- Top of page
The main imaging modality used in visualizing the fetal anatomy is two-dimensional (2D) ultrasound. While it has become a useful tool in assessing fetal organ structures and their anomalies, its limitations must be acknowledged: 2D ultrasound lacks adequate diagnostic capacity to predict in utero fetal organ function.
In cases of fetal renal pelvis dilatation, there is an ongoing search for the best renal parameter to predict neonatal outcome1, 2. A widely accepted parameter is the anteroposterior (AP) diameter of the fetal renal pelvis. The threshold value, however, to distinguish between physiological and pathological dilatation is still under debate; Corteville et al. proposed that a threshold of ≥ 4 mm before 33 weeks' gestation and ≥ 7 mm after 33 weeks' gestation should warrant postnatal follow-up3.
Postnatal renal volume and growth are important parameters for evaluating and monitoring several diseases in pediatric urology4. Yet, antenatal renal volume measurements are not integrated routinely into the assessment of the fetal kidney. Furthermore, the clinical applications of fetal renal volume, such as in the assessment of obstructive renal disease, and subsequent neonatal outcome have not yet been reported. The presently used AP diameter of the fetal renal pelvis is a measurement in one orthogonal imaging plane and is generally considered to be indicative of the volume of the whole fetal renal pelvis. Yet, even when two other fetal renal pelvis parameters, the transverse and longitudinal diameters, are included, renal volume estimation is inadequate: volume calculations based on the ellipsoid formula (volume = length × width × thickness × π/6) are known to be inadequate in irregular, non-symmetrically shaped organs and underestimate kidney volume with a 25% error rate5. With the latest new developments in the field of three-dimensional (3D) ultrasound, accurate volume estimations of different structures can be assessed and this technique has gained widespread application in different medical fields6, 7. To date, prediction of ovarian volume, fetal kidney volume and fetal trunk volume have been assessed using 3D ultrasound in reproductive and fetal medicine8–10. Before fetal renal pelvis volume calculations can be used as a better alternative, in comparison with the conventional renal parameters, to predict neonatal outcome in hydronephrotic patients, the reproducibility of fetal renal pelvis dilatation must be established. In this study we aimed to assess the reproducibility of fetal renal pelvis volume measurements in hydronephrotic kidneys using transabdominal 3D ultrasound.
- Top of page
Our study population consisted of 15 kidneys, one from each of 15 fetuses in the second and third trimesters of pregnancy, with hydronephrosis suspected on the basis of sonographic appearance. All ultrasound examinations were conducted at the Department of Obstetrics and Gynecology, University Hospital, Maastricht, The Netherlands. The criterion for hydronephrosis adopted for the purposes of this study was renal pelvis AP diameter ≥ 5 mm in the second or third trimester of pregnancy. The 15 patients enrolled in the study visited the ultrasound unit for routine or clinically indicated examinations and written informed consent to participate was obtained from the women.
Transabdominal ultrasound examination was performed with a 4–8-MHz array 3D probe using a Voluson 730 Expert (GE Medical Systems, Zipf, Austria) ultrasound system. Examinations and volume acquisition were performed by one trained observer, and contour definition and assessment of the fetal renal pelvis volume was done by six independent observers. By using one volume acquisition per kidney, variation due to actual physiological changes in renal pelvis size could not occur. After visualizing the fetal kidney in a sagittal 2D plane, the 3D mode was activated, and the volume box was placed over the entire fetal kidney. Care was taken to keep the transabdominal probe steady during volume acquisition. The fetal kidney was scanned in slow sweep mode (high quality 2) in order to obtain good resolution. The volume sweep angle was set at 65°. The acquired 3D volume was stored immediately using the Sonoview program until it was retrieved for further analysis. Only cases with clear 3D ultrasound images of the fetal kidneys, without fetal movement artifacts and acoustic shadowing of the vertebral column, were included in the final analysis.
The acquired volume dataset was analyzed using the Virtual Organ Computer-aided AnaLysis (VOCAL™) imaging program (version 4.0), which is integrated into the Voluson 730 Expert ultrasound system. Volume assessment was based on manual delineation of the fetal renal pelvis contour (Figure 1). A rotation step of 6° was used, resulting in the definition of 30 contours for each kidney, in order to provide a sufficient number of contours to calculate an accurate fetal renal pelvis volume. All kidneys were magnified by × 2.02 in order to obtain a clear view of the fetal renal pelvis. Each observer calculated target volumes for all 15 of the hydronephrotic kidneys to establish the interobserver reliability, and one observer (L.D.) acquired the fetal renal pelvis volume twice to assess the intraobserver reliability. Before the beginning of this study three of the six observers were already experienced with volume acquisition and contour definition using the VOCAL imaging program. The other three observers were all experienced sonographers, but unfamiliar with 3D ultrasound. These three sonographers learned the 3D measurement technique primarily for the purposes of this study by attending a short training program. All observers were blinded to their own volume acquisition after contour definition by covering the display on the computer workstation.
Figure 1. Three-dimensional ultrasound image showing measurement of the fetal renal pelvis volume after manual delineation of the fetal pelvis contour using the Virtual Organ Computer-aided AnaLysis (VOCAL™) imaging program.
Download figure to PowerPoint
Statistical analysis was performed using the Statistical Package for the Social Sciences (SPSS Release 11.5, Inc., Chicago, IL, USA). Departure from a normal distribution was assessed using the Kolmogorov–Smirnov one-sample test and normal plots of within-subjects residuals. If the data were skewed, a logarithmic transformation was performed before statistical analysis to produce a normal distribution. The intraobserver reliability, defined as the ability of a test to estimate the overall correlation between all possible values within the variable taken by a single observer, and interobserver reliability for observations by different observers, were assessed by intraclass correlation coefficients (ICCs). Bland–Altman plots were constructed11. A two-way mixed model was assessed for calculating the overall interobserver ICC for the six observers. To evaluate whether the measurements of experienced observers were more reproducible compared with the measurements of inexperienced observers, the interobserver ICC between pairs of observers were calculated. In addition, the coefficient of variation (CV) and the repeatability coefficient (r) were calculated. The CV is the ratio of the within-subject SD to the mean expressed as a percentage, and was calculated according to the method described by Bland and Altman12. The various components of variance required in the calculations were estimated by analysis of variance (ANOVA) tables. r is the maximum difference that is likely to occur between repeated measurements and can be defined as 1.96 × √2σω2, where σω2 is the within-subject variance.
- Top of page
After contour definition, the fetal renal pelvis volume was measured successfully in all 15 fetuses by each of the six observers and by one of these observers twice, with a total of 105 volume calculations. The fetal pelvis volume measurements ranged from 0.27 to 11.86 cm3. Table 1 presents the minimum, maximum, mean and SD values for the fetal pelvis volumes. The measurement error of each subject did not follow the assumption of a normal distribution, so data were log-transformed before statistical analysis.
Table 1. Descriptive statistics for fetal renal pelvis volume for each of six observers
|Observer||Renal pelvis volume (cm3)|
The intraobserver reliability, with a CV of 10.84% and a high ICC value of 0.996 (95% CI, 0.988–0.999), indicated good repeatability. Figure 2 displays the intraobserver difference between the first and the second set of measurements plotted against the mean according to Bland and Altman11. There was no apparent relationship between the intraobserver difference and the mean volume of the fetal renal pelvis.
Figure 2. Intraobserver difference plotted against the mean of two measurements for fetal renal pelvis volume with mean difference (——) and 95% limits of agreement () indicated.
Download figure to PowerPoint
The interobserver reliability for all six observers, with a CV of 15.67% and a mean ICC of 0.998 (95% CI, 0.998–0.999), indicated good repeatability. Figure 3 shows the interobserver difference in fetal renal pelvis volume based on the mean of pairs of measurements. Again, there was no apparent relationship between the interobserver difference and the mean volume of the fetal renal pelvis.
Figure 3. Interobserver difference plotted against the mean of pairs of measurements for fetal renal pelvis volume with mean difference (——) and 95% limits of agreement () indicated.
Download figure to PowerPoint
The ratio of the maximum difference between random measured volumes (r) was slighty lower within (r = 1.32) than between (r = 1.52) observers. Table 2 shows the interobserver reliability for all possible pairs of operators. The interobserver pairwise ICC varied between 0.994 and 0.999 and the accompanying CV and r showed slightly higher values when comparing experienced with inexperienced observers.
Table 2. Interobserver reliability: intraclass correlation coefficient (ICC), coefficient of variation (CV) and repeatability coefficient (r) for all possible pairs of observers
|Observers||ICC (95% CI)||CV (%)||r|
| 1 and 2||0.994 (0.981–0.998)||10.84||1.32|
| 1 and 3||0.994 (0.981–0.998)||10.84||1.32|
| 2 and 3||0.999 (0.997–1.000)||7.55||1.23|
| 1 and 4||0.994 (0.981–0.998)||10.84||1.32|
| 1 and 5||0.997 (0.990–0.999)||10.84||1.32|
| 1 and 6||0.996 (0.987–0.999)||19.53||1.62|
| 2 and 4||0.998 (0.995–1.000)||13.34||1.41|
| 2 and 5||0.997 (0.992–0.999)||10.84||1.32|
| 2 and 6||0.997 (0.994–0.999)||17.68||1.58|
| 3 and 4||0.999 (0.997–0.999)||13.34||1.41|
| 3 and 5||0.998 (0.994–0.999)||13.34||1.41|
| 3 and 6||0.997 (0.991–0.999)||15.68||1.51|
| 4 and 5||0.998 (0.992–0.999)||13.34||1.41|
| 4 and 6||0.995 (0.985–0.998)||10.84||1.32|
| 5 and 6||0.997 (0.991–0.999)||19.53||1.62|
In conclusion, repeatability between experienced and inexperienced observers was good and depended on the sonographer's experience with the 3D technique.
- Top of page
3D volume calculations have been used in various different research settings in order to assess their value in the clinical setting, for example in the estimation of fetal lung volume to predict neonatal lung hypoplasia13. The aim of this study was to demonstrate that 3D ultrasound is able to provide a reproducible measurement of the fetal renal pelvis volume in an offline setting. A possible clinical application of assessing fetal renal pelvis volume could be to provide a new parameter with which to distinguish fetal renal pathology from physiology, in order to predict neonatal outcome. In fact, fetal renal pelvis volume could be the sole marker in predicting the need for postnatal follow-up and treatment. Furthermore, Riccabona et al. recently reported that renal parenchymal volume and relative renal size in patients with hydronephrosis is comparable with split renal function determined by scintigraphy in the neonatal period14. The reproducibility of 3D ultrasound and magnetic resonance imaging in obtaining renal volume has been proved to be accurate in in-vitro and in-vivo models, and reference charts of fetal renal volume have been established by 3D ultrasound5, 6, 9, 15, 16.
As Järvelä et al. describe, there are two main sources of reliability involved in using the 3D technique: 3D volume acquisition and volume calculation after contour definition8. Clearly, the reliability of 3D volume acquisition affects the ability to examine the reliability of contour definition. In our study, ultrasound examinations and volume acquisition were performed by one trained observer, while contour definition and assessment of the fetal renal pelvis volume was done by six independent observers. Our study found that both intra- and interobserver reliability for fetal renal pelvis volume measurements after contour definition were very good, even when the fetal renal pelvis volume was small. Both intra- and interobserver ICCs were high and were accompanied by low CV values. Although the overall reproducibility in this study was good, for pairs of observers in which one was experienced and one was inexperienced, the measurement error was higher compared with that for pairs involving two experienced observers, as indicated by the higher CV scores. This illustrates that the 3D technique is dependent on the experience of the observer; training is essential to accustom an operator to the scanning technique and viewing of the modalities.
Besides experience of the observer, there are other potential sources of error that may distort reproducibility. 3D sonography and acquisition of a good volume relies on the quality of the 2D image. As in 2D ultrasound, maternal obesity, acoustic shadowing by the fetal bones and fetal movement can distort the 3D volume acquisition and delineation, leading to poorer reproducibility. Another intrinsic restriction lies in the technical aspects of the program used. With the VOCAL imaging program, the observer chooses a rotation step for manual delineation of the target volume. Raine-Fenning et al. showed that using a rotation step of 6° or 9° is a reliable and valid method for volume calculations and also stated that volume measurements of irregularly shaped objects are less valid compared with those of regularly shaped volumes7. Therefore, in this study we set the rotation step at 6°, rendering 30 images of irregularly shaped volumes to be delineated, but at the cost of making it more time-consuming.
Although we did not measure the actual time spent delineating the fetal renal collecting system and performing the volume calculation, we acknowledge that this method is more time-consuming compared with measuring the AP diameter of the fetal renal pelvis with 2D ultrasound. However, the high false-positive rate of the latter and its medical costs postnatally must be weighed against the time involved in 3D volume calculation prenatally and the benefits of early detection of those actually at risk of renal deterioration. Volume calculations could be made with a larger rotation step, which would speed up the time involved, but at the expense of poorer reliability. Future investigations are required to examine the feasibility of fetal renal pelvis volume measurements in a clinical setting.
Another approach for volume calculation is threshold segmentation, which calculates the proportion of the fetal renal pelvis volume to the total renal volume after delineating the contour of the fetal kidney. This technique is restricted to the same sorts of limitation as those mentioned above. However, calculating threshold volume can be less time consuming in the sense that delineation of the fetal kidney contour is easier than is delineation of the fetal renal collecting system. Also, the rotation step could be set at 9° instead of 6°, because the outer shell of the kidney is more symmetrical in shape than is the fetal renal collecting system. Reproducibility of fetal pelvis volume calculations with threshold segmentation, however, has not yet been established.
In conclusion, with 3D ultrasound it is technically feasible for different observers to reproduce fetal renal pelvis volume measurements in an off-line setting using the VOCAL imaging program, even when the fetal renal pelvis volumes are small. Further research to establish its clinical applicability in the prediction of neonatal outcome is warranted.