Abstract: To assess the interobserver reliability of antral follicle counts (AFC) using real-time two-dimensional (2D) ultrasound and three-dimensional (3D) ultrasound. Methods: Two observers conducted transvaginal ultrasound examinations in 51 subfertile women in the early follicular phase of the menstrual cycle. Antral follicles were counted using real-time 2D ultrasound and the time taken was recorded. A 3D volume was then acquired from each ovary and stored for subsequent offline analysis using the multiplanar view. The time taken for each step was recorded and the total time was calculated. Intraclass correlation coefficients (ICC) and limits of agreement were used to assess the reliability. Results: There was no difference in the mean antral follicle counts using real-time 2D (18.63±11.39) and 3D (18.73±11.74) ultrasound. According to the ICC, there was a significantly higher interobserver reliability for counts made using 3D (mean, 0.994; 95% CI, 0.990-0.997) as compared with real-time 2D ultrasound (mean, 0.979; 95% CI, 0.963-0.988), P<0.01. 3D ultrasound was also associated with narrower limits of agreement (-3.46, 3.35) than was 2D ultrasound (-6.78, 6.31). While the total time taken was significantly longer for the 3D technique ［(204.0±53.0) s  vs. (112.4±34.8) s, P<0.001］, the time required for the actual ultrasound examination was significantly less ［(48.0±7.5) s  vs. (112.4±34.8) s, P<0.001］.Conclusion: 3D ultrasound significantly improves the interobserver reliability of antral follicle counts. While this is at the expense of time overall, the duration of the actual ultrasound examination and the patient exposure is significantly reduced using 3D compared with real-time 2D ultrasound.

Key words: Ultrasonography, Imaging, three-dimensional, Ovarian follicle, Observer variation