Post-surgical Endometriosis Segmentation in Laparoscopic Videos
Andreas Leibetseder, Klaus Schoeffmann, Jörg Keckstein, Simon Keckstein
TL;DR
This paper tackles automating segmentation of dark endometrial implants in laparoscopic videos to support clinical training and post-surgical analysis. It adapts Mask R-CNN with a ResNet-101 backbone to a single-class lesion dataset derived from GLENDA, using augmentation to reach $mAP@0.50IoU = 0.642$ (and $0.324$ for the $0.50-0.95$ range) after 29 epochs. The processing pipeline outputs annotated frames and a per-frame detection summary bar, with detections and masks exportable in JSON for future interactive viewers. Overall, the work demonstrates feasibility of applying established deep-learning object segmentation to a real-world medical use case and provides a baseline for future post-surgical video-browsing tools and clinical education.
Abstract
Endometriosis is a common women's condition exhibiting a manifold visual appearance in various body-internal locations. Having such properties makes its identification very difficult and error-prone, at least for laymen and non-specialized medical practitioners. In an attempt to provide assistance to gynecologic physicians treating endometriosis, this demo paper describes a system that is trained to segment one frequently occurring visual appearance of endometriosis, namely dark endometrial implants. The system is capable of analyzing laparoscopic surgery videos, annotating identified implant regions with multi-colored overlays and displaying a detection summary for improved video browsing.