Reproducibility of three-dimensional ultrasound diagnosis of congenital uterine anomalies




To examine the reproducibility of the diagnosis of congenital uterine anomalies and the repeatability of measurements of uterine cavity dimensions using three-dimensional (3D) ultrasound.


The reproducibility of diagnosis of congenital uterine anomalies was examined by re-slicing stored 3D ultrasound volumes. Each data set was evaluated by two independent observers. Assessment of uterine morphology was performed in a standardized plane, with the interstitial portions of the Fallopian tubes used as reference points. Additionally, in 35 cases of congenital uterine anomalies the width of the uterine cavity (W), fundal distortion (F) and the length of unaffected uterine cavity (C) were measured. Intraobserver and interobserver variabilities were evaluated by each observer performing all three measurements twice.


Eighty-three 3D ultrasound volumes were examined. Both operators classified 27 uteri as normal, 33 as arcuate, 19 as subseptate and three as unicornuate. A single case of uterine anomaly was described as arcuate uterus by one operator and subseptate by another (kappa 0.97). The intraobserver variability for each of the three measurements (W, F and C) was satisfactory with limits of agreement ranging from ±1.43 to ±2.51 mm. The examination of the interobserver variability showed no significant differences between the two observers (F = 0.484, P > 0.05).


3D ultrasound is a reproducible method for the diagnosis of congenital uterine anomalies and for the measurement of uterine cavity dimensions. Copyright © 2003 ISUOG. Published by John Wiley & Sons, Ltd.


Congenital uterine anomalies are associated with adverse reproductive outcomes1–4. As a result, screening for uterine anomalies forms a part of routine clinical investigations of women with history of infertility, recurrent miscarriage and early preterm labor5. The conventional methods for the assessment of uterine morphology are hysterosalpingography, hysteroscopy and laparoscopy6. All these tests are invasive and the diagnosis of congenital uterine anomaly is based on the subjective impression of the operator performing the test. There is no agreement on the criteria that should be used to differentiate between various types of uterine anomalies, and none of these methods has ever been investigated for their reproducibility.

Three-dimensional (3D) ultrasound has recently become available in gynecological practice. It is a non-invasive, outpatient diagnostic modality, which enables a detailed assessment of uterine morphology7. A high level of agreement between 3D ultrasound, hysterosalpingography and laparoscopy in the classification of uterine morphology has been reported previously8, 9. In this study we have examined the interobserver and intraobserver variability of 3D-ultrasound diagnosis of congenital uterine anomalies. We have also assessed the reproducibility of the measurements of uterine cavity dimensions, which were obtained using the technique of planar reformatted sections.


All women referred to our Unit for a gynecological ultrasound examination between January 1998 and May 2002 were screened for the presence of congenital uterine anomalies. Whenever, a uterine anomaly was suspected on conventional two-dimensional ultrasound examination, a 3D transvaginal ultrasound scan was performed (Voluson 530 and Voluson 730, KretzTechnik, Zipf, Austria) in order to establish the final diagnosis. The technique of 3D ultrasound has been described previously10. Briefly, the uterus was visualized in the longitudinal plane and a 3D volume was generated by the automatic sweep of the mechanical transducer. The acquired volumes were in the shape of a truncated cone with the depth of 4.3–8.6 cm and a vertical angle α = 90°. The volumes were then analyzed online using the technique of planar reformatted sections. With this technique it was always possible to obtain the coronal view of the uterus, which is usually lying perpendicular to the ultrasound beam. The analysis of uterine morphology was performed in a standardized reformatted plane, with the uterus visualized in the coronal plane using the interstitial portions of the Fallopian tubes as reference points (Figure 1).

Figure 1.

A standardized view of a subseptate uterus in which the interstitial portions of both Fallopian tube are clearly visualized. Measurements of the uterine cavity width (W), fundal distortion (F) and the length of the unaffected uterine cavity (C) are illustrated.

In all cases with confirmed diagnosis of congenital uterine abnormalities 3D ultrasound volumes were permanently stored on removable hard disk cartridges (Magneto-Optic 3.0′, 640MB Olympus Europe, Hamburg, Germany). Of a total of 239 stored cases, 83 were selected for the examination of reproducibility of 3D-ultrasound diagnosis by a senior investigator (D.J.). The volumes were selected in such a way as to ensure a good case mix including normal uteri, and minor and major uterine anomalies. The volumes were then examined independently by two experienced operators (R.S. and B.W.), who described uterine morphology in accordance with the modified American Fertility Society Classification (Table 1)11.

Table 1. Criteria for the classification of congenital uterine anomalies
Uterine morphologyFundal contourExternal contour
NormalStraight or convexUniformly convex or with indentation < 10 mm
ArcuateConcave fundal indentation with central point of indentation at obtuse angle (> 90°)Uniformly convex or with indentation < 10 mm
SubseptatePresence of septum, which does not extend to cervix, with central point of septum at an acute angle (< 90°)Uniformly convex or with indentation < 10 mm
SeptatePresence of uterine septum that completely divides cavity from fundus to cervixUniformly convex or with indentation < 10 mm
BicornuateTwo well-formed uterine cornuaFundal indentation > 10 mm dividing the two cornua
Unicornuate with or without rudimentary hornSingle well-formed uterine cavity with a single interstitial portion of Fallopian tube and concave fundal contourFundal indentation > 10 mm dividing the two cornua if rudimentary horn present

In addition, in 35 cases of congenital uterine anomalies the reproducibility of measurements of uterine cavity dimensions was also examined. This sample would detect as significant a discrepancy in the mean difference of one standard deviation (1 SD) with a statistical power of 85% at the 0.05 significance level. Three measurements were taken in each case: the uterine cavity width (W), which was measured as the distance between the two internal tubal ostia; the depth of fundal indentation or the septum length (F), which was defined as the distance between the midpoint of the line adjoining the tubal ostia and the distal tip of fundal indentation or uterine septum; and the length of the unaffected uterine cavity (C), which was measured from the distal tip of the fundal indentation or uterine septum to the level of the internal os (Figure 1). The internal os was identified in the longitudinal section of the uterus using the point of reflection of urinary bladder as a reference point. Each of the two operators obtained two blinded readings of uterine cavity dimensions, which were recorded by an independent observer.

The interobserver agreement for the diagnosis of each uterine anomaly was assessed by calculating the kappa statistic. Intraobserver variation of the measurements of uterine cavity dimensions was analyzed by calculating standard deviations of differences between the 35 pairs of measurements of W, F and C taken by each observer. The differences were then plotted against the mean of the two readings on a scatter diagram12. Analysis of variance was used to assess the interobserver variation of the measurements of uterine cavity dimensions.


Reproducibility of diagnosis of congenital uterine anomalies

There was complete agreement between two operators in classifying the uteri as being normal or abnormal. The observers agreed on the diagnosis of arcuate uterus in 33 cases, subseptate uterus in 19 cases and unicornuate uterus in three cases. There was disagreement in a single case, which was classified as an arcuate uterus by Operator A and as a subseptate uterus by Operator B (kappa 0.97, 95% CI 0.94–1.0) (Table 2).

Table 2. Interobserver reproducibility of diagnosis of congenital uterine anomalies
Operator AOperator B
Normal27 0 0 027
Arcuate 033 0 033
Subseptate 0 119 020
Unicornuate 0 0 0 3 3
Total273419 383

Repeatability of measurements of uterine cavity dimensions

There was no relationship between mean measurement of any of the uterine dimensions and either intraobserver or interobserver difference. Inspection of scattergrams also showed that > 95% of differences were within mean difference ± 2 SDs, thus confirming their normal distribution. Limits of agreement for repeated readings ranged between ± 1.696 and ± 2.514 mm for Operator A, and between ± 1.762 and ± 2.07 mm for Operator B (Table 3).

Table 3. Intraobserver variability of measurements of uterine dimensions
MeasurementOperator AOperator B
Mean difference (95% CI)Limits of agreement (95% CI)Mean difference (95% CI)Limits of agreement (95% CI)
Uterine cavity width (W)− 0.186 (− 0.53 to 0.161)− 2.204 (− 2.804 to − 1.604)1.832 (1.232 to 2.232)0.406 (0.103 to 0.709)− 1.356 (− 1.88 to − 0.832)2.168 (1.644 to 2.692)
Depth of fundal indentation (F)− 0.149 (− 0.44 to 0.14)− 1.845 (− 2.35 to − 1.341)1.547 (1.043 to 2.051)− 0.171 (− 0.418 to 0.074)− 1.603 (− 2.03 to − 1.178)1.26 (0.835 to 1.685)
Uterine cavity length (C)− 0.171 (− 0.60 to 0.23)− 2.684 (− 3.432 to − 1.936)2.344 (1.596 to 3.092)0.2 (− 0.156 to 0.556)− 1.87 (− 2.48 to − 1.252)2.266 (1.652 to 2.88)

The examination of the interobserver variability showed no significant differences between the two observers (P > 0.05) (Table 4). Readings by Observer A were on average 0.149 to 0.33 mm higher than measurements taken by Observer B (Figures 2–4). However, this small bias is unlikely to be of clinical significance.

Figure 2.

Interobserver variability of uterine width measurement (W) (mean difference − 0.33 mm (95% CI − 0.642 to − 0.02), 95% limits of agreement − 2.145 mm (95% CI − 2.685 to − 1.603) to 1.483 mm (95% CI 0.942 to 2.024)).

Figure 3.

Interobserver variability of fundal depth/septum length measurement (F) (mean difference − 0.195 mm (95% CI − 0.441 to 0.051), 95% limits of agreement − 1.63 mm (95% CI − 2.062 to − 1.204) to 1.24 mm (95% CI 0.813 to 1.671)).

Figure 4.

Interobserver variability of uterine cavity length measurement (C) (mean difference − 0.149 mm (95% CI − 0.411 to 0.113), 95% limits of agreement − 1.677 mm (95% CI − 2.124 to − 1.23) to 1.379 mm (95% CI 0.932 to 1.826)).

Table 4. Analysis of variance of interobserver mean difference of intrauterine dimensions
Source of variationdfSums of squaresMean squaresFP
  1. df, degrees of freedom.

Measurements 3420.980.620.9440.562
Observers  20.630.320.4840.618
Residual 6844.40.65  


This study showed a good level of agreement between two different operators in the diagnosis of congenital uterine anomalies using 3D ultrasound. The description of congenital uterine anomalies was based on the American Fertility Society Classification11. However, the diagnostic criteria used to classify uterine anomalies were more detailed than previously described and they included cut-off levels for the fundal shape and distortion. Although these cut-offs were selected arbitrarily, they were necessary to differentiate between uterine anomalies with similar morphological features such as subseptate and arcuate uteri. Furthermore, without clearly defined diagnostic criteria it would not have been possible to examine the reproducibility of ultrasound diagnosis.

The American Fertility Society Classification11, which is used by most clinicians in routine practice, does not specify either morphological features or diagnostic methods to describe uterine morphology. It is therefore not surprising that to date no study has been published investigating the reproducibility of diagnosis of uterine anomalies using any of the traditional invasive diagnostic methods. Another likely consequence of the lack of agreed diagnostic standards is a significant interobserver variability in the diagnosis of uterine anomalies. Operators with varying levels of experience, using different diagnostic methods, are likely to disagree on both the type and severity of uterine anomaly in individual cases. An indirect proof of this is a wide variety of reported prevalence of uterine anomalies in different studies of both low-risk women and those with a history of adverse pregnancy outcomes13–16. Similar problems have also affected 3D ultrasound diagnosis of uterine anomalies, with different studies reporting very different prevalence rates of uterine anomalies in well-defined groups of women, such as those with history of infertility1, 17.

3D ultrasound provides a very different view of uterine anatomy, which further hampers comparisons with other diagnostic modalities. For example, arcuate uterus is the most common uterine anomaly in studies of uterine morphology using 3D ultrasound7. However, this anomaly is very difficult to diagnose on traditional invasive tests, and as a result there is only a handful of reports in the literature describing the diagnosis and clinical significance of this anomaly1, 4.

The only way to overcome these limitations is by quantifying the degree of disruption of uterine morphology and comparing the results to reproductive outcomes. Findings of good intraobserver and interobserver reproducibilities of measurements of uterine dimensions with 3D ultrasound are very encouraging in this context. Quantitative description of uterine morphology may show that not only the type of uterine anomaly, but also its severity, may be important in assessing the risk of adverse pregnancy outcomes. This could help to refine selection criteria for surgery, resulting in improved long-term outcomes in women with uterine anomalies.

In order to achieve this objective, a relatively large quantity of data needs to be collected. Bearing in mind the rarity of uterine anomalies, this can only be achieved by establishing a wide international collaboration involving a large number of research centers. 3D ultrasound is the only diagnostic method available to facilitate such a collaboration. Quantitative description of uterine morphology would ensure that the same diagnostic criteria were used in each case. Furthermore, by asking the participating centers to submit archived 3D volumes, both operator dependence and subjective bias in the assessment of uterine morphology would be almost completely eliminated. Until such a study has been conducted, we will have to face continued uncertainty about the clinical significance and optimal management of congenital uterine anomalies.