The interobserver reliability of off-line antral follicle counts made from stored three-dimensional ultrasound data:...

To assess the interobserver reliability of antral follicle counts (AFCs) made from stored three-dimensional (3D) ultrasound data using conventional two-dimensional (2D) images, 3D multiplanar view and 3D-rendered 'inversion mode'.

3D transvaginal ultrasound was performed in the early follicular phase (days 2-5) of the menstrual cycle in 41 subjects aged < 40 years, undergoing investigation for subfertility. From the stored 3D ultrasound datasets, the number of antral follicles of 2-10 mm in diameter in each ovary was independently measured, using all three methods by three investigators, each with a different level of experience. The image quality of each dataset was subjectively categorized into one of three groups, based on the proportion of the ovarian contour that could be seen clearly.

There was no significant difference in the mean AFC between the observers for any of the three different techniques. The intraclass correlation coefficient (ICC) for the 2D-equivalent mode, the 3D multiplanar mode, and the 3D-rendered inversion mode were indicative of good interobserver reliability for each method. The interobserver reliability for the 3D-rendered inversion mode was better with Grade 1 image quality than with Grade 3 image quality. There were no equivalent differences, however, between the three different grades of image quality with the 2D-equivalent and 3D multiplanar modes. The time taken for AFC measurement using 3D-rendered inversion mode was significantly longer than with the 2D equivalent and 3D multiplanar methods.

3D image displays and rendering techniques do not appear to offer any advantage over a conventional 2D display in terms of AFC measurement reliability. AFC measurement using the 3D-rendered inversion mode has an adequate interobserver reproducibility but is dependent on image quality.