Radiomics and deep learning for stratification of uterine fibroid growth risk from baseline ultrasound

Uterine fibroids are the most common benign tumors in reproductive-age women, exhibiting highly variable growth rates that pose a major challenge in clinical management. Accurately forecasting future fibroid growth using clinical assessment or conventional imaging is essentially impossible, leaving clinicians without reliable tools for individualized treatment planning. In this exploratory study, we evaluated whether baseline B-mode ultrasound-based radiomics— quantitative image feature extraction—could help predict fibroid growth patterns. We retrospectively analyzed 189 fibroids from 97 patients: 62 regressed, 69 grew moderately (