Intelligent System for Diagnosis of Erythema Migrans

Abstract

Lyme borreliosis is the most common human tick-borne infectious disease of the Northern Hemisphere. One of the first signs of the disease is erythema migrans, a skin lesion that appears within days to weeks after an infected tick bite. In this article, a novel intelligent system for erythema migrans recognition is presented based on image and text information, applicable for individual and clinical web-based use. Novelties of our approach include a combination of visual and textual attributes, a new combination of visual attributes (geometrical, color, and Gabor-filter based), and a new algorithm for calculation of color-based attributes. Procedurally, the intelligent system for erythema migrans recognition integrated in a web-based application facilitates provisional diagnosis of erythema migrans in the general population and assists general medical practitioners in their decisions. Several classification methods—Naïve Bayes, Support Vector Machine, Adaboost, and Random forest—were tested in order to achieve improved performance.