SEGMENTATION OF IMAGES FROM WIRELESS CAPSULE ENDOSCOPY

Marcin Kociolek1,2, Ram D. Sriram1.

1 National Institute of Standards and Technology, Gaithersburg, USA

2 Institute of Electronics, University of Lodz, Lodz Poland

The wireless capsule endoscopy (WCE) is a revolutionary technique which allows the visualization of the entire gastrointestinal (GI) track, especially the small intestine. The investigation of video recording from WCE is performed by trained clinicians. Because of huge amount of images to review (usually about 60.000 frames per video) the investigation takes a significant amount of time. This requires considerable concentration from the physician, so as not to miss lesions that can occur only in one or few frames. Our goal in this project is two fold: 1) extract features from WCE video, through a novel image interpretation technique, and 2) map these features onto an ontology, thus aiding in the semantic interpretation of images. Our software will search each frame and highlight the frames where possible lesions occur. By use of predefined ontology our system will recognize different (GI) diseases. Augmenting images with semantic information will also aid in large scale data base searches. This will aid clinicians to make queries to a database of cases to find semantically similar images.

Feature extraction is applied for each frame of the WCE video. It incorporates two main steps:

-       automatic image segmentation

-       feature vector calculation independently for each segment.

Proper segmentation is a key issue in this task. First, this process should be fully automatic because involving an operator causes prolongation of analysis time. Second, more accurate segmentation will result in better feature values.

This poster presents our approach to segmentation of WCE video frames. A future work plan is also presented

 

Category: Mathematics

 

Author information:

 

Marcin Kociolek

Guest Researcher

mentor: Ram D. Sriram

National Institute of Standards and Technology

MEL

Manufacturing Systems Integration Division (826)

Design Processing Group

office: Metrology (220), Room A122

adres:

100 Bureau Drive, Stop 8263

Gaithersburg, MD 20899-8263

phone: (+1 301) 975-5994

email: marcin@cme.nist.gov

fax: (+1 301) 975-4482