Computerized digital image analysis: an aid for melanoma diagnosis--preliminary investigations and brief review

Abstract

Both digital imaging and epiluminescence microscopy hold promise for improved early detection of cutaneous melanoma. Several centers have been actively working in these areas during the past decade. These experiences and preliminary work based on the image capture of 83 pigmented lesions at our center using a prototype digital imaging system (SKINVIEW) are described. This system is based, in part, on the analysis of lesional morphologic features, such as shape, border, and radii. Histopathologic correlation was matched against these features to assess the efficacy of diagnosis. At our center, these parameters alone were not sufficient to discriminate between benign and malignant lesions, in part, because the melanomas were, in general, early lesions and many of the nevi were sufficiently clinically atypical to require removal for discrimination from melanoma. In addition, technical improvements in the image capturing and processing mechanism are needed. Rapid progress in this area is anticipated.