A segmented color plane histogram based feature extraction scheme for automatic bleeding detection in wireless capsule endoscopy

Wireless capsule endoscopy (WCE) is a recently developed revolutionary video technology to visually inspect the entire gastrointestinal tract in a non-invasive way. However, a major problem associated with this technology is that to detect bleeding, a physician has to analyze the tremendous amount of image frames, which is both time consuming and due to oversight often leads to human error. These limitations give motivation for development of computer aided automatic bleeding detection schemes. In this paper, to investigate bleeding, the analysis of WCE image frames is carried out in normalized RGB (rgb) color space as human perception of bleeding is associated with different shades of red and rgb overcomes some of the drawbacks of conventional RGB color space. In the proposed method, at first, the WCE image frame is segmented based on different ranges of r-values. Then for a certain level of r-value, the variation in g plane is presented with the help of histogram. Features are extracted from the proposed r versus g plane histograms. For the purpose of classification, KNN classifier is employed. Extensive experimentation on several WCE image frames obtained from various publicly available WCE videos makes it evident that the proposed method outperforms some of the existing methods in terms of accuracy (98.12%), sensitivity (94.98%) and specificity (98.55%).