Automatic segmentation of pelvic floor ring on 3D ultrasound

The pelvic floor muscles are fundamental in stabilizing the structure of organs and tissue in the pelvic region. However, pregnancy as well as delivery can lead to con-siderable damage to the muscles, leaving organs in that region (such as the bladder, uterus or rectum) to possibly shift and descend. This is called pelvic floor prolapse. As this can lead to serious health problems, a pessary ring can be inserted in the vagina to stabilize the vaginal wall while the muscles recover.

In addition to having a low success rate of around 60%, some patients have complaints and experience pain throughout the procedure. In some cases the ring falls out of the vagina. As not much is known about the position or orientation of the ring, deter-mining the success of a pessary procedure is difficult. 3-dimensional ultrasound data of the patients pelvic floor region was gathered as part of the Gynius1 project. A first step to locating the ring in the vagina is to locate the ring within the 3-dimensional ultrasound data.

Some particular challenges faced in this project include the high noise level inherent to ultrasound capture per se, the inconsistent appearance of the ring in the data as well as the partial image of the ring in the given data due the ring diameter exceeding the maximum depth reachable by the ultrasound capture device.

In this presentation different algorithms are analysed, presented and evaluated based on their viability in determining the position of the ring in the data.