Deep Learning in Obstetric Sonography: A Literature Review of Techniques, Applications, and Challenges

Abstract

Objective: Fetal sonography is a cornerstone of prenatal care, enabling noninvasive monitoring of fetal growth and early detection of congenital anomalies. Despite its clinical value, its effectiveness can be limited by operator dependency, subjective interpretation, and variability in image quality. To address these issues, researchers have suggested using deep learning (DL) models as a maneuver to assist sonographers. Materials and Methods: A search of the current literature was conducted using PubMed, IEEE Xplore, SpringerLink, and Scopus, to create a targeted and current review on the use of DL in obstetric sonography. These platforms are renowned for their high-quality, peer-reviewed publications in both medical imaging and artificial intelligence. The timeframe is selected between 2019 and May 2025, which aligned with the period that DL experienced rapid expansion, into clinical applications. Results: The systematic review found that DL has numerous therapeutic uses in prenatal imaging, including classifying aberrant and normal anatomy and measuring fetal biometry. This review consolidated recent developments in the application of DL such as classification, segmentation, and automation in obstetric processes. Conclusion: Deep learning has a great potential to guide future diagnostic directions for sonographers, clinicians, and researchers involved in fetal sonography.