A New Approach Based on Local Binary Patterns Histogram and Fourier Descriptors as a Support Tool in Presumptive Diagnosis of Gastritis

Abstract

According to the World Health Organization (WHO) in developing countries around 23% of malignancies are caused by infectious agents. The infection with Helicobac-ter pylori (H. pylori) bacteria seems to be a major cause of stomach cancer (80%). An early diagnosis of gastritis may help decrease the risk of gastric cancer and other complications such as gastric ulcers. The aim of this paper is to evaluate the probability of providing specialists a diagnostic support tool based on computer vision. We used twenty-four endoscopic images of healthy patients and thirty-five images of patients suffering from gastritis to perform an automatic classification process. The suggested approach uses Local Binary Patterns (LBP), texture descriptors, and certain classifiers to perform the automatic classification. The results are promising and show 83% precision in identifying the disease.