3D pelvic floor ultrasound-based quantitative analysis for postpartum pelvic floor dysfunction

Objective Three-dimensional (3D) Pelvic Floor Ultrasound (PFU) imaging has been used in clinical practice to visualize the pelvic floor's anatomical structures and functional changes related to postpartum Pelvic Floor Dysfunction (PFD). This study aimed to evaluate the clinical value of standardized 3D PFU imaging and quantitative ultrasound parameters in assessing postpartum PFD. Methods We performed a retrospective cohort study that included parturients who delivered from January 2024 to June 2025, underwent standardized postpartum follow-up at five pelvic floor disease diagnosis and treatment centers. Participants were divided into a rehabilitation group (who received standardized PFM training starting at 6 weeks postpartum) and a control group (who did not receive structured rehabilitation intervention). A 3D PFU examination via transperineal approach was performed and multi-planar imaging data were collected, as well as dynamic imaging data. We retrieved quantitative ultrasound parameters, including anteroposterior diameter, transverse diameter, levator hiatus area, bladder neck mobility and urethral rotation angle at resting state or pelvic floor contraction or during Valsalva maneuver condition. Comparative analysis was used to evaluate group differences. Results The rehabilitation group demonstrated significantly larger levator hiatus-related measurements and dynamic mobility parameters compared with controls in all functional states (all p < 0.05). Because larger hiatal dimensions and greater bladder neck mobility may reflect more severe pelvic floor support impairment, these between-group differences were interpreted as quantitative imaging findings rather than direct evidence of treatment superiority. Conclusion 3D PFU imaging can visualize pelvic floor morphology and provide reproducible quantitative metrics for functional assessment. Standardized image acquisition and quantitative parameter analysis may provide a quantitative basis for future intelligent-assisted diagnostic systems; however, no artificial intelligence or machine-learning model was developed or validated in this retrospective study. Therefore, the findings should be interpreted as foundational imaging evidence rather than proof of an implemented intelligent diagnostic model.