Edge density based automatic detection of inflammation in colonoscopy videos. (1st May 2016)

• Record Type: Journal Article
• Authors: Ševo, I.; Avramović, A.; Balasingham, I.; Elle, O.J.; Bergsland, J.; Aabakken, L.
• Is Part Of: Computers in biology and medicine. Volume 72(2016)
• Page Start: 138
• Page End: 150
• Publication Date: 2016-05-01
• Subjects: Colonoscopy -- Inflammation -- Texture -- Automatic detection; Medicine -- Data processing -- Periodicals; Biology -- Data processing -- Periodicals; 610.285
• DOI: 10.1016/j.compbiomed.2016.03.017
• Languages: English
• ISSNs: 0010-4825

Abstract

Abstract: Colon cancer is one of the deadliest diseases where early detection can prolong life and can increase the survival rates. The early stage disease is typically associated with polyps and mucosa inflammation. The often used diagnostic tools rely on high quality videos obtained from colonoscopy or capsule endoscope. The state-of-the-art image processing techniques of video analysis for automatic detection of anomalies use statistical and neural network methods. In this paper, we investigated a simple alternative model-based approach using texture analysis. The method can easily be implemented in parallel processing mode for real-time applications. A characteristic texture of inflamed tissue is used to distinguish between inflammatory and healthy tissues, where an appropriate filter kernel was proposed and implemented to efficiently detect this specific texture. The basic method is further improved to eliminate the effect of blood vessels present in the lower part of the descending colon. Both approaches of the proposed method were described in detail and tested in two different computer experiments. Our results show that the inflammatory region can be detected in real-time with an accuracy of over 84%. Furthermore, the experimental study showed that it is possible to detect certain segments of video frames containing inflammations with the detection accuracy above 90%. Abstract : Highlights: A model based method for automatic inflammation detection in colonoscopy videos is introduced. The method relies on a high quality display provided by Olympus colonoscopy probe. The proposed method is suitable for parallel implementation and real-time processing of high-resolution colonoscopy videos. Real-time inflammation detection can provide the gastroenterologist with a useful tool to enable faster and more accurate diagnosis.