C.F.E.F. investigators: Marc Althuser, Bernard Benoit, Marie Pierre Bodin, Myriam Chami, Corinne Courtiol Borderie, Christian Delattre, Christine Eglin, Gracianne Gerves, Jean Guillon, Philippe Kolf, Eve Le Goff, Dominique Marchal André, Daniel Moeglin, Annick Réali, Claude Talmant and Monique Yvinec; France
Fetal lung volumetry using two- and three-dimensional ultrasound
Article first published online: 13 JAN 2005
Copyright © 2005 ISUOG. Published by John Wiley & Sons, Ltd.
Ultrasound in Obstetrics & Gynecology
Volume 25, Issue 2, pages 119–127, February 2005
How to Cite
Moeglin, D., Talmant, C., Duyme, M. and Lopez, A. C. (2005), Fetal lung volumetry using two- and three-dimensional ultrasound. Ultrasound Obstet Gynecol, 25: 119–127. doi: 10.1002/uog.1799
- Issue published online: 28 JAN 2005
- Article first published online: 13 JAN 2005
- Manuscript Accepted: 18 SEP 2004
- lung volume;
- pulmonary hypoplasia;
- three-dimensional ultrasound;
To compare methods of measuring fetal pulmonary volume and to establish nomograms of fetal pulmonary volume according to gestational age for the accurate diagnosis of pulmonary hypoplasia.
Three methods of measuring fetal pulmonary volume in 39 normal fetuses were compared: two-dimensional (2D) ultrasound measurement assuming that the lung is a geometrical pyramid, three-dimensional (3D) ultrasound using the VOCAL rotational method, and the conventional multiplanar 3D mode. Linear regression was used to construct an equation for 3D volume calculation from 2D measurements (the re-evaluated pulmonary volume equation (RPVE)). Lung volume measurements were recorded from 622 singleton fetuses in order to construct nomograms.
There was no statistically significant difference between the lung volume values obtained using the two 3D modes. However, in comparison with the 2D measurements the volumes obtained were larger (mean difference = 11.99, P < 0.1 × 10−6). The relationship between the 2D and 3D volumes was determined using a statistical linear regression method: RPVE (mL) = 4.24 + (1.53 × 2DGPV), where 2DGPV (2D geometric pulmonary volume) = (surface area right lung base (cm2) + surface area left lung base (cm2)) × 1/3 height right lung (cm). Two nomograms were constructed, one for use with 2D and one for 3D technology.
2D pulmonary volume assessment can be used in clinical situations where fetal prognosis depends on lung volume and its growth potential. It is routinely available and easy to perform particularly when repeat measurements are required in evaluation of lung growth. We therefore propose this method as an alternative to magnetic resonance imaging or 3D ultrasound. Copyright © 2005 ISUOG. Published by John Wiley & Sons, Ltd.
Pulmonary hypoplasia occurs frequently, being found in 7–10% of neonatal autopsies and in up to 50% of cases in which other associated congenital abnormalities are present1, 2. The clinical presentation ranges from acute respiratory distress with neonatal death to various degrees of chronic respiratory failure, and complications include pulmonary hemorrhage, bronchopulmonary dysplasia and transient respiratory distress. The causes of pulmonary hypoplasia are numerous and varied and include thoracic compression, lack of fetal movement and severe oligohydramnios. These disorders can occur when there are renal, skeletal or muscular abnormalities, in the presence of congenital diaphragmatic hernia or when membranes rupture prematurely. There are several pathological definitions of pulmonary hypoplasia: decreased dry weight of the lung, decreased lung- to body-weight ratio, decreased radial alveolar count and/or decreased DNA content of the lung1, 3, 4.
In practice, it is difficult to predict which fetuses are going to have a fatal outcome as a result of lung hypoplasia. Lung height, lung diameter, thoracic circumference, rib length, pulmonary surface area, pulmonary to abdominal surface area ratio and Doppler velocimetry of pulmonary vessels have been proposed as indirect estimates of pulmonary volume5–20. However, none of these parameters has been shown to be accurate or specific5. Although magnetic resonance imaging (MRI) may overcome the limitations of sonographic evaluation for fetal pulmonary hypoplasia21–28, this technique is limited by its cost, poor patient compliance and artifacts related to fetal movements. Since 1996, three-dimensional (3D) ultrasound has been advocated as a valid alternative to estimate fetal pulmonary volume29–39. This new technique is better tolerated by patients than is MRI and has been shown to measure fetal organ volumes accurately, both in vivo and in vitro40–42. However, there are no ultrasound-based nomograms that can be used reliably to define a risk threshold for pulmonary hypoplasia and 3D ultrasound is currently limited by the fact that it is not yet available to all practitioners. The assessment of organ volumes can also be obtained from equations that use two-dimensional (2D) measurements. These values, calculated by 3D ultrasound, make it possible to improve the accuracy of 2D equations43.
The aim of this study was two-fold: to establish a nomogram of fetal pulmonary volume according to gestational age, and to define centiles for the accurate diagnosis of pulmonary hypoplasia.
The pulmonary volumes of 39 different fetuses were measured three times to compare the different techniques: once by a 2D method, once using 3D VOCAL™ mode (Virtual Organ Computer-aided AnaLysis, 3D View, version 2.0; General Electric Medical Systems, KretzTechnik, Zipf, Austria) and once using 3D multiplanar mode. All measurements were made by the same practitioner (D.M.). To establish the pulmonary growth nomogram, we took measurements from these fetuses, along with a further 583 fetuses, i.e. from a total of 622 singleton pregnancies. None of the fetuses had any detectable malformation and all had normal routine biometry parameters.
The pulmonary volumes were measured in 2D by 16 C.F.E.F. investigators from all over France. The ultrasound views corresponding to these data were all checked by one of the authors (C.T.). All practitioners were specialist physicians, performing fetal ultrasound scans as their sole medical practice. All were also members of the ‘College Français d'Echographie Foetale’ (C.F.E.F.), a medical society devoted to fetal sonology. The institutional review board of the C.F.E.F. approved the protocol.
The mean age of the mothers was 29 (range, 17–42; SD, 5) years. The exact gestational age of each pregnancy was confirmed by an early scan and pulmonary volumes were measured between 17 and 34 weeks' gestation. Each fetus was scanned just once for the study.
Pulmonary volume was calculated using the assumption that the lung is a geometrical pyramid (2D geometric pulmonary volume (2DGPV)) (Figure 1): total pulmonary volume (mL) = (surface area of right lung base (cm2) + surface area of left lung base (cm2)) × 1/3 height of right lung (cm). The surface area of the lung bases was measured on the transverse thoracic view containing the four chambers of the heart (Figure 2a). The height of the right lung was measured on a right sagittal paramedian view. It was measured between the upper aspect of the apex of the diaphragm and the pulmonary apex (Figure 2b).
Three-dimensional sonography and data processing
Scans were performed using a commercially available ultrasound system (Combison 530D, KretzTechnik) with ‘3D view’ software (KretzTechnik). Images were acquired using a 5-MHz electronic probe with an integrated mechanical drive and the volume was constructed by interpolating 250 elemental slices. The angle sweep was 60° and provided a maximum volume of 2500 cm3. The protocol used was that previously described by Lee et al.30 and Pöhls and Rempen32. According to Pöhls and Rempen32, the best image is obtained using contiguous transverse views of the lung. The initial plane of acquisition was that containing the four-chamber view of the heart. The upper and lower limits of the volume box were the clavicle and the apex of the diaphragm, respectively. The data obtained were recorded on a hard disc coupled to the ultrasound scanner. Volumes were then calculated using integrated software.
Pulmonary volume was measured in 3D using two different techniques: multiplanar mode and VOCAL mode.
For conventional multiplanar mode (3D) we used the protocol of Lee et al.30 (Figure 3). Three perpendicular views of the lung were displayed simultaneously on the screen. The lungs and/or the heart can be segmented on each view and their volume then calculated. Total thoracic volume was measured first. The thorax was segmented using a contour and integrated software calculated its volume. The upper and lower limits were the clavicle and the apex of the hemidiaphragm, respectively. The ribs and vertebrae were excluded. Fetal heart volume was then measured. The heart was contoured on consecutive transverse views and the measurements obtained were converted into a volume. The total lung volume was obtained by subtracting the heart volume from the total thoracic volume.
The multiplanar mode has been used in several previously published articles30–34, 39. It is quite rapid, with pulmonary volume being obtained in less than 5 min. One drawback, however, is that once the contour is drawn, it cannot subsequently be modified.
For rotational-based volume measurements, we used VOCAL software (Virtual Organ Computer-aided AnaLysis, 3D View, v 2.0; General Electric Medical Systems, KretzTechnik,). This is integrated within the ultrasound scanner. Using this rotational technique the volume of an ‘object’ is obtained by contouring its surface, as described by Kalache et al.35. We measured the volume of each lung separately (Figure 4) and the total lung volume was obtained by addition of these two values. Our plane of acquisition was the frontal view of the three perpendicular views already obtained for the multiplanar technique.
This plane was rotated around the z-axis until the lung apex was above and the diaphragm was below. The next step was to calculate the surface geometry of the lung by rotating the longitudinal plane around the vertical axis and defining 2D lung contours on each plane. The 2D contours can be defined automatically or manually, and the rotation step for each contour plane can be selected with an angle ranging from 6° to 30°. A rotation step of 30° was chosen arbitrarily. The upper and lower contour points were then positioned automatically at the level of the clavicle and mid-diaphragm, respectively, after the first manual trace. These two landmarks were visualized in each of six planes obtained by counterclockwise rotation of the lung via the vertical axis. Area tracing was carefully performed, excluding the fetal heart, mediastinal structures, ribs and the spine, until a rotation of 180° was completed. A 3D volume model of the lung was generated and reviewed for possible inconsistencies. Lung volume was calculated after all contours were considered to have been adequately traced. Each lung was measured separately as described by Pohls and Rempen32, Bahmaie et al.33 and Kalache et al.35, and took 5–10 min.
Clinical examples of pulmonary hypoplasia
We reviewed retrospectively the ultrasound scans performed since 1995 by two of the authors (T.C. and M.D.). We identified nine fetuses meeting the clinical criteria for pulmonary hypoplasia. In all cases, the presence of pulmonary hypoplasia was verified on autopsy. In order to test our equation for calculating 3D volumes, it was applied to these nine fetuses.
Kolmogorov–Smirnov D and Lilliefors tests for normality showed that the data were normally distributed. We used the intraclass correlation coefficient (ICC) to calculate reliability between the 3D techniques. Pearson's correlation and Student's t-test were used to compare the relationships and the mean differences, respectively. Pearson's correlation was also used to examine the relationship between volume measurements and gestational age in weeks. Linear regressions were used to predict equations to obtain the re-evaluated pulmonary volume equation (RPVE) from the 2D measurements. The Akaike information criterion (AIC) was used to choose the most suitable model44. This is a statistical procedure to choose the best model that neither underfits nor overfits the data.
The graph of pulmonary growth for the 622 fetuses was produced by a local polynomial regression fitting procedure (‘Loess’ function45) on SPSS software (SPSS for Windows; version 10.0).
Comparison between 3D multiplanar mode and VOCAL mode (Figure 5a)
The ICC between the multiplanar and VOCAL techniques was high: 0.94 (95% CI, 0.88–0.96; P < 2.2 × 10−16). The highest Pearson's correlation for the two techniques was found with gestational age (r: multiplanar = 0.92, VOCAL = 0.94, P < 2.2 × 10−16). There was a high correlation between the volumes obtained by the two techniques (r = 0.93, P < 2.2 × 10−16), and the means of the two sets of measurements were not significantly different (mean difference = 0.65, P = 0.51). In summary, there was no statistically significant difference between the values obtained using multiplanar and VOCAL techniques.
Kalache et al.35 found a comparable degree of agreement between multiplanar and VOCAL techniques but with less interobserver variability for the former. This is why we chose to compare multiplanar rather than VOCAL with a 2D technique later in the study.
Comparison between 2D and 3D multiplanar techniques (Figure 5b)
The ICC between 2D and multiplanar techniques was moderate: 0.52 (95% CI, 0.25–0.71; P < 0.0003). The highest Pearson's correlation coefficient for these two techniques was with gestational age (r: multiplanar = 0.92, 2D = 0.92, P < 2 × 10−6). The Pearson's correlation coefficient was very high when comparing measures obtained using 2D and 3D techniques (r = 0.92, P < 2.2 × 10−16). However, the means for the two sets of values were significantly different, with lower values for the 2D technique (mean difference = 11.99, P < 0.1 × 10−6).
In summary, the relationship between pulmonary volume and gestational age was similar whether the volume was measured using 2D or 3D multiplanar techniques. However, the volumes obtained were smaller when using 2D measurements. It can be seen from these results that it is possible to define an equation that will allow extrapolation of the 3D volumes from a 2D measurement.
Extrapolating 3D volumes from 2D measurements (Figure 5b)
The relationship between the 2D and 3D volumes was determined using a statistical linear regression method. Several regressions were used and one was chosen using the AIC44. Selecting a model using a criterion such as the AIC is similar to using hypothesis models in statistics. There are a number of models to test. The value of the criterion is measured for each of them and the model chosen is the one with the lowest value. One advantage of the AIC is that two independent models can be compared. We selected two of the models obtained using linear regression. One had a polynomial and the other a simple form. The AICs were 264.34 with 36 degrees of freedom, and 262.36 with 37 degrees of freedom. The best model was that with the lowest AIC for the highest degree of freedom, i.e. for a minimum of parameters to be estimated.
The equation selected was:
where RPVE is re-evaluated pulmonary volume equation, and 2DGPV is 2D geometric pulmonary volume = (surface area right lung base(cm2) + surface area left lung base (cm2)) × 1/3 height right lung (cm). The ICC was recalculated using this model. There was no statistically significant difference between the RPVE and the volumes obtained using the 3D multiplanar technique (ICC = 0.92, P < 0.1 × 10−16). We applied this equation to a sample of 622 patients to establish our pulmonary volume nomogram.
Establishment of a nomogram relating pulmonary volumes to gestational age
Fetal pulmonary volumes were measured in 622 patients using 2D ultrasound. The RPVE was obtained for each patient using the equation above. We used local polynomial regression fitting45 to establish the growth nomogram, generating a graph with local adjustment of the raw data according to gestational age; the local adjustment was according to neighboring points, weighted by the inverse of the distance between them. Values were obtained for the 1st, 3rd, 50th, 97th and 99th centiles (Tables 1 and 2). Graphs showing the median and centiles are given in Figure 6.
|Gestational age (weeks)||n||Raw data (mL)||Smoothing data (percentiles) (mL)|
|Gestational age (weeks)||n||Raw data (mL)||Smoothing data (percentiles) (mL)|
Clinical examples of pulmonary hypoplasia
Our equation for calculating 3D volumes was applied to the nine fetuses meeting the clinical criteria for pulmonary hypoplasia (Table 3); the values for pulmonary volume in these patients are represented by the stars in Figures 6a and b. All were below the first centile. A multicenter study to recruit a larger number of fetuses with pulmonary hypoplasia would be necessary to validate these results.
|Gestational age (weeks)||Right lung base SA (cm2)||Left lung base SA (cm2)||Right lung height (cm)||2DGPV (mL)||RPVE (mL)||Diagnosis|
|20||1.3||0.95||1.3||0.98||5.73||Polyhydramnios with lack of fetal movement|
|21||1.15||1.12||2.3||1.74||6.90||Right kidney agenesis, left kidney dysplasia|
|22||2.09||1||2||2.06||7.39||Bilateral kidney agenesis|
|24||1.2||1.06||1.9||1.43||6.43||Neonatal death with pulmonary hypoplasia|
|26||4.7||2.7||2.4||5.92||13.30||Bilateral kidney dysplasia|
|27||5||3.67||2||5.78||13.08||Polymalformation syndrome, 9 pericentric inversion|
|29||2.8||0||2.5||2.33||7.81||Left diaphragmatic hernia|
Kalache et al.35 demonstrated that both 3D multiplanar mode and 3D VOCAL mode can measure pulmonary volume reliably in fetuses. Our results concur with this view and we feel that these two techniques are interchangeable when measuring fetal pulmonary volume. The VOCAL mode seems to have advantages, however; it allows finer contouring of the lung and subsequent modification of the contour. This is of particular interest when the outline of the lung is irregular, such as in congenital diaphragmatic hernia. In practice, however, extrapolating pulmonary volume from a 2D scan is more convenient compared with using 3D, as it is quicker and easier43.
When studying the fetal lung using a 3D technique, the thoracic volume is best estimated by starting the data acquisition on a transverse view of the thorax32. Some authors30, 31 have obtained the pulmonary volume by subtracting the heart volume from the thoracic volume. The thoracic volume is obtained by adding together the areas of several slices of thorax (transverse views), excluding the ribs and vertebrae. The heart volume is measured by contouring the shape of the heart on the same slices. The mediastinal structures (thymus, trachea, esophagus and great vessels) are included in the measurement when using this method and this therefore tends to overestimate the pulmonary volume. Measuring the volume of each lung individually has been proposed as a way of reducing the error32–34. It is of note, however, that when these authors excluded the mediastinal volume from their measurements, they also failed to include the lowermost part of the lung bases, situated below the apex of the diaphragm (Figure 1).
In our study there was no demonstrable difference between the volumes obtained using either 3D technique, and it can be imagined that the small portion of the lung not included when measuring the lung height from the apex of the diaphragm to the clavicle is pretty much equivalent to the mediastinal volume included in the technique of Lee et al.30. The multiplanar technique allowed satisfactory estimation of pulmonary volume, with much shorter measurement and calculation times compared with using a frontal surfacing technique such as VOCAL mode. VOCAL mode may, however, be more accurate as the whole of the lung volume is included in the measurement and subsequent modifications of the initial contouring are possible.
Previous studies have shown that whatever the method used, the difficulty in obtaining satisfactory lung volumes increases with gestational age32, 39. Inter- and intraoperator variability is higher after 30 weeks' gestation33. This is due to the greater difficulty in accurately contouring the lung after this age39. In practice, however, the detection of pulmonary hypoplasia and decision making are most important before 32 weeks and this limitation in methodology has few clinical implications.
Predicting 3D multiplanar volumes from a 2D scan
We did not evaluate interobserver variability using a 2D technique in this study. Chang et al.43 have shown that when measuring cardiac volume the reproducibility of volume measurement is lower using 2D compared with a 3D technique based on the addition of surfaces. However, because of the limitations of 3D ultrasound (not routinely available, more expensive and time-consuming), they concluded that 2D ultrasound should be the method of choice in clinical practice. The lack of widespread availability of 3D scanners justifies the approach taken by Chang et al. in trying to define a ‘correction coefficient’ to obtain a 3D volume from a 2D technique43. We used the same procedure, i.e. a pyramid as the geometric model for the lung.
The transverse thoracic view that includes the four chambers of the heart is quite easily obtained and is therefore fairly reproducible, which is why we chose this plane as our reference plane for the lung bases. Roberts and Mitchell16 showed that there is no significant difference in height between the two lungs and, because it is technically easier, we chose to measure the height of the right lung.
We found 11 studies in the literature that attempted to establish growth nomograms for fetal pulmonary volume. Three used MRI24, 25, 28 and eight used ultrasound29–34, 36, 39. The number of fetuses per gestational week was small (sometimes very small), much lower than in our study. The appropriate statistical parameters were not always calculated for each gestational week and the medians published varied widely, whether at 22 or at 32 weeks. Moreover, various regression models were used. Our results at 22 weeks are compatible with those of Duncan et al.24, Pöhls and Rempen32, Osada et al.34 and Sabogal et al.39. Our results at 32 weeks are more in keeping with those of Bahmaie et al.33.
In the largest study, conducted by Rypens28 (n = 215), the confidence intervals at 22 and 32 weeks' gestation were much wider than were those in our study: 13–47.5 vs. 12.97–13.6 at 22 weeks and 40.3–114.62 vs. 37.1–41.2 at 32 weeks, respectively. Comparing the medians between the two studies shows that the values obtained by MRI are 1.83 times higher compared with those obtained by ultrasound, both at 22 and 32 weeks.
MRI values are consistently higher than are ultrasound values in most studies, although this difference was not found in the most recent MRI article46. It would be interesting to design a further study to validate and understand better this significant difference between the two techniques.
The fetuses at risk of pulmonary hypoplasia presented in Table 3 and in Figure 6 all had pulmonary volumes below the 1st percentile on our nomogram. It is obvious that our technique can only be applied to situations in which the lung shape is close to that of a pyramid and is not much altered (e.g. in oligohydramnios secondary to premature rupture of membranes or renal disease or in dwarfism, where the thoracic volume is reduced). A more accurate 3D technique with the possibility of modifying lung contours (VOCAL mode) is mandatory when the lung shape is altered significantly (e.g. in congenital diaphragmatic hernia, primary intrathoracic disease with compression of lung tissue)37, 38.
In conclusion, we found no statistically significant difference between lung volume measurements obtained using the two 3D modes, although in comparison with the 2D measurements the volumes obtained were larger. We have produced an equation allowing calculation of 3D volumes from 2D measurements, and constructed two nomograms, one for use with 2D and one for 3D technology. Further epidemiological studies are being performed to determine the sensitivity and specificity of our method in assessing pulmonary hypoplasia.
We thank Doreen Raine for her linguistic review of the manuscript.
- 15The thorax. In Diagnostic Ultrasound of Fetal Anomalies Text and Atlas, NybergDA, MahonyBS, PretoriusDH (eds). Year Book Medical Publishers, Inc.: Chicago, 1990; 262–299., .
- 16Direct ultrasonographic measurement of fetal lung length in normal pregnancies complicated by prolonged rupture of membranes. Am J Obstet Gynecol 1990; 160: 1560–1566., .
- 45Local regression models. In Statistical Models, ChambersJM, HastieTJ (eds). Cole Advanced Books and Software: Pacific Grove, CA, 1992; Chapter 8., , .