Identification of Polyps and Tumors from Endoscopy Images with Deep Belief Network

Abstract

It has been demonstrated that a successful image processing strategy can process real-time endoscopic videos to assist professionals in making critical decisions about cancer patients. An effective diagnostic measure in gastrointestinal tract is endoscopy, which us an endoscope with a camera and a transmitter which sends video frames. Using the existing software, the images are taken from real-time endoscopic videos and fed into MATLAB for image processing. The processed videos' output is then supplied back into the host program. Image processing techniques have been used to identify and improve the visualization of polyps in the gastro intestinal system, which assists practitioners in making decisions. The system's goal is to let the physician or medical practitioner visualize and identify problematic structures such as polyps and bleeding spots during endoscopic operations. This suggested approach is tested on a recorded gastrointestinal dataset that has ten sequence videos of 7894 frames.