A retrospective cohort study published in the International Journal of Gynaecology and Obstetrics evaluated whether deep learning-based automated cervical length measurement could improve spontaneous preterm birth risk stratification in women with a borderline short cervix of 20–25 mm by conventional measurement. The study included 171 women evaluated between January 2019 and December 2023 at a tertiary referral hospital in South Korea. Women were reanalyzed using the previously validated CL-Net algorithm, which traces the cervical canal, and classified as deep learning-Short (≤25 mm) or deep learning-Normal (>25 mm).

The primary outcome was preterm birth before 37 weeks of gestation. Thirty-nine women (22.8%) were classified as deep learning-Short and 132 (77.2%) as deep learning-Normal. Preterm birth before 37 weeks occurred in 28.2% versus 6.1%, respectively (P<0.01), and a deep…