Dr P. Gowland, Sir Peter Mansfield Magnetic Resonance Centre, School of Physics and Astronomy, University of Nottingham, Nottingham, England, UK.
Objectives To compare fetal organ size measured using echo-planar magnetic resonance imaging and 2D ultrasound. To determine the relative accuracy with which each technique can predict fetal growth restriction.
Design A cross sectional, observational study comparing two different measurement techniques against a gold standard, in a normal clinical population and an abnormal population.
Setting and Population Seventy-four pregnant women (33 who were ultimately found to be normal and 37 with fetal growth restricted fetuses) were recruited from the City Hospital Nottingham UK to be scanned once (at various gestations).
Methods Each fetus had a standard ultrasound biometry assessment followed by magnetic resonance imaging measurement of organ volumes.
Main outcome measures For each measurement for both techniques, the normal population was plotted with 90% confidence intervals. Fetal growth restricted subjects were compared with the normal population using this plot; 2 × 2 tables were created for each measurement. This was used to calculate the relative sensitivities and positive predictive value of the different measurements. A Bland–Altman plot was used to compare the ultrasound and magnetic resonance imaging measurements of fetal weight.
Results Brain sparing was seen in ultrasonic head circumference measurements, but an overall reduction in fetal growth restriction brain volume was apparent using magnetic resonance imaging at late gestations. Across the whole range of gestational ages, ultrasound assessment of fetal weight was the best predictor of fetal growth restriction.
Conclusion Ultrasound fetal weight assessment appears to identify more fetuses with fetal growth restriction than abdominal circumference. The brain sparing apparent in ultrasonic head circumference measurements of fetuses with fetal growth restriction masks a reduction in brain volume observed with magnetic resonance imaging.
Poor fetal growth is associated with severe disability,1 significant but subtle impairments in motor and cognitive function2–9 and poor general health in later life,1,10 possibly related to programming of the autonomic nervous system.11 Fetal growth restriction is a significant risk factor for preterm delivery and is also associated with long term health risks, but only some of this morbidity results from the effects of preterm delivery and prematurity. Despite the clear link between fetal growth restriction and impaired neurodevelopment, there is considerable evidence of ‘brain sparing’ (asymmetric growth) from ultrasound in most cases of fetal growth restriction.12 Brain sparing is probably accomplished by the redistribution of the fetal blood flow from the trunk to the head.13 Preliminary magnetic resonance imaging data have also suggested that brain size is less affected by fetal growth restriction than liver size.14,15
Antenatal screening for fetal growth restriction primarily depends upon serial measurements of uterine size. Various charts have been used for screening with varying degrees of success. Overall, clinical methods predict fetal growth restriction with 60–85% sensitivity, 80–90% specificity and a positive predictive value of 20–80%.16 It has previously been shown that fetal growth restriction can be predicted (and sometimes distinguished from small for gestational age fetuses) using ultrasound biometry, Doppler blood flow assessment,17 fetal heart rate monitoring, magnetic resonance imaging volume measurements14,15,18 and magnetic resonance imaging placental structural and functional measurements.19–21 However, ultrasound does not appear to offer much benefit over clinical examination22 and it has not been possible to predict perinatal complications or subsequent neurological handicap.17 Similarly, no link has been found between placental pathology and neurological outcome in fetal growth restriction.23 It is unlikely that magnetic resonance imaging would ever replace ultrasound as the first line screening technique. However, it may have a part to play in studying fetuses with suspected fetal growth restriction on the basis of ultrasound in more detail.
The aim of this study is to compare ultrasound and magnetic resonance imaging measurements of fetal organ size in normal and fetal growth restricted pregnancies, to determine whether there are any differences in fetal organ size detected by the two techniques. We also aimed to determine whether magnetic resonance imaging had the potential to offer any advantage over ultrasound biometry in the prenatal diagnosis of fetal growth restriction.
The local Hospital Ethics Committee approved the study and informed written consent was obtained prior to imaging all volunteers. Two groups of subjects were recruited from the City Hospital Nottingham from 1995 to 1997.
Thirty-seven women were recruited by approaching women with clinical suspicion of a small baby or an ultrasound scan showing an abdominal circumference below the 10th centile. All of these fetuses were subsequently diagnosed as being growth restricted at birth, defined on the basis of an individualised birthweight ratio24 below the 10th centile. The individualised birthweight ratio is an index of fetal growth, calculated from gestational age at delivery, parity, fetal sex, maternal height, ethnic origin and birthweight that allows a better separation of truly growth restricted infants from those that are physiologically small. Doppler ultrasound was not routinely performed in the whole group, except when indicated for clinical reasons. Only one fetus was found to have an abnormal umbilical artery Doppler velocity profile and this fetus subsequently died before six weeks of postnatal age.
A further 37 women were recruited by advertisement at the antenatal clinic. Four of the fetuses from this group subsequently had individualised birthweight ratio <10th centile, and these cases were therefore included in the abnormal group. There were no abnormalities at six weeks of postnatal life in the normal group. There were no recorded chromosomal or syndromic abnormalities in either group. Maternal characteristics were similar in each group in relation to age and smoking status, although there were significantly more parous subjects in the normal group. The overall pregnancy characteristics are summarised in Table 1.
Table 1. Pregnancy characteristics. Values are given as median (IQR).
Fetal growth restricted (n= 41)
Normal (n= 33)
Maternal age (years)
Individualised birthweight ratio (centile)
Gestation at delivery (weeks)
Parity (% primips)
The fetuses from both groups were scanned antenatally with ultrasound immediately preceding magnetic resonance imaging. The normal and fetal growth restricted groups were matched so that women from each group were scanned at similar gestational ages between 26 and 40 weeks. Gestational age at scanning was estimated from the date of the last menstrual period, but corrected on the basis of first trimester ultrasound biometry if there was a discrepancy of more than seven days.
Ultrasound scanning was performed in a conventional manner using a GEC RTX-400 scanner. Measurements of head circumference, bipariteal diameter, femur length and abdominal circumference were obtained as the mean of three consecutive measurements. Fetal weight was estimated by the use of the abdominal circumference and femur length measurements in a standard formula.25 The same observer (KRD) made all the measurements. In two subjects at late gestation, the head circumference could not be measured by ultrasound.
Magnetic resonance imaging was performed using a 0.5-T purpose built magnetic resonance scanner using the multislice, echo-planar sequence. Echo-planar imaging is a particularly rapid form of magnetic resonance scanning, allowing an image to be acquired in about 130 ms, effectively freezing fetal motion.26,27 Echo-planar images of 13 contiguous slices were collected in 2.5 seconds with a slice thickness of 7 mm. The in-plane resolution was 3.5 × 2.5 mm. To scan the whole fetus, we obtained the initial multislice set of 13 images over at the fundus of the uterus. By moving the bed upon which the woman lay in 9.1 cm steps, and repeating the multislice scan at each step, trans-axial images were obtained until the cervix was seen. The number of sets of 13 images required ranged from three to five, depending on the size of the fetus. The overall scanning time was therefore between 7.5 and 12.5 seconds. The images were monitored as they were collected, and if the fetus was observed to move during scanning, the data acquisition was repeated. The magnetic fields employed were within safety guidelines of the UK National Radiological Protection Board.28
To calculate the fetal volume and fetal organ volumes, images showing the relevant organ were selected and regions of interest were manually drawn in each image using Analyze software.29 The area of each region of interest at each slice was measured automatically by counting the number of pixels covering the region of interest and multiplying this number by the area corresponding to a pixel. This was then multiplied by the slice thickness to give the volume. The same observer (KRD) made all the measurements, but the subjects were only identified by number for magnetic resonance imaging analysis, so that the observer would not be aware of the ultrasound results. The accuracy of the volumetric estimations was limited by the error in delineating the area of interest in each slice. The intra-observer error for repeated measures from the same data set was 2% for the brain and liver measurements and 4% for fetal volume. Repeated scans were also been performed for some volunteers to obtain four volumetric estimations 5 minutes apart to assess reproducibility. The percentage errors were found to be 2–4%.
For each measurement, for the normal subjects, the data were plotted against gestational age, and then fitted to a second-order polynomial using SPSS.30 The population 90% confidence intervals (CI) were also drawn, and the data from the fetal growth restricted pregnancies were superimposed on these curves. This is shown for the magnetic resonance brain volume and ultrasonic head circumference measurements in Fig. 1. The fractions of the normal population and the fetal growth restricted population lying below the line of best fit (50% CI) were calculated (Table 2). Similar plots were created for the ratio of magnetic resonance imaging measures of brain/liver volume,15 and ultrasonic head circumference/abdominal circumference ratio. The fractions of the normal population and the fetal growth restricted population lying below the 90% CI were calculated (Table 3). The data in Table 3 were used to calculate the sensitivity and positive predictive value of using ultrasound and magnetic resonance imaging measurements lying below the CI for the normal population, as a diagnostic test for fetal growth restriction. The magnetic resonance imaging head volume data were plotted against the ultrasonic head circumference data for the normal and fetal growth restricted pregnancies and a Bland–Altman plot was used to compare the ultrasound and magnetic resonance imaging techniques for measuring fetal weight.
Table 2. Demonstating differences in brain size measured with magnetic resonance imaging and ultrasound for the normal and abnormal groups.
Fetal growth restricted
Individualised birthweight ratio < 10th centile
Individualised birthweight ratio ≥ 10th centile
Magnetic resonance brain volume
Ultrasound head circumference
Table 3. Numbers of fetal growth restricted and normal subjects for which each measurement lay above or below the 10% CI drawn for the normal population. The resulting sensitivity and positive predictive value of using the 10% CI for each measurement as a diagnostic test is also shown.
Fetal growth restricted
Positive predictive value (%)
Individualised birthweight ratio < 10th centile
Individualised birthweight ratio ≥ 10th centile
Magnetic resonance brain volume
Magnetic resonance fetal weight
Ultrasound fetal weight
Magnetic resonance lung volume
Magnetic resonance liver volume
Ultrasound abdominal circumference
Magnetic resonance brain volume/liver volume
Ultrasound head circumference/abdominal circumference
Figures 1a and 1b show the MR brain volume and ultrasonic head circumference measurements plotted against gestational age for the normal group (with its 90% CI) and the fetal growth restricted group.Table 2 summarises the number of subjects lying below the line of best fit (50% CI). The ultrasonic head circumference measurements suggest that brain sparing has occurred particularly at later gestations, whereas the magnetic resonance imaging data demonstrate an apparent overall reduction in brain volume in fetal growth restriction at all gestations. Similar graphs for the other measurements showed that the magnetic resonance and ultrasound assessments of fetal weight, the magnetic resonance liver volume and the ultrasound abdominal circumference were all reduced in fetal growth restriction (in all cases, the data for all the fetal growth restricted subjects lay below the line of best fit for the normal subjects, except for one subject for ultrasound abdominal circumference). The fetal lung volume was less affected in fetal growth restriction particularly at late gestation (only 6/41 fetuses with fetal growth restriction had magnetic resonance imaging lung volume measurements outside the normal population defined by the 90% CI, and 8/41 were above the line of best fit.). The ratios of magnetic resonance brain/liver volumes and head circumference/abdominal circumference were elevated in the fetal growth restricted group, and some of the abnormal subjects showed considerable variation from the normal population 90% CI for the magnetic resonance brain/liver ratio.
Table 3 summarises the sensitivity and positive predictive value of the various magnetic resonance and ultrasound measurements are used as a diagnostic test, based on the normal population 90% CI. Specificity is not quoted as it would be biased because the CIs were defined on this group of normal subjects. The ultrasound fetal weight estimation was found to be the best predictor of fetal growth restriction.
Figure 2 shows the magnetic resonance imaging head volume data plotted against the ultrasonic head circumference data for the normal and fetal growth restricted pregnancies.Figure 3 shows a Bland–Altman plot that can be used to compare the ultrasound and magnetic resonance imaging techniques for measuring fetal weight.
This study compared fetal organ size measured using magnetic resonance imaging and two-dimensional ultrasound in normal and fetal growth restricted pregnancies. The parameters studied were chosen on the basis of a previous finding14 that a single magnetic resonance imaging fetal liver volume measurement, performed several weeks before delivery, could distinguish fetuses subsequently diagnosed as being growth restricted with greater accuracy than magnetic resonance imaging fetal weight estimations. This study failed to support the previous finding and instead found that ultrasound estimates of total fetal weight predicted fetal growth restriction best.
The Bland–Altman plot for the magnetic resonance imaging and ultrasound fetal weight estimations (Fig. 3) shows the 95% CI for the normal population. The fetal growth restriction data have also been superimposed and suggest that there is no difference in the way in which the two techniques measure the two populations. The cause of the increased scatter observed at large volumes is not clear. The fractional random error in the magnetic resonance fetal volume measurements should decrease with increasing fetal volume, as the definition of the exact position of the boundary becomes less significant to the total volume estimate.
The insensitivity of the liver volume measurements for predicting fetal growth restriction was due to a large scatter in the magnetic resonance imaging liver volume measurements in the normal population and hence, wide CI. This was probably due to the operator dependency in outlining the fetal liver. Echo-planar imaging is a very rapid imaging sequence that effectively freezes subject motion, overcoming motion artefacts. However, it has a relatively low signal-to-noise ratio per image. Other magnetic resonance imaging sequences have been used to study the fetus,15,31 and better results might be obtained with different imaging sequences. Furthermore, semi-automated methods of image segmentation (to separate out the different fetal organs) might reduce the variability introduced by the subjective methods used here. However, the intra-observer and intrasubject variability (∼4%) found in this and previous studies is insufficient to explain the difference in variability of the magnetic resonance imaging and ultrasound data.
The lungs of fetal growth restricted fetuses are of relatively normal volume suggesting the fetus preferentially diverts nutrition to these essential postnatal organs at times of intrauterine stress.
The majority of studies investigating fetal growth restriction with ultrasound have shown the abdominal circumference to be the most useful measurement. However, its predictive value as a screening test for fetal growth restriction has been reported as being as low as 21%.32 Reports of ultrasound estimates of fetal liver size have been based on ultrasound measurement of liver length.33 This one-dimensional measurement does not provide a superior means for the prediction of a small for gestational age fetus than abdominal circumference.34 In view of the poor predictive value of liver length for fetal growth restriction, some authors have questioned the assumption that the majority of fetal growth restricted fetuses have reduced liver size.35 They suggest that the small abdominal circumference measurement may reflect a reduction in size of other intra-abdominal organs, reduced amounts of fat or possibly an elevated diaphragm because of poor lung growth. This view is contradicted by the magnetic resonance liver and lung measurements reported in this study, which agree with known physiological effects. The liver comprises the bulk of the fetal abdomen. Animal and human studies have shown diminished hepatic glycogen stores and liver mass associated with fetal growth restriction and postmortem data from unexplained stillbirths often show reduced liver size.36
The most important result of this study is the reduction in brain volume observed with magnetic resonance imaging. Our ultrasound data confirmed the well described apparent ‘brain sparing’ effect, with maintenance of relatively normal ultrasonic head circumference measurements near term. However, in contrast, we demonstrated an overall reduction in brain volume measured with magnetic resonance imaging. It is important that this change in brain volume in fetal growth restriction is studied further using high-resolution magnetic resonance imaging anatomical imaging sequences. The reduction in brain volume detected by magnetic resonance imaging is also not that surprising given the known neurological significance of fetal growth restriction. The reason why this deficit is observed with magnetic resonance imaging but not 2D ultrasound is not clear. It could simply be that the head circumference is a poor indicator of brain volume.Figure 3 plots the magnetic resonance imaging fetal brain volume against ultrasound abdominal circumference. As a volume is being plotted against a linear dimension, it would be expected that the curve would be cubic in shape. However, it is apparently close to linear, which suggests a change in the shape of the skull during gestation. There is a difference in the regression lines for the normal and abnormal populations, suggesting that there is a systematic difference in the measurements made in normal and abnormal brains during gestation. Simple geometric considerations tell us that the fractional change in volume of a sphere is three times greater than the fractional change in its circumference. This could be tested by comparing growth curves measured using 3D ultrasound with those from 2D ultrasound. An alternative explanation for the difference could be that magnetic resonance is detecting a different boundary to ultrasound, which detects the edge of the cranium. However, the echo-planar images used are strongly ‘T2*-weighted’, making it difficult to distinguish between cerebrospinal fluid and the brain, so it is likely that echo-planar imaging is also measuring the internal volume of the skull. Future studies should be performed using alternative magnetic resonance sequences, which provide more contrast between different tissues to determine if any component of brain tissue is particularly reduced in fetal growth restriction.
This lack of brain sparing may be important in improving our basic understanding the effects of fetal growth restriction on fetal development, and may have implications for the management of fetal growth restriction pregnancies.
The development of rapid magnetic resonance imaging sequences such as echo-planar imaging allows the detailed assessment of the internal fetal anatomy. The study showed that ultrasound fetal weight measurements were the best method of distinguishing fetal growth restricted fetuses in this population. The study also found an important diminution in fetal brain volume in fetal growth restriction as measured by magnetic resonance imaging.
This work was supported by a grant from the Medical Research Council. Dr Sara Lewis provided helpful statistical support.