An android-based patternrecognition application for the characterization of epidermal melanoma

Abstract

Malignant melanoma is currently one of the leading cancers among white-skinned populations around the world, mainly due to the changes in life styles and the significant increase in ultraviolet radiation. Although the mortality rate due to melanomas was about 70%, forty years ago, nowadays, a survival rate of 70% is claimed, which is attributed to early diagnosis. Hence, early stage detection of melanoma is of major significance for increasing chances of long term survival of affected patients. The most effective method for early detection is skin self-examination, a procedure often underestimated by individuals, resulting in poor prognosis. Therefore, the aim of the present study is to address the need for early and accurate characterization of skin lesions through the design andimplementation of an Android application that enables users identify areas on their skin that may need attention from an expert physician. The proposed application uses the phone’s camera feature to take a picture of a skin lesion/mole, communicates with a remote specialized pattern recognition system, via a set of XML Web Services, and within seconds receives a risk analysis of their uploaded image being a melanoma. The proposed system was trained using an image database from New Zealand Dermatological Society, and was reviewed by an expert dermatologist