Prediction of esophagitis and z-line from wireless capsule endoscopy images using fusion of low-level features

Abstract

Wireless endoscopy capsule (WCE) pictures areoften used to recognize digestive tract illnesses as they enable immediate GI tract perception. In any case, it requires a clinician's long-lasting review because of an incredible number of pictures delivered by the system. In this manner, it might be valuable to devise an Automatic detection framework to enable clinicians to distinguish abnormal pictures. In this work, it is endeavour to plan an electronic plan intending to distinguish esophagitis in WCE pictures. The Esophagitis in WCE pictures show extraordinary variations in appearance. Scale-Invariant Feature Transform (SIFT) and Auto Color Correlogram (ACC) are two features that are used to coordinate the texture, color and shape characteristics collected from points of interest. Using Naïve Bayes, Support Vector Machine (SVM) and Random Forest, we assessed the performance with comprehensive tests on our current picture information consisting of 100 normal-z-line WCE pictures and 100 esophagitis. From the experimental analysis, it is promising to use the proposed plan to distinguish esophagitis and normal-z-line from WCE pictures.