between left minus right lateral ventricle

volume (in cm3) and Total Thought Disorder Index score

within the schizophrenic group (n=S had both measures:

Spearman’s rho = -0.93, p~O.01, two-tailed).

First, although both schizophrenics and normals were within the CSF volume ranges reported in the literature (e.g., Wyper et al., 1979; Grant et al., 1987, 1988) both were in the upper range, suggesting possible sample variation. Second, the schizophrenic patients studied by us had predomi- nantly positive symptoms (7/10, with the remaining three being of mixed symptom type), and, third, all were right-handed. These sample characteristics may represent an important departure from those samples where large ventricular volumes were found in schizophrenics, i.e., the strongest correla- tions in the CT literature has been between negative symptoms and enlarged lateral ventricles (see Shel- ton and Weinberger, 1986). In addition, Andreasen and co-workers (1982, 1985) reported that 3 1% of patients characterized by negative symptoms were left-handed, and in a recent MRI report by this group, schizophrenics with negative symptoms had significantly larger lateral ventricles than those with mixed or positive symptoms (Andreasen et al., 1990). It is also known that right handedness reduces variations in known asymmetries of brain morphology (LeMay, 1984).

Our negative findings, therefore, may reflect a difSerent subgroup of patients, possibly indicating very different pathology, who, though clearly chronic, were right-handed and showed a prepon- derance of positive symptoms. This line of reason- ing is compatible with current concepts of heterogeneity of schizophrenia and multiple sub-

groups (Tsuang and Simpson, 1988). We think it unlikely that our findings are due to differences between our computerized vs. previous manual measurements (see Jack et al., 1988), since our manual area measurements were highly correlated with computerized measurements of the lateral ventricles and neither method showed a statistically significant difference between groups.

A positive finding of interest in the current study was that normal control subjects showed a left > right asymmetry in lateral ventricle size. This finding is consistent with the literature using manual measurements (LeMay, 1984). The schizophrenic patients in our sample did not show a statistically significant left > right asymmetry; in fact 3 out of 10 showed the reverse asymmetry, right >left. The absence of asymmetry in the schizophrenic group is consonant with recent MRI reports. For example, Suddath et al. (1990) re- ported bilateral lateral ventricular enlargement in the ‘ill’ member of identical twins discordant for schizophrenia but found no asymmetry. A lack of asymmetry was also present in the schizophrenic samples of Rossi et al. (1990) and Stratta et al. (1989). Heckers et al. (1990) found no left-right asymmetry in their schizophrenic group as a whole compared to normal controls although a paranoid subgroup did show left >right asymmetry com-

pared with normals. Kelsoe et al. (1988) reported no asymmetry for the schizophrenic group, al- though a posterior > anterior enlargement was present. Crow’s (1990) study of post-mortem brains of schizophrenics found that the left tempo- ral horn was larger than the right in schizophrenic patients.

That the absence of asymmetry in the current study has clinical and functional significance is suggested by the high correlation of thought disor- der with the degree of departure from the normal L > R asymmetry (rho = - 0.93). We caution that the small size of the present sample makes uncer- tain the extent of generalizability of the finding to the entire population and makes important a repli- cation in a larger sample. Finally we note the departure of schizophrenics from the symmetry or asymmetry pattern shown by normals on a variety of other biological variables, including P300 elec- trophysiological measures (McCarley et al., 1991) post-mortem morphometric measures (Brown et al., 1986; Jakob and Beckman, 1989) in vivo MRI

112

studies of temporal lobe grey matter (Suddath et al., 1990) PET studies (Wiesel et al., 1987; DeLisi et al., 1989), attentional processes (Posner et al., 1988) and with general hypotheses that asymmetry abnormalities are a fundamental genetic/develop- mental abnormality in schizophrenia (Crow, 1990). Before giving undue weight to, and speculating on the functional significance of the current study’s finding of an abnormal asymmetry, we think it important to perform essential follow-up studies using the methodological and display techniques described here: (1) to determine whether or not these findings can be replicated in a larger subject group; (2) to look at smaller ventricular subdivi- sions (e.g., temporal horns); and (3) to evaluate whether grey-white brain parenchymal abnormali- ties are associated with the abnormal ventricular asymmetry.

ACKNOWEDGMENTS

The authors gratefully acknowledge the technical and administrative support provided by Brian Chiango, Maureen Ainslie, Eve Chiango, Mary Ellen Korvec, Mary Egan, and Susan Johnson; and the editorial comments and suggestions pro- vided by Steven Faux, Paul Nestor, Brian O’Don- nell, Seth Pollak, and Scott Smith.

REFERENCES

American Psychiatric Association, Committee on Nomencla-

ture and Statistics (1987): Diagnostic and Statistical Manual

of Mental Disorders, 3rd revised edn. American Psychiatric

Association. Washington, DC. Andreasen. N.C. (1985) Positive vs. negative schizophrenia: A

critical evaluation. Schizophr. Bull. 1 I, 380-389.

Andreasen. N.C. and Olsen, S. (1982) Negative vs. positive schizophrenia: Definition and validation. Arch. Gen. Psychi-

atry 39, 7899794.

Andreason, N.C.. Erhardt, J.C., Swayze. V.W., et al. (1990) Magnetic resonance imaging of the bram in schizophrenia:

The pathophysiological significance of structural abnormali-

ties, Arch. Gen Psychiatry 47, 35-44.

Besson, J.A.O., Corrigan, F.M.. Cherryman, G.R. and Smith. F.W. (1987) Nuclear magnetic resonance brain imaging in

chrome schizophrenia. Br. J. Psychiatry 150, 161-163. Bleuler, E. (1911/1950) Dementia Praecox or the Group of

Schizophrenias. [Zinkin. H.. Trans.] International Univperst-

ties Press, New York.

Brown, R., Colter, N.. Corsellis. J.A.N., et al. (1986) Post-

mortem evidence of structural changes in schizophrenia.

Arch. Gen. Psychiatry 43. 36-42.

Cline. H.E.. Dumoulin, C.L., Hart, H.R.. et al. (1987) 3D

reconstruction of the brain from magnetic resonance images

using a connectivity algorithm. Magn. Reson. lmaging 5.

345-352.

Clint. H.E., Lorensen, L.E. and Ludke. S. (1988) Two algo-

rithms for the three-dimensional reconstruction of tomo-

grams. Med. Phys. 15, 45552.

Crow. T.J. (1990) The continuum of psychosis and its genetic

origins: The sixty-fifth Maudsley lecture. Br. J. Psychiatry

156. 7888797.

DeLisi. LE.. Buchsbaum. M.S., Holcomb. H.H., et al. (1989)

Increased temporal lobe glucose use in chronic schizophrenic

patients. Biol. Psychiatry 25, 835-851.

Filipek, P.A.. Kennedy. D.N., Caviness, V.S.. et al. (1988)

Magnetic resonance imaging-based brain morphometry: De-

velopment and application to normal subjects. Ann. Neural.

25. 61-67.

Grant, R., Condon. B., Lawrence. A., et al. (1987) Human

cranial CSF volumes measured by MRI: Sex and age influ- ences. Magn. Reson. Imaging 5, 4655468.

Grant, R.. Condon, B.. Lawrence, A., et al. (1988) Is cramal

CSF volume under hormonal influence? An MR study. J.

Comput. Assist. Tomogr. 12, 36-39.

Heckers, S.. Heinsen. H.. Heinsen. Y.C., et al. (1990) Limbic

structures and lateral ventricle in schizophrenia: A quantita-

tive postmortem study. Arch. Gen. Psychiatry 47. 1016-

1022.

Hollingshead. A.B. (1965) Two factor index of social position.

Yale University Press, New Haven. CT., Published privately.

Jack, C.R., Gehring. D.G., Sharbrough. F.W., et al. (1988)

Temporal lobe volume measurement from MR images: Accu-

racy and left-right asymmetry in normal persons. J. Comput.

Assist. Tomogr. 12, 21-29.

Jacob, H. and Beckmann. H. (1989) Gross and histological

criteria for developmental disorders in brains of schizophren-

ics. J. Roy. Sot. Med. 82. 4666469.

Johnstone, E.C.. Crow, T.J., Frith, C.D.. et al. (1976) Cerebral

ventricular size and cognitive impairment in chronic schizo-

phrenics. Lancet 2, 924-926.

Kelsoe, J.R., Cadet, J.L., Pickar. D. and Weinberger, D.R.

(198X) Quantitative neuroanatomy in schizophrenia: A con-

trolled magnetic resonance imaging study. Arch. Gen. Psychi-

atry 45. 533-541.

Kikinis. R., Jolesz, F.A.. Gerig, G., et al. (1990) NATO AS1 Series Vol. F60 3D Imaging in Medicine. In: K.H. Hohne

et al. (Eds.). Springer-Verlag, Berlin. pp. 441-454. Kraepelin, E. (1919) Dementia Praecox and Paraphrenia. [Bar-

clay, R.M., Trans.] Chicago Medical Book, Chicago, IL.

Last, R.J. and Tompsett. D.H. (1953) Casts of cerebral ventri-

cles. Br. J. Surgery 40. 5255543. LeMay, M. (1984) Radiological. developmental, and fossil

asymmetries. In: N. Geschwind and A.M. Galaburda (Eds.).

Cerebral Dominance: The Biological Foundations. Harvard

University Press, Cambridge, MA.

Mathew, R.J. and Partain, C.L. (1985) Midsagittal sections of

113

the cerebellar vermis and fourth ventricle obtained with

magnetic resonance imaging of schizophrenic patients. Am,

J. Psychiatry 142, 970-971.

McCarley, R.W., Faux, SF., Shenton. M.E., et al. (1989) CT

abnormalities in schizophrenia: A preliminary study of their

correlation with P3OO;P200 electrophysiological features and

positive/negative symptoms. Arch. Gen. Psychiatry 46.

698708.

McCarley. R.W.. Faux, S.F., Shenton. M.E.. et al. (1991)

Event-related potentials in schizophrenia: Their biological

and clinical correlates and a new model of schizophrenic

pathophysiology. Schizophr. Res. 4. 2077229.

Nasrallah. H.A.. Jacoby, C.G., McCalley-Whitters, M.. et al.

(1982) Cerebral ventricular enlargement in subtypes of

chronic schizophremcs. Arch. Gen. Psychiatry 39, 7744777.

Pavlicek. W.. Medic. M. and Weinstein, M. (1984) Pulse

sequence and significance. Radiographics 4, 49-65.

Posner. M .I., Early, T.S., Reiman, E.. et al. (1988) Asymmetries

in hemispheric control of attention in schizophrenia. Arch.

Gen. Psychiatry 45. 814~821.

Rossi. A., Stratta, P.. D’Albenzio. L.. Tartaro, A.. et al. (1990)

Reduced temporal lobe areas in schizophrenia: Preliminary

evidences from a controlled multiplanar magnetic resonance

imaging study. Biol. Psychiatry 27. 61-68.

Sandor, T.. Albert, M.. Stafford. J., et al. (1988) Use of

computerized CT analysis to discriminate between Alzheimer

patients and normal control subjects. Am. J. Neuroradiol.

9. Il81~1187.

Sandor. T.. Jolesz, F., Tieman. J., et al. (1990) Extraction of

morphometric information from dual echo magnetic reso-

nance brain images. SPIE Proc. 1360, 665-675.

Scriptel Corporation (1985) Scriptel User Guide SPD Series

Graphics Tablets. Scriptel Corporation. Columbus. OH.

Shelton, R.C. and Weinberger. D.R. (1986) X-ray computerized

tomography studies in schizophrenia: A review and synthesis.

In: H.A. Nasrallah and D.R. Weinberger (Eds.), Handbook

of Schizophrenia, Vol. 1: The Neurology of Schizophrenia.

Elsevier, Amsterdam, pp. 207-250.

Smith. R.C.. Calderon. M., Ravichandran, G.K., et al. (1984)

Nuclear magnetic resonance m schizophrenia: A preliminary

study. Psychiatry Res. 12, 137-147.

Smith. R.C., Baumgartner. R. and Calderon, M. (1987) Mag-

netic resonance imaging studies of the brains of schizophrenic

patients. Psychiatry Res. 20. 33-46.

Solovay, M.R.. Shenton, M.E., Gasperetti, C., et al. (1986)

Scoring manual for the thought disorder index (revised

version). Schizophr. Bull. 12. 493-496.

Spitzer. R.L. and Endicott. J. (1978) Schedule for Affective

Disorders and Schizophrenia. Biometric Research, New York

State Psychiatric Institute, New York.

Stratta, P., Rossi. A., Galluci. M.. et al. (1989) Hemispheric

asymmetries and schizophrenia: A preliminary magnetic reso-

nance imaging study. Biol. Psychiatry 25. 275-284.

Suddath. R.L., Christison, G.W.. Torrey. E.F.. et al. (1990)

Anatomical abnormalities in the brains of monozygotic twins

discordant for schizophrenia. New Engl. J. Med. 322.

7899794.

Tsuang, M.T. and Simpson, J.C. (Eds.) (1988) Handbook of

Schizophrenia. Vol. 3, Nosology, Epidemiology. and Genet-

ics. Elsevier. Amsterdam.

Udupa. J.K., Srihari, S.N. and Herman. G.T. (1982) Boundary

detection in multidimensions. IEEE Trans. Pattern Anal.

Mach. Intell. 4. 41-50.

Weinberger, D.R.. Torrey. E.F., Andreasen. N.. et al. (1979)

Lateral cerebral ventricular enlargement in chronic schizo-

phrenics. Arch. Gen. Psychiatry 36. 7355739.

Wiesel. F.A.. Wik. G., Blomqvist, I.S.F., et al. (1987) Regional

brain glucose metabolism in drug free schizophrenic patients

and clinical correlates. Acta Psychiatr. Stand. 76. 6288641.

Wyper, D.J.. Pickard, J.D. and Matheson. M. (1979) Accuracy

of ventricular volume estimation. J. Neural. Neurosurg.

Psychiatry 42. 3455350.