ARTIFICIAL INTELLIGENCE IN ULTRASOUND IMAGING FOR BENIGN GYNECOLOGICAL DISORDERS

Abstract

Transvaginal ultrasound remains the first-line imaging modality for most benign gynaecological conditions, yet its diagnostic accuracy depends heavily on operator experience, a dependency that artificial intelligence has increasingly been applied to reduce. A 2025 review identified twelve studies developing machine learning or deep learning models, and in some cases combinations of both, to distinguish favourable from unfavourable clinical or anatomical states relevant to benign gynaecological ultrasound, including endometrial receptivity assessment, adnexal mass characterisation, and uterine cavity evaluation prior to embryo transfer, with reported model accuracies reaching as high as 0.94 for certain implantation-related classification tasks [1].