ABSTRACT
Traditional endoscopies have dominated the field of the gastro-intestinal tract investigation for decades. Nevertheless, since its emergence, the wireless capsule endoscopy imaging technique has gained more popularity as it can visualise the entire gastro-intestinal tract including the small bowel. However, this technique produces immense amount of images which have to be reviewed 92by the physicians. This task is exhaustive and time consuming. This chapter presents a novel approach adopted for colon red lesion and ulcer abnormalities detection using images issued by WCE. A simple Convolutional Neural Network architecture is proposed. Besides, we incorporated Parametric Recti-fied Nonlinear Unit (PRenu) activation function. Extensive experiments have been conducted on two datasets in order to show the efficiency of the proposed method.
