SEARCH

SEARCH BY CITATION

Keywords:

  • 3D ultrasound;
  • automated measurement;
  • follicle;
  • ovary;
  • SonoAVC;
  • VOCAL;
  • volume calculation

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Objectives

To assess the ability of the new software SonoAVC to measure follicular volume and to compare these volume calculations with those made by conventional methods.

Methods

Three-dimensional ultrasound imaging was used to acquire volumetric data from the ovaries of 51 women undergoing controlled ovarian stimulation as part of in-vitro fertilization treatment. All assessments were performed on the day of oocyte retrieval and the true volume of each follicle was ascertained by manual measurement of the follicular aspirate. SonoAVC was used to automatically measure the volume of follicles and to provide three perpendicular diameters (xyz diameters), which were used to estimate volume using the sphere formula. The sphere formula was also used to estimate the volume from manual measurements of follicle diameter derived from conventional two-dimensional (2D) displays. Virtual Organ Computer-aided AnaLysis (VOCAL) was also used to measure volume, and the validity of each technique was compared using limits of agreement.

Results

Two hundred and twenty-four follicles with a mean follicular volume of 3.7 (range, 0.4–16.2) cm3 were studied. SonoAVC provided highly accurate automatic follicular volume measurements in all cases. Volume estimations made from the automatic maximal follicular diameter measurements (xyz diameters) were less valid. VOCAL proved highly valid but was less accurate than SonoAVC. Volumes estimated from manually derived follicular diameter measurements were the least accurate.

Conclusions

SonoAVC provides highly valid, automatic measurements of follicular volume. These measurements are more accurate than volumes estimated from 2D manual measurements, automated measurements of follicular diameter and those calculated using VOCAL. Copyright © 2008 ISUOG. Published by John Wiley & Sons, Ltd.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Ultrasound imaging is used on a daily basis to identify pathology and confirm normality. This requires a subjective interpretation of the image display, which can be modified according to the object or area of interest. Objective assessment of an ultrasound image requires some form of measurement to be made, which should be performable in a reproducible manner and provide a valid result. Ultrasound findings are open to interpretation, therefore, and dependent on the observer regardless of whether a subjective or objective examination is being made. Objective examination at least allows assessments by different observers (interobserver) and by the same observer (intraobserver) to be analyzed mathematically. This can be used to define standards and to audit practice. Automatic data analysis has the potential to remove any observer bias and to reduce the time needed for measurements, but must be both valid and reliable.

In assisted reproduction treatment ovarian follicular volume is considered to be an important indicator of oocyte maturity. The outcome of conventional in-vitro fertilization (IVF) treatment is best when oocytes are obtained from follicles measuring between 1 and 7 mL, which corresponds to a diameter of 12 to 24 mm if the follicles assume the shape of a sphere1. However, follicles rarely have a spherical shape and most are elliptical owing to follicular overcrowding in a hyperstimulated ovary2. Although there is a high degree of correlation between the mean diameter and volume of a follicle, follicular volume calculated from the mean diameter has been shown to be less accurate, especially when the follicle is ellipsoid2. Three-dimensional (3D) calculation of follicular volume appears to be more accurate but requires more extensive measurements to be made and is therefore more time consuming3, 4.

SonoAVC (Automatic Volume Calculation; GE Medical Systems, Zipf, Austria) is a new software program designed to provide automatic volume calculations of fluid-filled areas. It is either incorporated into the ultrasound machine (Voluson E8, GE Medical Systems) or installed on a personal computer for offline analysis of the datasets acquired by any ultrasound machine of the same manufacturer. SonoAVC identifies and quantifies hypoechogenic regions within a 3D dataset and provides automatic estimation of their absolute dimensions, mean diameter and volume5. Because each different volume is color coded separately, SonoAVC is an ideal tool for studying follicular development within the ovary. Before the reliability of the software can be tested in prospective studies it is essential to show that it is valid and provides accurate measurements. This study was designed to assess the ability of this new software to measure follicular volume and to compare these volume calculations with those made by conventional methods.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

3D ultrasound imaging was used to acquire volumetric data from the ovaries of 51 women undergoing controlled ovarian stimulation as part of IVF treatment. All assessments were performed on the day of oocyte retrieval using a Voluson E8 ultrasound machine and a four-dimensional 5–9-MHz transvaginal probe. A probe program with default settings adjusted to provide the best two-dimensional (2D) gray-scale image was then loaded. These settings were qualitatively defined by the authors, and were as follows: gain, − 5; speckle reduction imaging (SRI), 2; CrossXBeam CRI (Compound Resolution Imaging), 3; CRI filter, high; enhance, 2; reject, 25; and harmonics, high. The probe was manipulated to show the maximal diameter of the stimulated ovary and then held still to stabilize the image. The 3D facility was selected and the truncated sector adjusted to include the whole ovary. An automated mechanical sweep of this region of interest was undertaken using the slow-sweep mode and the 3D data were stored to the hard drive of the machine. Follicles were chosen randomly, and we limited assessment to between two and three follicles within each ovary to ensure that each follicle studied was clearly identified. A manual record of each follicle was made and the follicle identified on the ultrasound machine by labeling a 2D image with a number that correlated with the embryologists' records. This was achieved by taking a JPEG image of the follicle(s) after each was labeled on screen with the same identifying name and number as that used by the embryologists. This image was saved and also printed out. The 3D acquisition was performed immediately after the 2D image was captured to ensure that the same follicle was measured in each case.

The ultrasound data were subsequently analyzed by a single observer (J.C.) who was unaware of the true volume. Three different methods were used to estimate the volume of each follicle and these were applied in a random fashion so that only one method was used for each dataset at any one time. Each dataset had to be opened three times, therefore, and the observer (J.C.) was blinded to the origin of the dataset by another observer (K.J.) who controlled the order in which the datasets were analyzed. All measurements were made using the multiplanar view. This provides visualization of the three orthogonal planes simultaneously and allows the dataset to be manipulated so that the central point of the follicle is consistent for all three images (Figure 1). Once the dataset had been positioned correctly and magnified to provide the best view, one of three different measurement methods was employed: manual measurement, with volume calculated by means of the sphere formula using the mean 2D diameter; manual measurement using Virtual Organ Computer-aided AnaLysis (VOCAL); and automatic measurement using SonoAVC.

thumbnail image

Figure 1. Multiplanar view demonstrating the three orthogonal views of a stimulated ovary, with the follicle of interest highlighted by a centrally placed marker.

Download figure to PowerPoint

The first technique used to estimate follicular volume from the mean follicular diameter employed the formula for the volume of a sphere: 4/3 × π × radius3. The mean diameter was calculated in three different ways based on an unpublished prestudy questionnaire and a review of descriptions in the literature. The questionnaire was sent to all fertility units and ultrasound departments within the UK as part of a postal survey designed to ascertain current practice, and included a picture of a follicle for the person completing the form to indicate how he or she would measure its diameter. The first 2D technique involved a single measurement made obliquely in the longitudinal plane (A plane) to provide a subjective estimate of the mean diameter. The second 2D technique involved measuring the follicular diameter twice in the longitudinal plane (A plane). The A-plane assessment included measurement of the perceived maximal diameter and the diameter at 90° to this. The final 2D technique involved three assessments of the follicle, which was measured twice in the longitudinal (A) plane, as described above, and once in the transverse (B) plane after the dataset had been rotated manually about the x-axis to provide a transverse view. This third measurement was made perpendicular to the first two measurements.

The second measurement technique was also manual but used VOCAL to define the wall of the follicle (Figure 2). VOCAL allows the user to manually define any volume of interest by tracing around its border in a number of steps as the dataset is rotated about 180°. This has been shown to be a highly accurate and reproducible technique in both the in-vivo and in-vitro settings. For this study 9° rotation steps were used, which made 20 planes available for the measurement of each follicle, as this has been shown to offer the best compromise between validity, reliability and time taken6.

thumbnail image

Figure 2. Virtual Organ Computer-aided AnaLysis (VOCAL) was used to manually define the contour of the follicle.

Download figure to PowerPoint

The third measurement technique used the new software to automatically measure the follicle. The dataset was prepared in the same way as above, and the SonoAVC facility was activated when it had been correctly positioned and magnified. The settings of growth and separation within the software were kept uniform at default value of ‘mid’ for all follicle measurements. SonoAVC identifies the follicle by giving it a specific color and provides automated measurements of its mean diameter (relaxed sphere diameter), its maximum dimensions (xyz diameters) and its volume (Figure 3)5.

thumbnail image

Figure 3. Automatic volume calculation (SonoAVC) was used to automatically calculate the volume of the follicle.

Download figure to PowerPoint

Three measurements were performed for each measurement method and the mean value used for statistical analysis. The true volume of each follicle was ascertained by manual measurement of the follicular aspirate. Each aspirate was measured three times for completeness and the mean measurement taken as the ‘true volume’. Limits of agreement were used to examine the validity of measurements based on this true measure7. Although the mean difference was calculated and is presented, the range between the upper and lower limit of agreement provides more information on the validity of the technique. Techniques with the narrowest range between the upper and lower limits of agreement were considered to be the most accurate3. We have also presented the median value and range to allow appreciation of the distribution of the data. As the absolute size of an object may affect the reliability and validity of measurements we conducted subgroup analysis based on absolute follicular size. To do this, follicles were categorized into four subgroups by each quartile of their volumes and the limits of agreement were calculated within each group to assess the effect of the follicular volume on the measurement accuracy of each technique.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Two hundred and twenty-four follicles with a mean follicular volume of 3.7 (range, 0.4–16.2) cm3 were studied in 51 women. Measurements of aspirate volume were highly repeatable with a mean (95% CI) intraclass correlation coefficient of 0.9998 (0.9996–0.9999). The automated software provided highly accurate automatic follicular volume measurements in all cases (Table 1). VOCAL provided the second most accurate measurement of the follicular volume and, based on the difference between the upper and lower limits of agreement, proved more accurate than measurements made from the automatically calculated xyz diameters. When the measurements were made manually using 2D images, the accuracy was better when three orthogonal diameters were measured than when volume was estimated from a single diameter or two perpendicular diameters.

Table 1. Measurement reliability of follicular volume using different methods: limits of agreement for different methods calculated using true follicular fluid volume as control
Measurement methodMean ± SD (median; range) volume (cm3)Mean ± SD difference (cm3)LLA (cm3)ULA (cm3)Range between LLA and ULA (cm3)
  1. 2D, two dimensional; LLA, lower limit of agreement; ULA, upper limit of agreement; VOCAL, Virtual Organ Computer-aided AnaLysis.

True follicular volume (control)3.70 ± 2.60 (3.1; 0.4–16.2)0000
Automated volume3.67 ± 2.51 (3.1; 0.6–15.8)− 0.04 ± 0.25− 0.520.460.98
Automated xyz diameter4.11 ± 2.87 (3.6; 0.7–18.4)0.41 ± 0.45− 0.471.281.75
VOCAL3.73 ± 2.52 (3.2; 0.7–15.8)0.02 ± 0.33− 0.620.671.29
2D single diameter4.40 ± 3.42 (3.4; 0.7–18)0.70 ± 1.50− 2.243.645.88
2D two diameters4.08 ± 3.22 (3.4; 0.2–17.4)0.38 ± 1.37− 2.303.065.36
2D three diameters3.73 ± 2.65 (3.1; 0.12–17.5)0.03 ± 0.94− 1.801.863.66

Subgroup analysis was performed to assess the effect of follicular volume on measurement accuracy (Table 2). There was a distinct trend towards a decrease in accuracy as the follicular volume increased with all of the manual 2D measurements, as indicated by an increasing range between the upper and lower limits of agreement. However, volume measurements made using SonoAVC and VOCAL showed much better accuracy across the subgroups than manual measurements.

Table 2. Measurement reliability of follicular volume using different methods: limits of agreement for different methods according to the follicular volume divided by quartiles into four subgroups
Follicular volume (cm3)Number of folliclesMean difference (LLA, ULA) (cm3)
Automated volumeAutomated xyz diameterVOCAL2D single measure2D two measures2D three measures
  1. 2D, two dimensional; LLA, lower limit of agreement; ULA, upper limit of agreement; VOCAL, Virtual Organ Computer-aided AnaLysis.

0.4–1.8570.090.230.170.14− 0.08− 0.05
  (−0.43, 0.62)(−0.17, 0.64)(−0.44, 0.78)(−0.78, 1.05)(−1.31, 1.16)(−1.37, 1.28)
1.9–3.054− 0.010.280.070.550.580.27
  (−0.49, 0.48)(−0.13, 0.68)(−0.61, 0.75)(−1.42, 2.52)(−1.29, 2.45)(−1.42, 1.69)
3.1–4.756− 0.020.33− 0.010.200.350.14
  (−0.23, 0.22)(−0.05, 0.71)(−0.41, 0.38)(−2.05, 2.44)(−1.67, 2.36)(−1.05, 1.33)
4.8–16.257− 0.200.78− 0.131.910.66− 0.24
  (−0.70, 0.30)(−0.59, 2.15)(−0.85, 0.60)(−2.27, 6.09)(−3.71, 5.03)(−3.08, 2.59)

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

This is the first study to test the validity of SonoAVC as a tool for the automatic calculation of follicular volume. The software proved easy to use and was highly accurate regardless of the absolute volume. Volume calculations made with SonoAVC were significantly more valid and reliable than those made manually with VOCAL or through application of the sphere formula after measuring the mean follicular diameter. There are no published studies to which these data can be compared, but the results are promising and suggest that the software can be used in the clinical setting.

A high degree of accuracy in measuring the volume of irregular objects using the 3D VOCAL technique has been demonstrated in an in-vitro study6. Follicular shape is less likely to affect the volume measurement with VOCAL as this allows the user to outline the contour of the follicle serially through different planes as the volume is rotated 180° about a central axis. Measurement accuracy is best when a small rotation step is used as this generates multiple planes for the user to measure for each single volume calculation. The technique is limited, however, as the manual measurements take longer than conventional measurements. SonoAVC addresses this and proved more accurate than VOCAL.

Using diameter as a measure assumes sphericity. The lower accuracy of follicular volume measurements made using 2D images may relate to technical difficulties encountered in defining the diameters of distorted follicles within a hyperstimulated ovary4. This may also explain why volume calculations based on the mean automated xyz diameter also proved less valid. These measures represent the longest diameters of a follicle and assume that the volume is elliptical. In contrast to manual measurements using 2D and the automated xyz diameter, SonoAVC utilizes the number of voxels within the color-coded volume of interest to calculate the volume of a follicle. It is a true measure of volume, therefore, and not based on any mathematical assumptions.

This study was designed specifically to assess the validity of the new software and was limited to examination of single follicles. This allowed confirmation of a standard which was derived by measuring the actual follicular aspirate. Assessment of single follicles is relatively straightforward, but the findings now allow the software to be tested in prospective studies examining multiple follicles, which is a more challenging task but has more clinical relevance. Even if we accept that the software performs at least as well as current manual methods, its true potential lies in the fact that it provides automated measurements and this has clinical implications if it can be used to assess the ovary as a whole.

In conclusion, the new software SonoAVC provides highly valid automatic measurements of follicular volume. These measurements are more accurate than volumes estimated from 2D manual measurements and automated xyz measurements of follicular diameter and those calculated using VOCAL.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

SonoAVC was developed by GE Medical Systems in association with K plus Competence Center (Advanced Computer Vision) and partly funded by the K plus Program.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References