| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, Beth Israel Deaconess Medical Center, 330 Brookline
Ave., Boston, MA 02215.
2 Department of Radiology, Harvard Medical School, Boston, MA.
3 Clinical Research Program, Children's Hospital Boston, Boston, MA 02115.
4 Department of Ambulatory Care & Prevention, Harvard Medical School and
Harvard Pilgrim Health Care, Boston, MA.
Received December 8, 2006;
accepted after revision March 26, 2007.
Supported by National Institutes of Health, National Institute of
Biomedical Imaging and Bioengineering grant 01998.
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. In this prospective study, 50 consecutive fetuses at 17-37 weeks of gestational age referred for MRI for VM underwent fast spin-echo T2-weighted imaging. Supratentorial brain parenchyma, lateral ventricles, and extraaxial and cerebellar volumetric measurements were manually obtained in three planes by three radiologists. Inter- and intraobserver variability were assessed. The relationship between volumes and gestational age, and lateral ventricular diameter were assessed.
RESULTS. Volumes increased with gestational age. The presence of VM correlated with increased lateral ventricle diameter. The effect of imaging plane was negligible. Inter- and intraobserver variability were low.
CONCLUSION. Supratentorial parenchyma and lateral ventricular volumes can be reliably measured on fetal MRI, and imaging plane was not an important factor in measurement. Further studies are needed to correlate these indexes with long-term postnatal outcomes.
Keywords: brain • central nervous system • fetuses • MRI • ventriculomegaly
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Follow-up of cognitive and motor development in cases of sonographically diagnosed fetal VM has been reported in a limited fashion [4, 7, 19-21]. However, it is difficult to generalize these preliminary results due to small sample sizes, inclusion of various anomalies associated with mild VM, differences in evaluation techniques, and short terms of follow-up.
In studies limited to follow-up of sonographically isolated mild VM, developmental delays were reported in 0-36% of cases [1, 6, 22, 23]. This wide range of outcomes can be explained at least in part by differing methods of evaluation of development. The degree of VM is correlated to outcome, with fetuses with ventricular a width less than 12 mm having a better outcome [21, 24].
As part of an ongoing study to assess outcomes in fetuses with VM, we hypothesized that the use of qualitative and quantitative MRI indexes will improve the diagnostic usefulness of MRI for predicting outcomes compared with a reliance on sonographic data alone. Specifically, in fetuses with VM, higher ventricular volume and lower cortical volume (with respect to gestational age) will each correlate negatively with normal cognitive and motor development.
However, to test this hypothesis, we must first validate the method of performing fetal brain volumetry. This study was performed to determine the imaging plane that is most reproducible for the performance of brain volumetry measurements and to determine inter- and intraobserver variability in assessing fetal brain volumes in fetuses referred for VM.
|
|
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Study Population
Fifty fetuses from 50 consecutive pregnant women referred for MRI because
of sonographically diagnosed VM were studied between July 1, 2003 and October
18, 2004, as part of a study funded by the National Institutes of Health (NIH)
at Beth Israel Deaconess Medical Center. Historical data include age by dates
in weeks and referral diagnosis. Age by dates was defined as either the last
menstrual period or the estimated due date if the date had been recalculated
after a first-trimester sonogram.
Sonography
All fetuses underwent sonography at our institution on the day of the MRI.
Sonography was performed with an ATL 5000 (ATL) unit with 2.5-5.0-MHz
transducers by a radiologist who had 12 years of experience in high-risk
obstetric sonography. When the fetus was in the cephalic position, a
transvaginal scan was also obtained to better assess intracranial anatomy.
Sonographic data obtained included biometric measurements in weeks: biparietal diameter (BPD), head circumference, and average sonographic age (calculated from the average of BPD, head circumference, abdominal circumference, and femur length using the methods of Hadlock et al. [25]; and the lateral ventricular diameter (measured on a transverse image at the atrium). When VM was asymmetric, the larger ventricular diameter was used for data analysis. Sonographic diagnosis of VM (defined as a lateral ventricle measurement > 10 mm) and other CNS anomalies were recorded.
MRI Protocol
MRI examinations were performed on a 1.5-T superconducting system (Signa,
GE Healthcare) using an 8-element phased-array surface coil with a radiologist
monitoring the study as it was being performed. A three-plane scout view was
obtained and used to orient the first diagnostic imaging sequence. Single-shot
fast spin-echo imaging was performed in the fetal sagittal, coronal, and axial
planes using the following parameters: TR/TE, single-shot/60; field of view,
from 30 x 30 to 34 x 34 cm; matrix, 256 x 256 or 512 x
512; slice thickness, 4-5 mm (35 with 4-mm slice thickness, 14 with 5-mm slice
thickness, and one with both 4- and 5-mm slice thicknesses); and sequence
acquisition time, 29-45 seconds. The number of slices in each sequence
depended on fetal size and orientation of imaging plane to cover the region of
interest (ROI). Each sequence was used as the scout for subsequent
sequences.
Sequences were repeated as needed to show pertinent anatomy, but because these studies were performed to assess CNS anatomy, and not necessarily to perform volumetry, sequences were not repeated if only a few slices were affected by motion. T1-weighted sequences were obtained for diagnostic purposes but were not used for the volumetry portion of this study. Scanning time for single-shot fast spin-echo imaging (including scout) ranged from 5 to 20 minutes (mean, 10 ± 3 minutes).
MRI Diagnosis
MR images were reviewed by a radiologist with 11 years of experience in
fetal MRI, and measurements of the size of the lateral ventricle (in mm) at
the level of the atrium and at the frontal horns obtained using electronic
calipers were recorded. The MRI diagnosis of VM (lateral ventricle measurement
> 10 mm) and other CNS anomalies was recorded.
Comparison of Measurements
The sonographic measurement of the ventricle was used to determine whether
a fetus was normal (< 10 mm, no other CNS anomalies seen on sonography or
MRI) or had VM, in which case the degree of VM (10.0-12.0, 12.1-15, or > 15
mm) was used to stratify fetuses. A comparison was made of ventricular
measurements on sonography and MRI for these subgroups.
Volume Calculation
MR images were transferred to an Advantage Windows workstation (ADW 3.0, GE
Healthcare) and volumes were assessed using Advantage Windows Volume Analysis
software (Voxtool 3.1, GE Healthcare).
Radiologist A, who had 3 years' experience in fetal MRI, selected the best sagittal, coronal, and axial sequences for volumetric analysis. If it was determined subjectively that a sequence showed motion that would interfere with accurate tracing, or if the entire brain was not included in a sequence, then that sequence was not used for data analysis. If a particular imaging plane was obtained more than once during an examination, then the sequence with less motion was chosen for review. A total of 135 sequences in 50 fetuses were evaluated (46 axial, 44 coronal, 45 sagittal). Forty fetuses underwent volumetry with three slice orientations, six with two slice orientations, and three with one slice orientation. No fetuses were excluded from review because of motion on all sequences.
Segmentation was performed using hand-tracing of a free-form ROI on individual consecutive slices in the axial, sagittal, and coronal (Figs. 1A, 1B, and 1C) planes around the outer aspect of supratentorial cortex, lateral ventricles, and cerebellum. In addition, the extraaxial CSF was traced in the axial plane (Fig. 2). The area of each ROI obtained per image was automatically calculated (on the basis of the cross-sectional area and slice thickness) and summed to determine the total ROI volume. Lateral ventricular volume included the choroid plexus. Supratentorial parenchymal volumes were obtained by subtracting the volume of the lateral ventricles from the volume of the supratentorial cortex tracing. In two images from the same 17-week-gestational-age fetus, the two measures were small and close together (16-17 cm3) and the difference was negative; these points were not used in analysis.
|
Radiologist A performed volumetry on all 50 fetuses at two separate intervals, 6 months apart, to minimize recall bias. The second set of volumetric analyses was performed without knowledge of the volumes obtained during the first analysis. Two additional radiologists, radiologists B and C, each with 1 year of experience in fetal MRI, performed volumetric analysis a single time on the same sequences as radiologist A had.
The amount of time to manually trace the ROIs was calculated for radiologist C. Times were obtained for drawings of the supratentorial brain parenchyma, ventricles, and posterior fossa in each imaging plane. Time for assessment of extraaxial CSF in the axial plane was also measured. Times were compared with respect to imaging plane and gestational age.
Statistical Analysis
Each volume measurement (supratentorial parenchyma, ventricles, cerebellum,
and extraaxial CSF) was analyzed separately, as was the ratio of ventricular
to supratentorial parenchymal volume. We used a comprehensive mixed-effects
regression model to assess simultaneously the influences of gestational age,
plane of observation, random variability among subjects, and random
variability between and within reviewers. Fetal gestational age was
represented in four alternate ways: by dates, by sonographic biometry, by BPD,
and by head circumference. We constructed separate regression models using
each of these four indexes to adjust for fetal age, all other variables being
the same. The regression model also included terms for interaction of plane
with reviewer, testing whether reviewer agreement was better in one plane of
observation than another; and interaction of plane with fetal size, testing
whether the relationship of volume with respect to gestational age or brain
size was the same as observed from all directions.
To compare volume among subgroups according to prenatal diagnosis, we added diagnostic category to the regression model. Measurement times were analyzed similarly by mixed-effects regression, assessing the influences of gestational age and plane and adjusting for subject variability.
The volume measurements, volume ratios, and times were log-transformed for analysis to compensate for the pronounced skew in those quantities. For reporting, we retransformed group means to the original units. Regression contrasts and variance components are expressed as percentages; for example, the patient-to-patient SD of log10 parenchymal volume was 0.116, which we report as a relative SD of 100% x (100.116 - 1) = 31%. Gestational age was entered linearly in the model (using each of the four different methods of gestational age estimate), and accordingly the relationship of volume to gestational age is expressed as percentage per week; for example, the regression coefficient for log10 parenchymal volume was 0.0636 per week, which we report as 100% x (100.0636 - 1) = 15.8% per week. Head size was log-transformed for entry into the regression model, making the rate increase of volume with respect to head size the exponent of a power-law relationship; for example, the regression coefficient for log10 parenchymal volume was 2.7 per log10 BPD, which we report as volume being proportional to the 2.7 power of BPD.
As a general index of reliability for each volume measurement, including all sources of random variance, we calculated the intraclass correlation coefficient (ICC) as (subject variance) ÷ (subject + interobserver + intraobserver + residual variance). To assess the correlation of ventricular volume with MRI measurements of ventricular size and parenchymal thickness, we used the Spearman's correlation coefficient, adjusted for gestational age, because of its resistance to bias from skewed data.
We specified p < 0.05 as the threshold for statistical significance except when assessing nominal variables (e.g., three planes of observation, five diagnostic categories), in which case we specified 0.05 for the overall comparison and applied the Bonferroni rule (0.05 ÷ number of preplanned comparisons) for pairwise comparisons between levels. We interpreted correlations greater than 0.6 as strong, less than 0.4 as mild. SAS software, version 9.1, was used for all analyses. Mixed-effects regression was performed with the SAS MIXED procedure, with 0 as a lower boundary on random variance estimates. In cases in which the optimal estimate was on the boundary, we reported variance as 0 or "negligible."
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
Volumes with Respect to Gestational Age
Gestational age by dates ranged from 17 weeks to 36 weeks 5 days. The
dependence of fetal brain volume measurements on gestational age is detailed
in Table 2 and illustrated in
Figures 3,
4,
5A, and
5B. Supratentorial parenchymal
volume ranged from 1.6 to 306 cm3 and was 16% higher for every
additional week of gestational age, as measured by either dates or sonographic
variables (p < 0.0001, Fig.
3). Ventricular volumes ranged from 1.6 to 76 mL and increased by
4% per week, a statistically significant rate (p = 0.002), although
more slowly than the parenchyma volume increased
(Fig. 4). Cerebellar volume
(Figs. 5A and
5B) and extraaxial CSF
increased with respect to gestational age at rates comparable to the
parenchyma, 14% and 17% per week, respectively (p < 0.0001).
|
|
|
|
|
Parenchymal volume increased in proportion to the 2.7 power of BPD (p < 0.0001), whereas ventricular volume increased only as the 0.6 power (p = 0.019). Cerebellar volume (Figs. 5A and 5B) and extraaxial CSF increased similarly to parenchyma, in proportion to the 2.4 and 2.7 power, respectively (p < 0.0001). Similar results were obtained for rates of increase with respect to head circumference, with slightly higher exponents (Table 2).
Two- Versus Three-Dimensional Measurements
A moderately strong and significant positive correlation was seen between
ventricular volumes and ventricular diameter at the left and right atria
(Spearman's r = 0.66 and 0.63, respectively, adjusted for sonographic
gestational age), with lesser (r = 0.40-0.41) but still significant
correlations at the left and right frontal horns (p < 0.0001 for
all).
Imaging Plane
The mean measured volumes did not differ systematically by plane of
observation, nor did the rates of increase with gestational age by sonography,
as detailed in Table 3. Similar
results were seen for rate of increase of volume with gestational age by
dates, head circumference, and BPD. Interobserver variability did not differ
among the three planes, as indicated by small or negligible estimates for
reviewer x plane interaction.
|
Inter- and Intraobserver Variability
The SD among the three reviewers was 12% for measurements of parenchyma, 6%
for ventricles, 4% for extraaxial CSF, and negligible variation for cerebellum
(Table 4). Negligible
variability was seen for the two measurements on different occasions by a
single reviewer. The overall reliability of measurement was best for
ventricular volume (ICC = 0.87) and poorest for cerebellar volume (ICC =
0.34).
|
Volumes
Table 5 shows the
differences in MRI measurement of parenchymal volume, ventricular volume, and
ventricular percentage according to prenatal diagnostic categories by
sonography, with degrees of VM subdivided as in
Table 1. Mean parenchymal
volume, adjusted for gestational age, was somewhat higher in normal fetuses
but not significantly so. Ventricular volume and percentage differed
significantly among the categories, increasing with degree of VM, in
corroboration of the sonographic finding.
|
Time for Measurement
To trace the outside of the supratentorial parenchyma, radiologist C
required a median of 4.5 minutes (range, 1.5-13.0 minutes). The measurement
time varied significantly by plane of observation (p = 0.0003), being
greatest in the coronal plane and least in the sagittal plane, and increased
by 5.4% per week of gestational age (p < 0.0001)
(Table 6). Shorter times but
similar patterns were recorded for tracing the ventricles. The median time for
tracing the cerebellum was 1 minute in all planes, and for tracing extraaxial
CSF (performed in the axial plane only), more than 5 minutes. All regions
required more time for more mature fetuses, the rate of increase ranging
between 1.3% and 5.9% per week of gestational age.
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Many articles have been published on the use of volumetry for assessment of fetal lungs using sonography [17, 26-28] and MRI [17, 29-33]. However, less attention has been given to brain volumetry. In a study of postmortem fetuses by Kinoshita et al. [34], the maximum volume of the ventricles was 2.6 mL at 23 weeks. We found a similar volume early in gestation, ranging from 1.9 mL in our in vivo population. Our maximum ventricular volume was 74 mL, much greater than that reported by Kinoshita et al., which is expected given our VM population. In a preliminary report of 3D modeling of the fetal brain, Schierlitz et al. [35] assessed eight fetuses for brain morphology and ventricular volume measurements. However, ventricular volumes were not given for comparison. In a case series by Andreas et al. [36], volumetry was performed in three fetuses with brain malformations and compared with brain weight standards from pathology, illustrating how this information may be useful in counseling patients.
An important finding in our study is that the effect of the imaging plane is negligible. This means that if volumetry is to be performed, time can be saved by performing it in a single imaging plane, using the plane with the least motion. The time needed to perform volumetry of the supratentorial parenchyma in a single imaging plane ranged from 3 to 19 minutes, which is well within the range of an acceptable time for performance of volumetry in clinical practice. As software enhancements of semiautomatic tools for aid in tracings are improved, the time to perform volumetry should also decrease. However, power could be an issue in this study, as could model overparameterization. Although we did not find planar differences, it is possible that with more evaluators and more subjects, a particular difference could be found to be significant.
The potential for accurate supratentorial volumes obtained in any imaging plane also has important consequences for scanning time in the magnet. If diagnostic information is obtained on sequences obscured by motion, as long as one sequence in a fetal examination has relatively little motion, then volumetry can be performed. In our study, the images were monitored by a radiologist interested in ensuring that diagnostic information was obtained, but without regard for the performance of volumetry. In this small series, at least one sequence was suitable for volumetry in every case. It is possible that, in the future, if volumetry becomes a standard method of evaluation, more attention will need to be paid to motion artifacts; or that lower resolution, faster sequences will be needed.
Our study has shown that volumetry of the parenchyma and ventricles in fetuses referred for VM can be performed in a reliable manner because inter- and intraobserver variability are low. This is important if volumetry of the brain is to become a standard method for assessment of fetuses with VM. The reliability of measurement of the cerebellum was the poorest in this study. This was believed to be due to the small size of the cerebellum. Small differences in tracing technique multiplied by the relatively large slice thickness (with respect to cerebellar size) likely led to this finding.
A limitation of our study is the slice thickness of 4-5 mm. When examining early-gestational-age fetuses, partial volume averaging is a problem, and it is difficult to obtain appropriate volumes with large slice thickness. However, this is the standard method of performing fetal MRI. This slice thickness is less a problem with later gestational age fetuses with larger brains. Outcome studies will be needed to ascertain whether volume data obtained in the currently described manner are more predictive of outcome than 2D ventricular measurements and whether this predictive ability is improved at later gestational ages. In the future, with faster 2D and 3D imaging techniques, we should be able to obtain thinner slices of the fetal brain. It is possible that we can then use registration-based reconstruction methods to reconstruct higher-resolution MR images for fetal volumetry [37].
A third limitation of our study was that we did not compare our results with outcome data. This was intentional because we were following a large cohort of fetuses with VM with standardized neurologic and psychologic testing. This article is meant to elucidate the methodology for performing fetal volumetry. We have shown the low inter- and intraobserver variability, and that only a single imaging plane must be used. This suggests that we have a reliable method to obtain quantitative data from MRI that might be useful in formal outcomes analysis in large cohorts in the future.
Use of a standardized comprehensive method of evaluation is needed to correlate degree of VM, age at diagnosis, and associated findings with postnatal development. Parenchymal and ventricular volumetry are tools that may aid in this process. Further studies are needed to correlate longer-term outcomes with these findings.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
