Processing algorithms for the analysis of videocolonoscopy images

The cancer is the second cause of death in Venezuela. In Lara State, the malign tumors of the digest organs, including the colorectal area, represent the 4.13% of the registered death. The specialists ensure that early detection of this disease increases the probability of healing. This paper presents a collection of algorithms developed for the automatic analysis of endoscopic color images of tissues of the colon. The analysis begins with the preprocessing of the image, then segmentation and features extraction are applied and finally an artificial neural network is used for the classification. In the experiments an architecture of 60x6x1 resulted the most appropriate with accuracy of 97% and 72% in the training and test sets respectively. These results are very important because the classifier is not provided with another previous data of the patient, such as gender, age and medical history.