Bleeding detection in wireless capsule endoscopy based on MST clustering and SVM

Wireless capsule endoscopy (WCE), a widely used tool for gastrointestinal tract lesion detection, produces a large number of images. However, it is difficult and inefficient to manually search for and find bleeding lesions among these images. In this study, a new method for the automatic identification of bleeding frames in WCE has been proposed. The minimum spanning tree (MST) clustering algorithm was applied to reduce the background noise. Then, the images´ features are extracted from the clustering result and classified by support vector machine (SVM). The experimental results, suggest that this method can correctly identify and mark the bleeding areas in WCE images. The final sensitivity, specificity and accuracy of our proposed method are 91.69%, 94.59%, and 94.10%, respectively.