Key words fetus - 3 D ultrasound - biometry
Introduction
Fetal growth is associated with a number of factors: population characteristics, genetics,
parity, nutrition, and environmental parameters. Various investigators have previously
constructed fetal growth charts with the use of 2 D ultrasound and the majority of
these growth charts are integrated in the current ultrasound machines. However, the
major part of these growth charts is not comparable due to the use of different study
designs, data acquisition prior to 1990 (equipment with low image resolution), different
types of measurement [i. e., BPD measurements: bone to bone as outer-to-inner [1 ]
[2 ]
[3 ], outer-to-outer [2 ]
[3 ]
[4 ]
[5 ], middle-to-middle measurements [7 ], or skin-to-skin measurements [8 ]
[9 ]] derived from different populations [6 ]
[10 ]
[11 ]
[12 ] and different statistical models and approaches [13 ]
[14 ]. A systematic review of the methodology used by ultrasound studies of fetal biometry
confirmed considerable methodological heterogeneity [15 ].
Since the publication of our first age-related reference graphs and tables for the
head and abdomen parameters and the long limb bones in 1996 [8 ], the quality of ultrasound machines and abdominal and vaginal probes has improved
continuously, enabling a more precise measurement of biometric parameters. Furthermore,
3 D ultrasound [16 ] offers a major benefit over 2 D ultrasound not only in the detection and assessment
of fetal malformations [17 ] but also in controlling and correcting biometric planes by means of the two perpendicular
planes in the multiplanar display.
Patients and methods
The study design was conducted as a prospective cross-sectional study from 2000 to
2020 and included only singleton pregnancies with gestational age confirmed by CRL
measurements before 10 weeks of gestation. All patients were of Caucasian origin and
each patient was included only once. Exclusion criteria were multiples, fetal malformations,
chromosomal abnormalities, intrauterine growth restriction or macrosomia, oligo- or
polyhydramnios, and maternal diseases (diabetes mellitus, hypertension, nicotine or
alcohol abuse, and drug consumption). Informed consent was signed by all patients.
3 D measurements were performed between 10 + 0 and 41 + 0 weeks+days of gestation
by two experienced operators, using Voluson 730, Voluson E8, Voluson E8 Expert, or
Voluson E10 (GE Zipf, Austria) ultrasound equipment with an abdominal 3 D (4–8 MHz)
or transvaginal 3 D probe (5–9 or 6–12 MHz). The measurements were performed in the
A-plane of the multiplanar display after aligning the planes under control of the
two perpendicular planes into correct standard anatomical planes. BPD and OFD were
measured in the exact axial plane of the fetal head at the level where the midline
echo is broken by the cavum septi pellucidi in the anterior third and the anterior
and posterior horns of the lateral ventricles are seen [4 ]
[18 ]. ATD and ASD were measured through a transverse section of the fetal abdomen at
the level of the stomach with a short demonstration of the umbilical vein entering
the portal sinus as indicated by Hansmann [4 ] and Campbell and Wilkin [19 ]. Head and abdomen diameters were taken as outer-to-outer measurements, i. e., exactly
from the outer boundary of the skin to the outer boundary of the skin ([Fig. 1 ], [2 ]). Measurements of the long limb bones included the ossified shaft while the bone
was visualized in a horizontal position to the probe ([Fig. 3 ]). HC was computed using the formula 20 ]
[21 ]. For AC the approximate elliptical formula AC = (ATD+ASD)/2 × 3.142 [8 ]
[21 ] was used. Finally, only cases in which all 12 biometric parameters (BPD, OFD, HC,
ATD, ASD, AC, Fe, Ti, Fi, Hu, Ra, and Ul) were recorded were taken into account.
Fig. 1 Multiplanar demonstration of BPD and OFD measurement (a = axial plane, b = coronal plane, c = sagittal plane). 27 + 0 weeks of gestation.
Fig. 2 Multiplanar demonstration of ATD and ASD measurement (a = axial plane, b = coronal plane, c = sagittal plane). 27 + 0 weeks of gestation.
Fig. 3 Multiplanar demonstration of femur measurement (a = sagittal plane, b = axial plane, c = coronal plane). 27 + 0 weeks of gestation.
Statistical Methods
The distribution-free method we used for the construction of reference bands consists
of three major steps.
Step 1:
Fitting a nonlinear regression model in order to determine a central line around which
the band has to be spanned. The form of this regression function must reflect the
way in which the age-specific mean values of the quantity Y under analysis change over time. The model which turned out to be particularly suitable
for the fetal growth characteristics considered in this paper is given by the equation:
The symbols appearing in this formula have the following meaning:
t predicted according to the model;
t in the population (in the database underlying this paper, the values of t’ and t’’ are 10.0 and 41.0 weeks);
c = scale parameter;
d = shift parameter;
= incomplete beta integral with parameters a , b evaluated at x.
The model parameters a, b, c, and d have to be estimated by means or ordinary least squares fit from the data.
Step 2:
Determining the ratio ρ between the width of the reference band at t’’ and t’ . This constant has to be chosen in a way reflecting possible differences in variability
of Y at the boundaries of the time range.
Step 3:
Calculating the lower and upper reference limit at time t from the formulae
where λ
l
and λ
u
are computed by means of an iteration algorithm as the smallest values for which
the percentage of data points falling below and above the curve corresponding to y
⁎ (t ) and y
* (t ), respectively, is at least 5 %. In other words, the ordinates of the data points
falling on the lower and upper boundary of the band correspond to smoothed 5th and 95th percentiles, respectively. A detailed description of the statistical method can be
found in [14 ].
Assessment of the reliability of the measurements making up our database was done
calculating for the i th patient of a selected subsample of size n = 100 the quantity
RelDevi = 100*|X
i
(1) – X
i
(2) |/ (X
i
(1) + –
X
i
(2) ),
where X
i
(1) and –
X
i
(2) denotes the value noted by Examiner 1 and Examiner 2, respectively. The information
contained in these percentage interobserver deviations was summarized by calculating
their means and standard errors.
Results
A complete biometric profile including all 12 parameters was obtained in 10 225 fetuses
with normal outcome.
The mean percentage of the interobserver deviation between the two examiners was 0.504 ± 0.04,
0.534 ± 0.05, and 0.554 ± 0.06 for BPD, ATD, and Femur, respectively.
[Fig. 4 ] shows the reference band constructed for HC, together with a scatterplot of all
individual measurements contained in our database for this quantity. The 90 % reference
bands for 12 ultrasound parameters are shown in [Fig. 5a–f ], [6a–f ]. A complete account of the numerical results behind these graphical representations
can be found in [Table 1 ].
Fig. 4 Scatterplot of head circumference raw data between 10 and 41 weeks of gestation with
superimposed fitted 5th percentile, mean, and 95th percentile.
Fig. 6 Fetal long bones (a Femur, b Tibia, c Fibula, d Humerus, e Radius, f Ulna) with fitted 5th percentile, mean, and 95th percentile.
Table 1
Estimated model parameters and numerical results determining the reference bands shown
in [Fig. 4 ], [5 ], [6 ]. (For the definition of the symbols appearing in the column headings, see the statistical
methods section.)
Y
a
b
c
d
ρ
λl
λu
BPD
1.14 943
1.52 649
84.0244
0.19 919
1.72 041
4.1650
4.1719
OFD
1.18 451
1.82 477
100.730
0.18 258
1.73 397
4.4766
4.6094
ATD
0.96 609
1.10 274
100.748
0.11 067
2.36 895
4.2346
4.2031
ASD
1.01 940
1.18 073
100.158
0.11 562
2.28 844
4.5117
4.4277
HC
1.16 200
1.69 207
305.053
0.18 792
1.65 794
12.6719
13.2891
AC
0.99 287
1.14 246
315.606
0.11 314
2.20 884
11.8350
12.3223
Fe
0.91 345
1.35 330
75.5703
–0.00 847
1.38 145
2.9467
3.0211
Hu
0.85 826
1.43 654
66.1801
–0.01 244
1.12 215
3.1406
3.0684
Ti
1.00 207
1.57 249
64.2636
–0.00 593
1.25 579
2.8223
2.8438
Fi
0.98 402
1.52 876
63.2465
–0.01 658
1.23 750
2.8887
2.8330
Ra
0.88 751
1.62 679
54.6429
–0.03 555
1.19 848
2.8699
2.8906
Ul
0.81 378
1.42 009
63.6335
–0.04 321
1.12 596
3.0898
3.0749
The fitted values (5th percentile, mean, and 95th percentile) of all parameters are presented in Appendix-Table 1, 2 .
The comparison of the described new growth charts with our charts published in 1996
(8) shows slightly higher values in the head and abdomen parameters after 24 weeks
of gestation in the new charts but similar values in the long limb bones ([Fig. 7a–c ]).
Fig. 7 Comparison of our old fitted growth charts of a HC, b AC and c Femur (red color) [7 ] with the new fitted growth charts (blue color) (5th percentile, mean, and 95th percentile).
The comparison of our new growth charts with a selection of growth charts published
by other authors [1 ]
[4 ]
[5 ]
[9 ]([Fig. 8a–c ]) demonstrates higher mean values of the head parameters in our study, while the
mean abdomen values are similar to the mean values reported by Hadlock [1 ], Knitza [9 ] and Snijders [5 ], at apparently higher mean values reported by Hansmann [4 ]. The mean values of femur length are similar to the values reported by the compared
other growth charts.
Fig. 8 Comparison of our new growth charts for a HC, b AC and c Femur (5th percentile, mean, and 95th percentile (blue color)) with the corresponding growth charts (50th percentiles) of Hadlock [1 ] (red color), Hansmann [4 ] (yellow color), and Knitza [8 ] (green color).
The comparison of the variability of our new growth charts with the data of Snijders
et al. [5 ] confirmed a lower variation of the 90 % range towards the right-hand boundary of
the range of gestational age in the head and abdomen parameters as well as in the
femur lengths in our charts ([Fig. 9a–c ]).
Fig. 9 Comparison of our new growth charts for a HC, b AC and c Femur (blue color) (5th percentile, mean, and 95th percentile) with the 5th , 50th and 95th percentiles of the growth charts of Snijders et al. [5 ] (red color).
Discussion
Ultrasound has undergone tremendous improvement with regard to 2 D and 3 D technology
and image quality over the past three decades. This has had a significant impact on
the delineation of fetal structures, enabling more precise biometric measurements.
Furthermore, 3 D ultrasound permits storage of volumes and volume manipulation with
the use of the multiplanar mode. The demonstration of all three perpendicular 2 D
planes (A, B, and C plane) on the monitor allows identification of the correct biometric
plane, and correction of the A plane with the rotation controls in all three dimensions.
Consequently, control of the A plane by the two perpendicular planes enables the operator
to detect and correct any imprecise standard plane before performing the measurement.
Updates in growth charts are not only useful with regard to an improvement of the
conditions for measurements to obtain more precise data, but also in the detection
of fetal growth alterations when comparing the new charts with the older ones. In
a recently published study, Knitza et al. [9 ] found an increase in fetal growth within one generation. A similar finding was noted
in our study due to the fact that we used the same standard measurements with the
same caliper placements as in our previous study. Comparing our new charts with the
growth charts we published in 1996 [8 ], we identified a slight increase in head and abdomen circumference but not in the
long limb bones.
The comparison of our new growth charts with charts published in the literature is
shown in [Fig. 8 ], [9 ]. The greatest difference is found in HC, due to different measurements recorded
for the biparietal diameter (BPD) ([Fig. 8 ]) as well as for the range of the reference bands ([Fig. 9 ]).
A strength of our study is the homogeneity of the population from which the sample
was taken. The resulting reference bands are comparatively narrow and allow early
detection of deviations from a normal growth process. This is in contrast to the results
of the fetal growth longitudinal study of the INTERGROWTH-21st project [6 ], a multiethnic, population-based project intending to ensure worldwide applicability
of reference limits. However, pooling biometric measurements from different ethnic
groups with anthropological differences resulted in reference bands of increased width
as compared with the bands obtained in our study. Therefore, it seems doubtful whether
basing reference limit estimation on mixed populations is suitable for establishing
reference percentiles of worldwide applicability for purposes of early detection of
growth abnormalities. A key feature and major strength of the statistical approach
used in this study for the construction of reference bands is that it provides direct
control over the proportions of data points to be rated as being either abnormally
small or large. This means that the specificity of the diagnostic procedure relying
on the age-specific reference limits established here in the (very large) sample of
observed pregnant women precisely represents the targeted value of 95 % (except for
slight, practically irrelevant exceedances due to the discreteness of observed proportions).
In contrast, other well-established statistical approaches to the estimation of age-dependent
reference limits [22 ]
[23 ]
[24 ]
[25 ] have to rely on a specific model for the distribution of the measured quantity Y under assessment in order to enable maintenance of specificity at least in the long
run (i. e., in terms of the distribution of the coverage proportion arising from a
large number of repeated applications of the procedure to different datasets).
The model given by Equation (★) used to determine the regression line about which
the reference band is spanned, showed reasonably good fit for all measurements, with
a mean squared error being almost identical to that of a 3rd degree polynomial. In contrast to a polynomial, functions of that form are monotonic
which makes the corresponding model perfectly suitable for the analysis of data on
variables subject to a growth process.
Conclusions
The new reference charts for fetal head and abdomen parameters as well as for the
long limb bones derived from our prospective cross-sectional study enable the operator
to closely observe the growth profile of fetuses (12 growth parameters) from 10 to
41 completed weeks of gestation. Three-dimensional ultrasound – in comparison to 2 D
ultrasound – allows the demonstration of the different biometric parameters in exactly
controlled standard planes and thus enables precise measurements.
The comparison of the new charts with our charts published in 1996 [8 ] reveals a slight increase in head and abdomen size over the past two decades, while
no significant differences were observed for the limb bones over this time.
The comparison of our new growth charts with a selection of charts published in the
literature demonstrates a difference in mean values and variation within the 90 %
range.
The data from this study could be an integrative component in future automated measurement
programs controlling fetal growth profiles.
Fig. 5 Fetal head and abdomen parameters (a BPD, b FOD, c HC, d ATD, e ASD, f AC) with fitted 5th percentile, mean, and 95th percentile from 10 to 41 weeks of gestation.
