Blue-white veil and dark-red patch of pigment pattern recognition in dermoscopic images using machine-learning techniques

Abstract

The proposed work presents a method for the computer-aided detection in dermoscopic images of two of the most significant patterns in the diagnosis of melanoma, the blue-white veil (an irregular, structureless area of confluent blue pigmentation with an overlying white “ground-glass” haze) and dark-red patch of pigment. The development has been made with the help of supervised machine learning techniques, in a two steps process: firstly, obtaining the conditions that must satisfy the pixels, and secondly, obtaining the features that the image of the skin lesion should have, in relation to the region and the image itself. Tested over a database of 887 images, it has been obtained a results of 89.06% correctly detected. Moreover, as a part of the proposed method itself (derived from the close relationship with the blue-white veil) has been developed a method for obtaining the pixel rules for the dark-red patch of pigment pattern (“a patch of dark red pigmentation”) recognition. Tested over 80 images, it has been obtained a results of 95.14% correctly detected.