Near Real-Time Automated Quantification of Urethral Mobility During Ultrasound Imaging

Objectives: Urethral mobility is an important factor in stress incontinence and ultrasound imaging can provide clear visualization of this movement. Thus far, its quantification has been a manual process, that focuses on only a small part of the overall motion. An automated system to track urethral movement could provide richer information about urethral mobility and also reduce inter-observer effects on measurements. We hypothesize that it is possible to automatically track and quantify urethral mobility in near real-time ultrasound video clips.

Methods: As a proof of concept for technique development, we used transperineal video clips obtained on commercial ultrasound scanners (GE Healthcare and Philips Healthcare) from consented research volunteers with (n=2) and without (n=4) incontinence. We developed the tracking software with Python programming language and Open Source Computer Vision Library, using available corner detection (goodFeaturesToTrack) and optical flow tracking algorithms (calcOpticalFlowPyrLK). We tracked movement in three regions of interest (ROIs) along the length of the urethra relative to a 4th reference ROI; the pubic bone. For each ROI, its upper left corner was used as the origin in a Cartesian coordinate system.

Results: Examples during cough (a&b) and Kegel exercise (c&d) are shown in Fig. 1, displaying rest (a&c) to maximal urethral motion (b&d). Visual inspection of the real-time videos with superimposed tracking ROIs confirmed excellent fidelity to visualized movement. Movement quantification during the cough is seen in Fig. 2. The relative translations of ROIs in the (a) horizontal (x-axis) and (b) vertical (y-axis) directions, and (c) rotation of the 3 ROIs are shown. Horizontally (x-axis), the distal (ROI 1) and proximal (ROI 3) portions of the urethra move in the opposite directions with respect to the pubic bone. Similarly, the urethra’s motion in the vertical direction is not uniform as well. Finally, while all ROIs rotate counterclockwise with respect to the pubic bone, the distal part of the urethra rotates by almost twice of the proximal part measuring urethral bending (“kinking”).

Conclusions: These proof of concept analyses demonstrate the ability to automatically track movement (translation and rotation) of several urethral segments relative to the pubic bone in a matter of seconds ( < 5 s). This novel system, once perfected, should allow analysis of the full details of urethral movement on any video file, independent of the platform used for capture. It should prove useful in studying the relationship between the dynamics of urethral closure pressure and urethral motion.
Supported by NIH-RC2-DK-122379.