Clinical Report: Endo-MedSAM for Uterine Segmentation in Pelvic MRI
Overview
Endo-MedSAM, an adaptation of MedSAM-2, demonstrated improved uterus segmentation in pelvic MRI for endometriosis management. It achieved significant performance enhancements over zero-shot MedSAM-2, particularly with bounding-box prompts.
Background
Endometriosis affects approximately 10% of women of reproductive age and is a leading cause of chronic pelvic pain and subfertility. Accurate diagnosis and management are critical, yet often delayed due to the complexity of symptoms and imaging findings. Automated uterus segmentation in pelvic MRI can enhance diagnostic accuracy and streamline treatment planning.
Data Highlights
Configuration
Mean 3D Dice Score
Bounding-box prompts
0.81–0.88
1-click point prompts
0.68–0.76
Key Findings
Endo-MedSAM achieved mean 3D Dice scores of 0.81–0.88 with bounding-box prompts.
Improvements of approximately 0.27–0.34 in 3D Dice scores were noted compared to zero-shot MedSAM-2.
Performance was consistent across multicenter and single-center settings.
Endo-MedSAM supports both bounding-box and low-interaction point prompting.
Automated segmentation can reduce manual contouring burden and standardize measurements across different imaging protocols.
Clinical Implications
The implementation of Endo-MedSAM can facilitate faster and more reproducible uterus delineation in clinical practice. This may lead to improved consistency in volumetric assessments and enhance the overall management of endometriosis.
Conclusion
Endo-MedSAM represents a significant advancement in automated uterus segmentation for pelvic MRI, offering robust performance that can enhance endometriosis care.
