Locating Standard Planes in Endometrium Ultrasound Videos via Clinical Prior Knowledge of Key Anatomical Structures

Accurate localization of standard planes is crucial for sonog-raphers diagnosing and evaluating endometrial function from ultrasound videos. Traditionally, this process is performed manually, resulting in significant inter- and intra-observer variability. To address these challenges and enhance diagnos-tic consistency, we propose an automated model that localizes endometrial standard planes by detecting key anatomical structures. Specifically, our model segments the endometrium and identifies uterine cavity lines, facilitating standard plane localization based on clinical prior knowledge. Experimen-tal results demonstrate that our model outperforms existing methods, achieving an AUC of 0.922 and an accuracy of 0.869. The model's predictions provide consistent and in-terpretable results, supporting sonographers in their assessments.