Bleeding Detection in Wireless Capsule Endoscopy Videos using Temporal Features

Y. Chen, A.K. Lakshmanan, and J. Lee (USA)


Wireless capsule endoscopic application, bleeding detection.


The wireless capsule endoscopy (WCE) has been widely used to diagnose the gastrointestinal tract (GI) diseases in the human digestive tract because of its ability to view the entire small intestine non-invasively. Bleeding is one of abnormalities that the GI specialist looks for. However, a typical problem associated with WCE is the huge amount of image data to be examined visually, which consumes a lot of productive time of the GI specialist. In order to speed up the viewing process, we propose a methodology for bleeding detection based on temporal features. By utilizing a sequence of images, we analyzed the temporal change over the HSV color space for suspicious bleeding detection. This way we could reduce the number of false positives caused by capsule glitches. The key aspect of the proposed algorithm is to define the color range for blood over the HSV color space. In addition a threshold value is updated automatically based on the non-bleeding images to identify potential bleeding cases. Algorithms for identifying bleeding spots and severe bleeding are implemented and experimented in this work.

Important Links:

Go Back