Prefrontal Cortex, Thalamus, and Cerebellar Volumes in ...

BRAIN VOLUMES IN YOUTHS 1593
cooperation, and image quality. We required that the midsagittal slice
show full visualization of the cerebral aqueduct and the anterior and
posterior commissures, in which a line was estimated requiring the
anterior-commissure-posterior-commissure line to be within three degrees
of 180. If these criteria were not met, the subject was realigned until these
criteria were met. A three-dimensional spoiled gradient recalled acquisition
in the steady-state pulse sequence was used to obtain 124 contiguous
images with slice thickness of 1.5 mm in the coronal plane for region of
interest measures (TE � 5 msec, TR � 25 msec, flip angle � 40 degrees,
acquisition matrix � 256 � 192, number of excitations � 1, field of vision
� 24 cm). Coronal sections were obtained perpendicular to the anteriorcommissure-posterior-commissure
line to provide a more reproducible
guide for image orientation. Axial proton density and T 2-weighted images
were obtained to enable exclusion of structural abnormalities on MRI. A
neuroradiologist reviewed all scans and ruled out clinically significant
abnormalities. All subjects tolerated the procedure well. No sedation was
used.
The imaging data from the coronal sections were transferred from the
MRI unit to a computer workstation (PowerMacintosh, Apple Computer)
and analyzed with the use of IMAGE software (version 1.61) developed at
the NIH (Rasband, 1996) that provides valid and reliable volume measurements
of specific structures using a manually operated (hand tracing)
approach. Trained and reliable raters who were blinded to subject information
made all measurements. These methods were previously described
by our group (De Bellis et al., 1999; De Bellis et al., 2002; De Bellis et al.,
2001) and are briefly presented here.
Intracranial volumes were calculated by first manually tracing the
intracranial volume of each coronal slice after exclusion of skull and dura,
then summing these areas of successive coronal slices, including GM and
WM and cerebral spinal fluid (CSF) volumes, and multiplying by slice
thickness. These measures included frontal, parietal, temporal, occipital
cortex, subcortical structures, cerebellum, and brainstem.
Cerebral volumes were measured after manual exclusion of CSF volumes,
cerebellum, and brainstem in the same manner and included cortical
and subcortical structures.
Prefrontal cortex volumes were calculated by summing up areas of
successive coronal slices, including gray and white matter and CSF volumes
and multiplying by slice thickness. The anterior boundary of the
prefrontal cortex was defined as the most anterior coronal section containing
gray matter. The coronal slice showing the genu of the corpus
callosum was used to mark the posterior limit of the prefrontal cortex
(Rosenberg et al., 1997). Prefrontal white and gray matter volumes were
calculated by using a semiautomated segmentation algorithm. This computerized
segmentation technique uses mathematically derived cutoffs for
gray matter–white matter–CSF partitions with histograms of signal intensities.
This computerized segmentation technique is both labor intensive
and manually operated. It uses an interactive method in which mathematically
derived cutoffs for gray matter–white matter–CSF partitions from
histograms of signal intensities are used to individually select gray matter,
white matter, and CSF areas from each coronal slice. Gray matter, white
matter, and CSF areas are thus separately calculated and multiplied by
slice thickness for the individual subject’s gray matter, white matter, and
CSF volumes. In this way, we can minimize the inherent limitations on
qualifying white matter signal hypointensities as gray matter on T 1weighted
MRI scans by visual inspection of slices (i.e., so that hypointensity
artifacts in corpus callosum or cerebral white matter are not calculated
as gray matter). This approach has been validated by using both a stereology
technique for brain morphometric measurements and a phantom
with known absolute volumes (Keshavan et al., 1995) and has been used in
several published neuroimaging studies (De Bellis et al., 1999; De Bellis et
al., 2001; Rosenberg et al., 1997). See Fig. 1.
The boundaries of the thalamus were previously described (Gilbert et
al., 2000). The boundaries of the thalamus were defined as follows: The
mammillary bodies and the interventricular foramen defined the anterior
boundary; the internal capsule defined the lateral boundary; the third
ventricle defined the medial boundary; the hypothalamus defined the
inferior boundary; the lateral ventricle defined the superior boundary; and
Fig. 1. Manual tracings of PFC. PFC volumes were calculated by summing up
areas of successive coronal slices and multiplying by slice thickness. The anterior
boundary of the prefrontal cortex was defined as the most anterior coronal
section containing gray matter. The coronal slice showing the genu of the corpus
callosum was used to mark the posterior limit of the prefrontal cortex. Prefrontal
white and gray matter volumes were calculated by using a semiautomated
segmentation algorithm (not shown).
the crus fornix defined the posterior boundary of the thalamus. Every
coronal slice that included the thalamus was manually traced, and total
volume was computed by summing up successive areas and multiplying by
slice thickness.
The volumes of the cerebellum, vermis, and brainstem were calculated
by summing up areas of successive coronal slices after tracing the region
of interest and excluding CSF as previously described (Hardan et al.,
2001). Briefly, measurements began as the cerebellum appeared laterally
to the pons. The tentorium cerebelli acted as the superior limit and the
base of the cerebellum itself as the inferior limit. The cisterna magna and
transverse sinus were excluded. The last slice included was the one at
which the cerebellum was no longer distinguishable from the transverse
sinus or disappeared. The measurement of the vermis began at the slice
where the anterior and/or inferior posterior lobes appeared. The gray
matter of the vermis structures, determined by mathematically derived
cutoffs for gray matter–white matter–CSF partitions from histograms of
signal intensities as previously described (De Bellis et al., 2001), were
traced separately until the slice where the fourth ventricle was no longer
visible. The cerebellar medullary body and the deep cerebellar nuclei were
excluded. Measurements were made around the vermis until it was no
longer visible. The first slice of the brainstem was measured where the
pons first appeared within the suprasellar cistern. The cerebellar peduncles,
including the brachium pontis (middle), were included in the
brainstem. In the anterior plane, the superior limit of the pons was a
straight line connecting the ambient cisterns from left to right. Posterior
measurements included the cerebral aqueduct and the superior colliculus.
No separate volumetric measurements were made for the midbrain, pons,
and medulla oblongata. See Fig. 2.
Intraclass correlation of interrater and intrarater reliability for independent
designation of regions on segmented images obtained from 20
subjects were 0.99 and 0.99 for intracranial volume, cerebral volume,
cerebral CSF, prefrontal lobe volume, prefrontal lobe gray matter, prefrontal
lobe white matter, prefrontal CSF; 0.96 and 0.98 respectively, for
right, left, and total thalamus volumes; 0.91 and 0.95 respectively, for right,
left, and total cerebellum; 0.89 and 0.94 respectively, for pons/brainstem;
and 0.91 and 0.92 respectively, for cerebellar vermis.
Statistical Methods
Demographic variables were compared by means of Student’s t test and
Pearson 2 as appropriate. Histograms were obtained to assess normality of
the data and to observe any outlying observations. There were no significant
outliers, and we did not exclude any cases in our data analyses.
Formal hypothesis testing was carried out by t tests in two stages, first with
the raw data, then again adjusting for total cerebral volume, to determine
differences between AUD subjects and control subjects. In testing for
covariate effects such as sex and interactions (sex-by-group), multivariate
regression analysis was used. Brain structural means, which differed significantly
between the groups (e.g., PFC volumes), were adjusted for
cerebral volume to correct for individual differences in brain size with
regard to gender and then correlated with clinical data using Spearman
correlation coefficients. All significance testing involving the main hypoth-