ABSTRACT
Gastrointestinal (GI) diseases such as bleeding, ulcer and tumor can be cured or controlled in their early stages, otherwise they will deteriorate into cancers or some other fatal diseases. The best way to diagnose these diseases in their early stages is to directly view the GI tract. Wireless capsule endoscopy is a disposable, pill-shaped device that consists of one lens, Light Emitting Diodes (LEDs), a Complementary Metal Oxide Semiconductor (CMOS) camera, two batteries and a radio transmitter. Since the capsule travels along the digestive tract with the physical peristalsis, the examination of the entire small intestine has become comfortable and painless to the patient. This chapter introduces a novel method for polyp recognition in WCE images based on improved Locality-Constrained Linear Coding (LLC) method and Probabilistic Latent Semantic Analysis (PLSA) model. Results show that the proposed method outperforms existing polyp recognition methods in terms of classification criteria with our current datasets.
