Classification of Bleeding Images in Wireless Capsule Endoscopy using HSI Color Domain and Region Segmentation

Wireless Capsule Endoscopy (WCE) is one of the latest technologies that has evolved in the area of medical endoscopy. The far old endoscopy techniques such as colonoscopy push enteroscopy and intraoperative enteroscopy helped doctors to visualize up to stomach, from the upper part and terminal ileum and colon from the bottom. There exists no method to view most of the small intestine without surgery. WCE is a new technology that enables close examination of the interior portion of the entire small intestine without any surgery. The capsule, small and similar to that of a multi-vitamin capsule has a life expectancy of approximately 8 hours. During that time, it captures around 55,000 images, and transmits the images into a small data receiver worn by the patient on the belt. In spite of this technical revolution in medical history there exists a big drawback. The physician has to spend more than two hours to monitor the WCE video. This drawback has lead to many researchers working on reducing the time assessment of the video. In this paper, we propose a new method that classifies bleeding images in WCE video using the HSI (Hue, Saturation and Intensity) color domain and region segmentation method. In order to classify the bleeding images, we first convert raw data into HSI color domain since it is closer to human perception than the other color domains. Second, we segment each image into bleeding and non bleeding regions using JEDISON region segmentation algorithm. Finally we classify the segmented images into bleeding and non-bleeding by color threshold technique and rigidness of bleeding region. It is assessed that the sensitivity and specificity of our proposed approach are 70.96% and 97.97%, respectively. The result shows that our effort is to achieve classification of bleeding images with higher accuracy compared to existing algorithm, i.e., Suspected Blood Indicator (SBI) in Given Imaging.