Automatic identification of Discoid Lupus Erythematosus

We propose in this paper, an algorithm which carries out the identification of Discoid Lupus Erythematosus (DLE) and classifies it into one of two possible phases, the active and cicatrisation (scarring) phases. The images used in this study are provided by the Department of Dermatology, University of Campinas, Unicamp. In the pre-processing step, those raw images were segmented by applying the K-means Clustering algorithm [13] in order to separate the area of interest (skin, lesion and scar). 3 different classifiers were considered in this study, namely KNN — K-Nearest Neiboard, QDC — Quadratic Discriminant Classifier and UDC — Uncorrelated Discriminant Classifier. The best performance is obtained from using the QDC classifier. Therefore, the QDC classifier was used to perform the clinical patient monitoring. Some images were altered artificially by a graphic software, simulating an increase (degradation) and decrease (improvement) of the lesion.