Wireless capsule endoscopy represents a color imaging technology in the field of medical endoscopy that is extensively used to detect lesions of the human digestive tract. It is the golden standard in evaluating small bowel lesions, offering a set of digital snapshots difficult to get using other investigation methods. Its major drawbacks are the time consumed for image analysis and the burden for the physicians that must spot and classify lesions within more than 55000 images. This paper carries out a study on the detection of telangiectasia in the small bowel, based on an adapted color slicing technique applied not only on unique frames, but on series of successive frames, performing a global analysis suitable on partial sequences or entire wireless capsule endoscopy movies. We have quantified the extracted features and determined a weighting algorithm to find telangiectasia lesions. For frames containing potential lesions, we have determined features not only for the global image, but also for the normal mucosa surrounding the lesion extracted from the image. This approach allows the physician to see variations of parameters within a frame or a sequence that contains lesions. Experimental results prove that the algorithm is effective in detecting telangiectasia patterns of different images, with an accuracy of 93.88%, reducing thus the time spent for the analysis of the images acquired by wireless capsule endoscopy.
Keywords: Wireless Capsule Endoscopy; color slicing technique; medical image analysis; telangiectasia detection.