Presented by: Jin Yong MMed (Surg)
Singapore General Hospital, Singhealth, Singapore

Introduction

Overactive bladder (OAB) is an increasingly prevalent urological condition. Routine urodynamics study (UDS) is invasive, inconsistent and simply impractical in every outpatient setting. We aim to prove that detrusor instability in OAB can be identified based on keypoint movements of the vascular network of the detrusor using our novel cystoscopy based machine vision software.

Materials

We prospectively collected 158 cystoscopy videos and extracted 30-60 second clips from each video. The recordings were first filtered to remove videos with significant artefacts, leaving 109 cystoscopy videos (31 previously UDS confirmed DO, 78 non-OAB). Thereafter, the videos were enhanced to improve resolution and enlarge the vascular network on the detrusor. Mosaic stitching and 3D Mapping was performed to stitch single frames into a large panorama that was reconstituted into a 3D sphere to emulate the shape of the bladder. Thereafter, the vascular network underwent UNet Segmentation to create keypoints for analysis, generating an average of 300 keypoints per frame. Finally, the movement of these keypoints over time was generated into a heat map using Keypoint Motion Spectra, as a surrogate for areas of detrusor microcontraction.

Results

,

Results demonstrated a greater amount of overall movement of keypoints in the OAB group as compared to the non-OAB group. Combined heatmap also evidenced a greater amount of movement per frame in each cystoscopy video (Fig 1). This is reiterated numerically by the greater mean pixel deviations per frame in the OAB group of 84.71 compared to 26.23 in the non-OAB group (Fig 2).

Conclusion

Our novel machine vision augmentation model yields promising results in identifying DI in OAB patients, using movements of keypoints on the blood vessel network in our cystoscopy based software. Differentiating this degree of movement can potentially allow for diagnosis of OAB using a cystoscopy based tool in the future, as well as allow for targeted areas of intra-detrusor treatment to minimize side effects and maximize efficacy of drugs.

Funding

None

Co-Authors

Yuguang Tan, MRCS
Singapore General Hospital, Singhealth, Singapore

Shauna Jia Qian Woo, MB.BS
Singapore General Hospital, Singhealth, Singapore

Lay Guat Ng, MMed (Surg)
Singapore General Hospital, Singhealth, Singapore

Mark Kei Fong Wong, PHD
Endosiq Technology Singapore