Skin Disorder Diagnosis Assisted by Lesion Color Adaptation

A technique that counterbalances the variations in the color features of skin disorder lesions is tested in the framework of a skin disease diagnosis application. This application attempts to recognize a skin disorder by photographs displaying a body part with normal skin and lesion. These color feature variations are caused by different lighting condition or different settings used during the analysis of the photograph. Although the color features of the lesions depend on the stage of the disease, the color of the patient, etc, a significant accuracy improvement can be achieved if a simple color adaptation proposed in this paper is also taken into account. A sensitivity improvement of up to 150% can be measured (e.g., from 45% to 68% in Papillomas) in most of the skin disorders that have been tested. The Skin Disease application that employs the proposed color adaptation technique can be implemented on any low cost smart phone. It can serve as a diagnosis assistant for dermatologists or a tool for remote monitoring of skin disorders. An important advantage of the Skin Disease application is that it allows the end user to customize the disorder recognition rules or extend the supported set of diseases.