Detection of Diseases in Intestinal Lumen by Using SVM - Classifier Algorithm
The wireless capsule endoscopy is an emerging technology which is used to detect the abnormalities in Gastrointestinal (GI) tract (i.e. small intestine and large intestine). A endoscopic video consists of more than 57000 images and it is very difficult for physicians to examine the intestinal diseases like ulcer, cancer, bleeding, tumor out of thousand endoscopic images makes the task very difficult and expensive. Our goal is to develop a detection method by using support vector machine classifier. SVM is a non-linear classification which performs efficiently using kernel trick and mapping their inputs into high-dimensional feature spaces, which showed good results in the medical diagnostics and other fields. First, frames presenting intestinal images are detected by an SVM classiÃ¯Â¬Âer using textural and color information. Secondly, intestinal images are segmented into regions. We show a detailed validation using a large dataset and there may not be large differences in accuracy, there is difference between them in complexity by using SVM algorithm. This paper shows that SVM can be very efficient and yield high accuracy rate of 98.6% and precision of 95.5%.