Computer-Aided WCE Diagnosis Using Convolutional Neural Network and Label Transfer

Automatic diagnosis of gastropathy disease plays an important role in supporting the clinician and speeding up the examination process. While it is hard for clinicians to accurately detect gastrointestinal disease due to its great dependence on doctors experiences. In this paper, we propose an algorithm for stomach disease detection using deep learning features. First, images are partitioned into patches using a superpixel segmentation and points on superpixel edges are randomly selected as the center of image patches. Then a Convolutional Neural Network (CNN) is utilized to get more generic features of Wireless Capsule Endoscopy (WCE) stomach images. Finally, the detected patch label is transferred to superpixel labels to get a more accurate lesion detection area. Experiments on our database show that the proposed method can get more accurate disease regions and it outperforms the previous state-of-the-art methods based on hand-crafted features.