Computer assisted small bowel capsule endoscopy video segmentation related to anatomical landmarks of the gastrointestinal tract

Introduction: Small bowel capsule endoscopy (SBCE) propose a new non-invasive method for the diagnosis of digestive tract disorders. However, the medical reading of long segment video data is time consumptive. The aim of our present study was to develop a new computer algorythm as to assist the physician to interpret a SBCE video by automatic segmentation into different anatomic parts of the digestive tract. Methods: A new computer assisted video segmentation scheme is proposed to locate the anatomical border between the stomach, small intestine, and large intestine. To achieve this we utilize the different color feature of the SBCE video color bar and we draw a dissimilarity curve to obtain an approximate localization of the anatomical segment borders on the SBCE video. The color histogram in the HSI color space is used to segment the stomach, small intestine and large intestine, which was based on computer spectral analysis of RGB intensities in the color bar (CB) of Miroview SBCE analysis system with the software Colors. Our software analyzed the average and relative intensity in the color histogram of the blue, green and red color in the (RGB) components (GCI = G/R+G+B, and RCI = R/R+G+B). Results: The relative intensity in the histogram of the red color (RCI) was significantly higher in the duodenum as compared to the antral region, 0.49+0.03 vs. 0.44+0.02, respectively (p < 0.000001). Similarly, the relative intensity in the histogram of the green color (GCI) was significantly higher in the cecal as compared to the ileal segment, 0.35+0.01 vs. 0.31+0.01, respectively (p < 0.00001). The average rise of RCI at the pylorus was 10.9+2.8%, wheras the average rise of GCI at the ileo-cecel valve was 14.1+4.0%. The average precision were 100% for the stomach/small intestine and 90% for the small/large intestine discrimination. Conclusions: Our new computer assited video segmentation method can accurately locate the border between different anatomical regions of the gastrointestinal track on the SBCE video, and may result a simplicity of SBCE examination with an automatic detection of the end of the small bowel study during ambulatory SBCE investigation.