For obstetric decision-making, it is important to have as accurate an estimation of fetal weight as possible. The large number of methods published in this respect is clear evidence of the importance of prenatal weight calculation. Both fetal macrosomia and intrauterine growth restriction increase the risk of perinatal mortality and morbidity, with corresponding long-term postnatal consequences. Ultrasound methods do not estimate fetal weight directly, rather they do so indirectly by measuring defined segments of the body. Two-dimensional (2D) ultrasonography is used routinely for this purpose, and the estimated weight of the fetus is calculated using appropriate tables or integrated computer programs. The most frequently used parameters include the biparietal diameter (BPD), transverse abdominal diameter, abdominal circumference and femur length. The large number of weight formulas described also provides clear evidence that none of those usually employed is accepted universally. Even in ideal ultrasound conditions, there are still considerable measurement inaccuracies, with a mean error of 7–10%. The error is greatest at the two ends of the weight scale. In the majority of studies, low birth weights have been overestimated and high birth weights underestimated1.
None of the established formulas2–5 takes soft-tissue thickness into account, despite evidence that abnormal tissue content may be a reliable indicator of fetal growth aberrations6–8. Catalano and colleagues showed that, although the neonatal fat mass constitutes only 12–14% of birth weight, it explains 46% of its variance9. Fetal fat deposition represents approximately 90% of caloric accretion at term10. In addition, the quantity of neonatal adipose tissue may be one of the causes of underestimation of birth weight on 2D ultrasonography in cases of fetal macrosomia9, 11. In another study, Bernstein et al. found that there was a correlation between increased body fat in infants of diabetic mothers and sonographic overestimation of fetal weight12. Fetal fat deposition in the extremities was found to be characterized by an exponential increase when plotted against gestational age13.
Soft-tissue measurement using two-dimensional ultrasound examination
Several studies have used a variety of ultrasound measurements to take the soft-tissue content into account, with varying degrees of success14–29. Hoffbauer and coworkers were among the first to include fetal thigh diameter in a weight formula30. A subsequent comparative ultrasound and anatomical study of six stillborn fetuses in the third trimester found that sonographic measurements were systematically larger than the comparable anatomical measurements28. Exact positioning of the measurement plane, however, appeared to be unnecessary, as results within 1–2 cm of the reference plane in the upper thigh were of similar accuracy. The authors concluded that circumference measurements of the fetal thigh could be made in a reliable manner, and could be used to detect changes in the soft-tissue mass and possibly improve fetal weight estimation28. In an attempt to further improve fetal weight estimation, Vintzileos and colleagues performed a stepwise polynomial regression analysis including the fetal thigh circumference8. The best results were obtained by combining measurements of standard 2D parameters and thigh circumference. In the authors' opinion, thigh circumference formulas may be helpful in improving weight estimations in growth-restricted or macrosomic fetuses with quantitative disturbances in the soft-tissue mass8. A drawback of this study, however, was the lack of an evaluation group to test the newly described formula prospectively. Jeanty et al. assessed fetal growth and nutrition by calculating arm and thigh volumes from subcutaneous tissue in the limbs23. In an anatomical study, the authors compared the typical image of the transverse limb to a target, with the hyperechogenic ‘bull's eye’ representing the bone. The muscles comprised the hypoechogenic space between the bone and the hyperechogenic subcutaneous and cutaneous tissues. The obliquity of the limb section examined affected the measurement to a greater degree than did the exact plane at which it was obtained. Limb volume was found to be independent of the standard parameters commonly used for predicting intrauterine growth restriction, and was therefore able to provide complementary information23. Balouet and co-workers considerably improved fetal weight estimation by incorporating the soft-tissue thickness when measuring the fetal thigh17, 31. The cutaneous and subcutaneous limb circumference proved to be better predictors of actual weight at birth than did the abdominal circumference. Landon et al. obtained similar results in a study of fetal humeral soft-tissue thickness in 93 pregnant women with gestational diabetes24. Of the sonographic variables examined, the soft-tissue thickness in the humerus proved to be the most accurate predictor of excessive fetal size. The authors concluded that this new parameter might be able to distinguish between large fetuses with truncal obesity and symmetrically large fetuses. In the light of evidence that ultrasound methods significantly underestimated neonatal weights over 4000 g, Sood and colleagues conducted a study of 95 women at risk of having a large-for-gestational-age fetus, to determine the value of an objective assessment of humeral soft-tissue thickness in estimating macrosomia27. The soft-tissue humeral thickness correlated significantly with birth weight and ponderal index, and was able to explain up to 40% and 20%, respectively, of the variation in these values. However, a new formula incorporating tissue thickness did not result in any significant improvement in predicting fetal weight in the selected group of patients studied27.
Fetal volume calculation from two-dimensional measurements
Other attempts to assess fetal weight using volume measurements have been described, such as three-dimensional (3D) head and trunk reconstruction18, 32, 33, ultrasound measurements based on neonatal specific gravity and volume26, and modification of the two-compartment model of fetal volume19. The rationale behind these efforts was that fetal weight ought to be directly proportional to fetal volume19. In all of the formulas used, however18, 19, 26, 32, 33, the results were obtained by reconstructing 2D measurements rather than from direct volumetry.
With the recent advent of 3D ultrasonography, reproducible measurements of circumference and volume have become possible through simultaneous visualization of three orthogonal fetal limb sections. This technique has the advantages of efficiency and speed in comparison with conventional ultrasonography34.
In the first few years after it was introduced, 3D imaging was used mainly to obtain limb circumference measurements at the exact midpoint of the humerus and femur length20, 21. Favre et al. developed four models including thigh and upper-arm circumference: one for the whole population and one for three subgroups previously established on the basis of the abdominal circumference centile20. The data from the study groups confirmed the value of measuring fetal thigh circumference in small-for-dates fetuses, and arm circumference in the other groups.
Chang, Liang and coworkers were among the first to introduce true 3D volumetry for fetal weight measurement into clinical practice14, 15. Direct limb volumetry compared favorably with 2D measurement techniques as the fetal limb is not a perfect cylinder14. In these studies, single-parameter volumetry of the fetal arm and thigh were more accurate in predicting weight at delivery than were conventional 2D equations based on several biometric parameters14, 15. In principle, these results were confirmed by other studies of fetal volumetry, although application of the published 3D weight formulas in a predominantly Caucasian population led to significant overestimation of fetal weight16. Further research demonstrated that use of a polynomial 3D equation incorporating arm and thigh volumes, as well as the standard 2D abdominal transverse diameter, led to superior results in comparison with conventional weight formulas16. In another study, Schild et al. showed that a combination of three different volumes (upper arm, thigh and abdomen) with the BPD produced the highest measurement accuracy for the estimated birth weight35.
Lee and colleagues reported on a pilot 3D study assessing the volume of a specified abdominal and thigh cylinder25. The former was defined as the volume 1.5 cm adjacent to the reference plane for abdominal measurements on each side, whereas the latter was measured as a 2-cm cylinder in the mid-thigh region, in both cases including the subcutaneous tissue. Preliminary findings in 18 term fetuses suggested that accurate birth weight prediction appeared to be feasible using fetal volume parameters. However, the results of this study were obtained from the same population as that from which the formula had initially been derived, introducing an inherent bias into further conclusions25.
Song et al. used a modified 3D measurement of the thigh volume to produce a new weight formula, which allowed more accurate measurement of birth weight in the evaluation group than that obtained with the Hadlock and Shepard formulas36. The authors themselves addressed what represents a fundamental problem with their own and many other studies of weight estimation, that in many cases the newly described weight formulas and those used for comparison are not derived from the same population, so that direct comparison is not justified.
Instead of measuring volumes in the whole upper arm and thigh, Lee et al. measured the fractional limb volume, which is based on 50% of the diaphyseal length37. The authors considered that the advantages of this new parameter were its shorter measuring time and reduced susceptibility to error caused by acoustic shadowing at the extreme ends of the diaphysis. The best predictive model for birth weight consisted of a combination of the abdominal circumference and fractional thigh volume. However, the evaluation group was relatively small, with 30 fetuses, and the formula was established and evaluated only in fetuses with previously normal growth.
In a small recently published study, it was hypothesized that upper-arm and thigh volumes measured by 3D ultrasound scan correlate strongly with weight estimation in large and small fetuses, and can be used to estimate fetal weight38.
Pang et al. compared the accuracy of 3D volumetry using various numbers of image planes to outline the region of interest39. In this in vitro study there was a high level of accuracy in measuring regularly and irregularly shaped bodies. The most accurate volume measurement was obtained with the maximum number of levels. However, reducing the number of imaging planes did not lead to a significantly poorer result and had a substantial advantage in that it was able to reduce the time needed for measurement by 50%.
Attempts to estimate the overall volume of the embryo or fetus have so far had to be limited to the first and early second trimesters of pregnancy, because the fetus is too large for whole-body calculations during the third trimester40–46. However, this should be qualified by noting that only two of these studies took the extremity volumes into account, which at this gestational age represent an estimated 10% of the total volume or weight40, 42. Aviram et al. carried out the volume measurements using a commercially available software program and did not provide any more detailed explanation of their measurement methodology40. Blaas et al. were able to assess the overall volume of the fetus using new software with simultaneous segmentation of the head, trunk and extremities42. The authors recommend that all 3D ultrasound devices should be equipped with software capable of using the same dataset to provide a combination of several volume measurements of various regions of the body at the same time point42.
In general, the same potential errors occur with 3D volumetry as with 2D measurements of the fetus, as reviewed previously47. In addition, acoustic shadowing at the extreme ends of the diaphysis can complicate soft-tissue volumetry37. Furthermore, a single formula is not capable of covering the entire range of fetal weights; instead, special formulas are required for various weight groups, particularly at the two extremes of the weight scale48. Further support for this view is provided by the clinical experience that the weight of small fetuses tends to be overestimated, whereas that of large fetuses tends to be underestimated37. Studies relevant to this topic are shortly to be completed.
Magnetic resonance imaging and fetal weight measurement
Fetal volumetry has also become a field of interest for magnetic resonance imaging (MRI), as the increasing number of studies on the subject shows1, 49–54. There have been reports on the use of several fast acquisition protocols, including echoplanar MRI and T1-weighted and T2-weighted imaging for fetal volume calculation1, 49–54. The data so far do not suggest any dependency of the measurement accuracy on the protocol used54. Fetal volume can be calculated either using the Cavalieri principle or semiautomatically with special software. None of the programs available currently allows fully automatic volume assessment54. For weight calculation, fetal volume is multiplied by fetal density, although the exact value for the latter is not known. In addition, fetal density is also a function of gestational age owing to changes in the proportions of tissue represented by muscle, bone and fatty tissue, for example54. In the majority of cases, the authors selected a density value of 1.031 g/mL51, 52, 54, 55, but values of 1.0 g/mL50 and 1.07 g/mL53 were also used. Weight calculations using MRI showed greater accuracy than ultrasound estimates1, 51, 53, 54. However, the great majority of the measurements were carried out at term1, 51–54. As less adipose tissue is present at early gestational ages and in smaller fetuses, weight predictions by MRI carried out earlier than 37 gestational weeks may be less accurate. Therefore, future studies also need to include fetuses of earlier gestational ages and those with and without growth restriction51. A different density value may need to be calculated for this purpose. Reported disadvantages of the method include its higher cost and the longer time needed, which averages 45 min, including postprocessing time54.
As volumetric measurements can improve fetal weight estimation, modern prenatal medicine would now be inconceivable without these methods. Despite this, several technical and methodological problems need to be overcome before this technique can be adopted into everyday clinical practice. It is currently unclear which fetal organ systems should be included in volumetry in order to achieve as accurate a weight estimate as possible. In addition, it needs to be clarified whether volume calculations alone, or only those in combination with conventional 2D parameters for weight estimation, should be used. As none of the formulas is capable of covering the entire range of the weight scale, specific formulas for defined weight classes or gestational intervals should be produced. It would be desirable for the currently available software to be optimized in such a way that simultaneous segmentation of the whole fetal body is possible, particularly at early gestational ages. MRI technology also promises further improvements in fetal weight estimation.