Automatic segmentation of pelvic floor ring on 3D ultrasound

Finished: 2020-03-03

BSc assignment

Pregnancy and delivery have a high impact on the pelvic floor, causing problems like incontincence and pelvic organ prolapse. In the case of pelvic organ prolapse it might be benificial to put a ring in the vagina to prevent the prolapse of the vaginal wall or the uterus. Sometimes this works amazing and most pelvic floor problems are gone. In other cases however the ring falls out and the problems are still there. Little is understood about why this works so well in some cases and doesn't work at all in other cases. 

To better understand this problem we collected a dataset, within the Gynius project, which contains ultrasound data directly after the placement of the ring and after 3 months. To analyse this data the ring should be segmented, ideally automatically. The ring appears dark on ultrasound and therefor the segmentation should be doable with (semi)automatically with classical machine learning algorithms and, if needed, automatically with deep learning.

The goal of the bachelor assignment is to segment the pelvic floor pessary ring automatically on 3D pelvic floor ultrasound data