To evaluate the repeatability of a novel automated technique called Smart ERA (Smart Endometrial Receptivity Analysis) for the automated segmentation and volume calculation of the endometrium in patients with normal uteri,, and to compare the agreement of endometrial volume measurements between Smart ERA, the semi-automated Virtual Organ Computer-aided Analysis (VOCAL) technique and manual segmentation. This retrospective study evaluated endometrial volume measurement in infertile patients who underwent frozen-thawed embryo transfer (FET). Transvaginal three-dimensional ultrasound scans were performed using a Resona R9 ultrasound machine. Data was collected from patients between 2021 and 2022. Patients with normal uteri and optimal ultrasound images were included. Endometrial volumes were measured using Smart ERA, VOCAL at 15° rotation, and manual segmentation. Intra-observer repeatability and agreement between techniques were assessed using the intraclass correlation coefficient (ICC) and Bland–Altman analysis. A total of 407 female patients were evaluated (mean age 33.2 ± 4.7 years). The repeatability of Smart ERA showed an ICC of 0.983 (95% CI 0.984–0.991). The agreement between Smart ERA and the manual method, Smart ERA and VOCAL, and VOCAL and the manual method, as assessed by ICC, were 0.986 (95% CI 0.977–0.990), 0.943 (95% CI 0.934–0.963), and 0.951 (95% CI 0.918–0.969), respectively. The Smart ERA technique required approximately 3 s for endometrial volume calculation, while VOCAL took around 5 min and the manual segmentation method took approximately 50 min. The Smart-ERA software, which employs a novel three-dimensional segmentation algorithm, demonstrated excellent intra-observer repeatability and high agreement with both VOCAL and manual segmentation for endometrial volume measurement in women with normal uteri. However, these findings should be interpreted with caution, as the algorithm's performance may not be generalizable to populations with different uterine characteristic. Additionally, Smart ERA required significantly less time compared to VOCAL and manual segmentation.
Read full abstract