Inter-rater agreement in the diagnosis of adenomyosis by 2- and 3-dimensional transvaginal ultrasonography.

Abstract

To assess the inter-rater agreement of 2-dimensional (2D) and 3-dimensional (3D) transvaginal ultrasonography (TVUS) in the diagnosis of adenomyosis. This prospective study included premenopausal women (n = 96) with heavy menstrual bleeding, menstrual pain, or both who were scheduled for hysterectomy or transcervical resection of the endometrium. All women underwent real-time 2D TVUS and subsequently offline 3D TVUS, which was blinded to 2D TVUS, by a single expert rater and a single nonexpert rater for the diagnosis of adenomyosis based on standardized pattern recognition and junctional zone measurements. Three-dimensional TVUS was done on a computer with 3D volumes recorded during 2D TVUS by both raters. The expert rater reported the image quality of all 3D volumes (n = 192). Inter-rater agreement (Cohen's κ) was assessed for both techniques, and the improvement over time was assessed for 2D TVUS. Diagnosis of adenomyosis showed good (κ = 0.69) and poor (κ = 0.21) inter-rater agreement with 2D and 3D TVUS, respectively (P < .05). The agreement with 2D TVUS improved over time. The agreement with 3D TVUS was slightly better for expert-recorded 3D volumes (κ = 0.40), which also had better image quality (P < .05). The most reproducible 2D and 3D features were anechoic lacunae (κ = 0.52) and junctional zone irregularity (κ = 0.27), respectively. Standardized pattern recognition during real-time 2D TVUS may result in good agreement between expert and nonexpert raters for the diagnosis of adenomyosis. Offline 3D TVUS is less reproducible, and junctional zone measurements do not improve the inter-rater agreement. The low inter-rater agreement may be related to a lack of experience and low image quality of nonexpert-recorded 3D volumes.