SEARCH

SEARCH BY CITATION

Keywords:

  • first-trimester screening;
  • intracranial translucency;
  • semi-automated measurement;
  • spina bifida

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. References

Objectives

To assess the reproducibility of fetal intracranial translucency (IT) measurements performed manually or with SonoNT®, a semi-automated caliper placement technique recently introduced for nuchal translucency thickness (NT) measurement.

Methods

This was a retrospective study using 116 stored images of the head (mid-sagittal plane) from normal fetuses in dorsoposterior position at 11–13 weeks. Two experienced operators each measured the IT separately, twice manually and twice using the semi-automated software. Intraoperator and interoperator repeatability were assessed. The mean of the two manual measurements of the more experienced Operator 2 was considered as the ‘gold standard’.

Results

Seven cases were excluded as the IT could not be recognized by the semi-automated software. In the remaining 109 cases, the interquartile range of the mean IT measurement was 1.9–2.4 mm for Operator 1 and 1.8–2.3 mm for Operator 2 for both the manual and the semi-automated IT measurements. The intraoperator SD for manual measurements was 0.091 mm for Operator 1 and 0.088 mm for Operator 2, and for semi-automated measurements it was 0.054 mm for Operator 1 and 0.067 mm for Operator 2. Concerning interoperator bias of the manual measurements, the mean difference between Operator 1 and Operator 2 was − 0.09 (95% CI, − 0.11 to − 0.07) mm. With respect to the gold standard, the mean bias of the semi-automated measurements was 0.01 (95% CI − 0.01 to 0.03) mm for Operator 1 and − 0.09 (95% CI − 0.11 to − 0.07) mm for Operator 2, indicating good agreement.

Conclusions

Manual IT measurements are reproducible. In addition, IT can be assessed reliably using the semi-automated NT algorithm, leading to standardization of the IT assessment process. Copyright © 2012 ISUOG. Published by John Wiley & Sons, Ltd.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. References

Ultrasound examination at 11–13 weeks has become a screening tool not only for the measurement of nuchal translucency thickness (NT), but also for the detection of severe fetal malformations. In order to facilitate detection of open spina bifida, we recently introduced measurement of the intracranial translucency (IT), which can be visualized in the same mid-sagittal plane as that used for assessment of NT and the nasal bone1–3. The IT represents the anteroposterior diameter of the developing fourth ventricle and can be identified reliably as an anechoic region with two horizontal echogenic borders. The anterior border of the IT represents the posterior border of the brainstem, and the posterior border represents the choroid plexus of the fourth ventricle. The increasingly widespread use of IT measurement is an indication of the simplicity of this sign, probably due to its similarity to NT, appearing as two parallel echogenic lines with fluid in between. Reports on semi-automated NT measurement4–7 have shown that automated placement of calipers may in future be of increasing interest in fetal ultrasound, since this can contribute to a reduction in the intra- and interoperator variability of measurements in both normal and abnormal cases4, 8, 9. Recently, it was shown that semi-automated NT software can also be applied to other structures, for example to measure umbilical vein diameter10. It is therefore possible that the semi-automated NT software could also be of interest in the assessment of IT.

The aim of this study was to examine whether at 11–13 weeks of gestation the assessment of IT can be further standardized by using the semi-automated NT algorithm, SonoNT®, in comparison to manual measurement.

Patients and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. References

Singleton pregnancies undergoing routine screening at 11 + 0 to 13 + 6 weeks and showing NT within the normal range, as well as normal fetal anatomy and normal appearance of the posterior brain at the anomaly scan, were included in this study.

As a standard requirement of our institution, all patients provided signed informed consent for fetal examination and agreed to the storage of digital images and measurement data for anonymous quality control and later data evaluation. During the first-trimester scan, images of the mid-sagittal view of the face were obtained with optimal visualization of the posterior brain1, including the thalamus, midbrain, brainstem, myelencephalon, fourth ventricle (measured as IT), choroid plexus of the fourth ventricle, fluid in the future cisterna magna and occipital bone. Images were stored on the hard drive of the ultrasound equipment for later assessment.

After findings at the second-trimester ultrasound examination were established to be normal, 116 first-trimester images were selected arbitrarily for retrospective IT diameter measurement. All were from examinations performed transabdominally, with the fetus in a dorsoposterior position, using a commercially available Voluson E8 (GE Medical Systems, Zipf, Austria) ultrasound system equipped with a high-resolution RM-6C transducer. IT measurement was performed both manually and semi-automatically with the SonoNT software, which is integrated in the ultrasound machine. Care was taken to ensure that only one image per patient was selected, and that no image showed any previous measurements.

The principle of semi-automated measurement using SonoNT has been explained in previous studies4, 7 and is illustrated in Figure 1. A region of interest (ROI) box or frame is placed by the operator over the fluid-filled region, which has to be well delineated by two horizontal echogenic borders. The software uses the original image within this frame and produces a corresponding ‘edge image’ that reflects the differences in brightness rather than the brightness itself. These images are used together to define the echogenic lines delineating the translucency that should be used for its measurement. According to the presets chosen, the upper and lower calipers were placed automatically on the inner borders (inner–inner measurement). The measurement algorithm connects every point on one of the two echogenic lines to all points on the other line7. For each point on the first line, it then eliminates all except the shortest connection to the second line, and from these it selects the longest as the final translucency measurement. The measurement obtained is shown on the screen as a red line within the ROI. Since it is a semi-automated measurement, the examiner can confirm the result or can change the lateral and vertical lengths of the frame to redefine the (largest) translucency thickness.

thumbnail image

Figure 1. Mid-sagittal plane of the fetal face showing use of the region-of-interest frame for the semi-automated assessment (SonoNT®) of the intracranial translucency at its largest diameter.

Download figure to PowerPoint

Two operators (K.K., R.C.) participated in this study. Both are specialists in fetal medicine and have obtained the NT certification of The Fetal Medicine Foundation, London. Both are experienced in evaluation of the posterior brain throughout gestation and in semi-automated NT measurement at 11–13 weeks. Before commencing this study, they each examined 15 images that were not evaluated further, in order to determine whether the semi-automated NT algorithm (SonoNT) could be applied to recognize the IT edges and deliver acceptable measurements. It was found that the IT has a more complex shape than does the NT and that the neighboring structures can lead to incorrect recognition of borders. It was agreed to place the ROI frame in the core ROI, with the vertical borders close to the echogenic edges of the brainstem and choroid plexus and the lateral borders rather narrow over the IT region, but always including the echogenic anterior and posterior borders (Figure 1). The IT frame had a rather square shape when compared to the rectangular shape described previously for semi-automated NT measurement4, 7. The size of the frame was therefore smaller in comparison to that used for semi-automated NT measurement.

The operators performed IT measurements separately on different days and were unaware of each other's results. As in the previous studies of Abele et al.4 and Moratella et al.7, numeric displays on the screen were covered so that the operators were blinded to the actual measurements and were unaware of previous results obtained by themselves. IT was first measured semi-automatically using the SonoNT software. The operator was allowed to change the size of the frame before confirming the measurement and storing it. A second semi-automated measurement was then performed immediately and also stored. If after several attempts it was not possible to obtain two ROI frames with identification of the IT borders, the case was excluded from the study. Once semi-automated measurements were completed, the examiner started the manual measurements, blinded to the results of the semi-automated measurements.

Manual IT measurement was performed in a way similar to NT measurement, by magnifying the image and placing the cursor on the inner border of the echogenic posterior brainstem border (upper caliper) and on the inner border of the choroid plexus of the fourth ventricle (Figure 2). The longest distance in the middle portion of the IT was chosen for the measurement. A second manual measurement was then performed immediately and also stored.

thumbnail image

Figure 2. Mid-sagittal plane of the face of the same fetus as in Figure 1, showing placement of calipers for manual measurement of intracranial translucency diameter.

Download figure to PowerPoint

Images with measurements were exported and the four values of each operator were entered into a database. For statistical analysis only cases with a complete set of measurements from both examiners were considered.

Statistical analysis

For each operator, the intraoperator repeatability of the two manual and the two semi-automated measurements was assessed by intraoperator SD, calculated as the SD of the differences between the two measurements divided by √2. In addition, the intraclass correlation coefficient (ICC) with 95% CI was used to assess reliability. For assessment of interoperator and intermethod reliability, the mean of the two manual measurements of Operator 2 (R.C.) were considered as the gold standard due to his experience in the assessment of NT and IT (> 1500 cases). Manual measurements of Operator 1 (K.K. > 500 cases) and semi-automated measurements of both operators were compared with this gold standard by calculating the mean difference between the mean of each pair of measurements, with 95% CIs.

In addition, 95% limits of agreement according to the method of Bland and Altman11 were computed to assess the intra- and interoperator repeatability.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. References

IT was assessed in a total of 116 cases, seven (6.0%) of which were excluded from further evaluation because one or both operators were unable to obtain appropriate semi-automated measurements. Interestingly, all excluded cases were within the first 30 patients evaluated, which may reflect the learning curve of IT measurement using the SonoNT software. Thus, in 109 (94.0%) cases, both operators were able to obtain all required measurements and these were included in the final analysis.

The median crown–rump length at the time of NT assessment was 70.0 (interquartile range, 65.1–73.4) mm. The median manual and semi-automated IT measurements were 2.10 (1.85–2.35) mm and 2.15 (1.90–2.40) mm, respectively. Table 1 shows the individual distribution of IT measurements of the two operators, indicating good overall agreement between measurements, with a trend towards larger measurements using the semi-automated approach.

Table 1. Distribution of manual and semi-automated intracranial translucency measurements
OperatorMode of measurementIntracranial translucency (mm, median (IQR))
  1. IQR, interquartile range.

1Semi-automated2.15 (1.90–2.40)
1Manual2.05 (1.90–2.40)
2Semi-automated2.15 (1.80–2.20)
2Manual2.10 (1.80–2.30)

Intraoperator repeatability

The intraoperator SD for manual measurements was 0.091 mm for Operator 1 and 0.088 mm for Operator 2 and for semi-automated measurements it was 0.054 mm for Operator 1 and 0.067 mm for Operator 2. For both operators, ICCs were significantly higher for semi-automated compared with manual IT measurements: 0.982 (95% CI, 0.974–0.988) vs 0.940 (95% CI, 0.914–0.959) for Operator 1 and 0.971 (95% CI, 0.959–0.980) vs 0.939 (95% CI, 0.912–0.957) for Operator 2. The 95% limits of agreement were − 0.26 and 0.23 mm for manual IT measurements and − 0.17 and 0.17 mm for semi-automated IT measurements.

Interoperator and intermethod repeatability

The mean difference between the manual measurements of Operator 1 and Operator 2, the latter being considered as the gold standard, was − 0.09 (95% CI, − 0.11 to − 0.07) mm. The mean difference between the semi-automated measurements for Operator 1 with respect to the gold standard was 0.01 (95% CI, − 0.01 to 0.03) mm and for Operator 2 it was − 0.09 (95% CI, − 0.11 to − 0.07) mm, indicating good agreement between the semi-automated measurements and the gold standard.

The 95% limits of agreement for the manual measurements of Operator 1 versus the gold standard were − 0.34 and 0.17 mm, and for the semi-automated measurements of both operators versus the gold standard they were − 0.20 and 0.23 mm.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. References

This study shows that IT measurements are reproducible and can be further standardized by using automated caliper placement algorithms (SonoNT). The intraoperator repeatability in particular can be increased, while in this study the interoperator repeatability was similar between manual and semi-automated IT measurements. This may reflect the fact that both operators are experienced in the assessment of IT. It has been shown that less experienced operators may benefit most from automated caliper placement algorithms, with the repeatability of their measurements being increased to the level of experts4.

High intra- and interoperator repeatability is a prerequisite for the widespread use of IT as a screening tool for open spina bifida at 11–13 weeks. Our study shows that SonoNT, as one of the commercially available softwares for semi-automated measurement of NT, can be used reliably for the assessment of IT diameter. We are confident that other algorithms developed for NT measurement6, 10, 12 can also be applied for IT measurement, since both IT borders are as easily identifiable as are the NT borders.

This study of semi-automated IT measurement confirms the similar results of semi-automated NT measurement. Abele et al.4 showed that semi-automation of NT measurement reduces substantially the intra- and interoperator variability. They concluded that less experienced operators in particular would benefit from this tool as differences between experts and non-experts disappeared. Especially for cases with moderately increased NT, in which small differences between two measurements may change the calculated risk dramatically, automated caliper placement could help to standardize the measurement technique and to reduce operator dependency9. Moratalla et al.7 investigated whether SonoNT may also improve caliper placement accuracy among expert sonographers and found that the interoperator SD was about seven times smaller with the automated caliper placement system and the intraoperator SD was more than halved.

One of the major limitations of semi-automated IT measurement compared with NT measurement is the slight differences in shape at different gestational ages and individual variations in the IT in different fetuses. Whereas NT is large in the middle portion, we observed that the IT can be curved or cone-shaped and occasionally cranial or caudal borders can be obscured by shadowing from neighboring structures. Therefore, in the preparation for the study, the examiners suggested placing the ROI frame over the middle portion, which is easiest to visualize, and not over the whole fourth ventricle because its upper and lower borders are not always clearly identifiable. Despite the fact that the IT values obtained by SonoNT showed a surprisingly reliable intra- and interoperator reproducibility, seven of 116 (6%) cases had to be excluded as reliable recognition of the IT was not possible. It should, however, be borne in mind that the algorithm of SonoNT and other similar automated NT systems were not developed for IT assessment.

Results of NT measurements are translated into patient-specific risks and small differences of even 0.1 mm can influence the decision-making process. For IT, the preliminary experience has concentrated on qualitative assessment rather than on a quantitative measurement of the amount of fluid and, so far, measurements have not been translated into patient-specific risks. Since the first introduction of IT assessment, we have observed that in some cases of spina bifida there may be some fluid in the posterior brain region1, 13. However, in our experience, in these cases either the posterior border was not identifiable or the IT had a reduced diameter, which emphasizes the need for a standardized quantitative assessment of IT (Figure 1). Interestingly, an apparently normal IT, even in retrospective studies, has been shown to exclude the presence of open spina bifida14.

The diameter of the brainstem is smaller than is the distance from the brainstem to the occipital bone in normal fetuses, and assessment of this feature can provide an additional clue for the detection of open spina bifida at 11–13 weeks of gestation15. Especially in those few cases with open spina bifida and presence of some fluid in the posterior brain, the brainstem appeared thicker and the distance to the occipital bone shorter, independently of whether fluid was still seen in the fourth ventricle1. Automated measurement of these two distances may help in such cases and we are confident that the increased use of semi-automated measurement of NT will push ultrasound companies to develop an algorithm for the semi-automated assessment of the posterior brain including brainstem, IT and brainstem–occipital bone distance.

In summary, we have shown that IT measurement can be performed with high intra- and interoperator reproducibility when performed manually. Further standardization can be achieved by using semi-automated caliper placement algorithms. Due to the similarity to NT shape, a commercially available semi-automated system developed for NT measurement can also be used reliably for IT assessment. An interesting approach for the future would be to encourage the development of an algorithm for semi-automated measurements within the posterior brain region.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and methods
  5. Results
  6. Discussion
  7. References