Reproducibility of three-dimensional gel installation sonohysterography in the assessment and classification of intrauterine abnormalities

Objective: Purpose of this study is to determine the interobserver and intraobserver variability of 3D GIS in the assessment of intrauterine abnormalities. Study design: Forty five 3D volumes were randomly selected from a larger prospective cohort study that studied the diagnostic accuracy of 3D GIS in addition to 2D GIS. To study interobserver agreement volumes were reviewed by two independent examiners. One examiner reviewed these samples twice with an interval of 1 month in a random order. Interobserver and intraobserver agreement were tested with Cohen's kappa coefficient and shown in Bland and Altman plots. Quality of the 3D volumes was evaluated. Results: Cohen's kappa for interobserver variability for type of abnormalities (none, polyp, fibroid, other) was 0.64 and for presence of a fibroid (fibroid yes/no) 0.77. Agreement on type of fibroid was 0.59. Intraobserver agreement was almost perfect for type of abnormality (Cohen's kappa of 1.0) and good for fibroid diameter. Quality of the 3D volumes was poor in 11 out of 45 cases. Reproducibility increased when poor quality images were excluded. Conclusion: Substantial interobserver and intraobserver agreement for 3D GIS in the diagnoses of intrauterine abnormalities was found. 3D GIS interobserver and intraobserver agreement are good for fibroid diameter and moderate for volume and protrusion. © 2014 Elsevier Ireland Ltd. All rights reserved.