Cookies on this website
We use cookies to ensure that we give you the best experience on our website. If you click 'Continue' we'll assume that you are happy to receive all cookies and you won't see this message again. Click 'Find out more' for information on how to change your cookie settings.

© 2020 IEEE. Ureteroscopy has evolved into a routine technique for treatment of kidney stones. Laser lithotripsy is commonly used to fragment the kidney stones until they are small enough to be removed. Poor image quality, presence of floating debris and severe occlusions in the endoscopy video make it difficult to target stones during the ureteroscopy procedure. A potential solution is automated localization and segmentation of the stone fragments. However, the heterogeneity of stones in terms of shape, texture, as well as colour and the presence of moving debris make the task of stone segmentation challenging. Further, dynamic background, motion blur, local deformations, occlusions and varying illumination conditions need to be taken into account during segmentation. To address these issues, we compliment state-of-the-art U-Net based segmentation strategy with the learned motion information. This technique leverages difference in motion between the large stones and surrounding debris and additionally tackles problems due to illumination variability, occlusions and other factors that are present in the frame-of-interest. The proposed motion induced U-Net (MI-UNet) architecture consists of two main components: 1) U-Net and 2) DVFNet. The quantitative results show consistent performance and improvement over most evaluation metrics. The qualitative validation also illustrate that our complimentary DVFNet is able to effectively reduce the effect of surrounding debris in contrast to U-Net.

Original publication

DOI

10.1109/ISBI45749.2020.9098608

Type

Conference paper

Publication Date

22/05/2020

Volume

2020-April

Pages

212 - 216