Wireless Capsule Endoscopy Image Classification Model to Detect Gastro Intestinal Tract Diseases Using Visual Words Based on Feature Fusion

Abstract

This work emphasis on Gastrointestinal (GI) disease diagnosis using Wireless Capsule Endoscopy (WCE) images. The recent researches in medical field are widely based on severe health illness such as cancer, diabetes, Alzheimer, Parkinson and heart diseases. These researches mainly focus on finding a new diagnosing method after the occurrence of the diseases. Consequently, there is a need to determine a new methodology for early stage diagnosing of diseases. Most diseases such as bleeding, ulcer and tumour in the GI tract can be cured or controlled in their early stages, otherwise it will deteriorate into cancer or some other vital diseases. In this work, the detection of pylorus, polyps, Z-line, cecum, ulcer and esophagitis are carried out by using WCE images. Although, these abnormalities are not harmful, it is pivotal to diagnose in the initial stage in order to avoid the development of severity in future. The WCE provides a number of images per second and hence it is difficult to identify the type of disease from the numerous identical images. Therefore, the classification of GI Tract Diseases from the images requires a new technique for accurate diagnosis. In this paper, features are extracted by Centre Symmetric-Local Binary Pattern (LBP), Auto Color Correlogram and Speeded Up Robust Feature (SURF) and it is further followed by K means clustering, Support Vector Machine (SVM) and Particle Swarm Optimization- Support Vector Machine (PSO-SVM) classification. The efficiency of these techniques is analysed by means of performance. The results obtained shows that the proposed method of PSO-SVM with the combined CS-LBP+ACC+SURF gives better classification results and well suited for particular set of images. In future, a greater number of images have to be processed for accurate analysis as well as to try with some other classifiers.