Differentiation of Pigmented Skin Lesions Based on Digital Processing of Optical Images

Abstract

A method for differentiation of pigmented skin lesions is proposed that is based on digital processing of optical images. The optical images are detected using a digital camera and a standard for its color and spatial calibration. The method involves segmentation and automatic recognition of neoplasms on the basis of five parameters: diameter, area, color, shape, and border sharpness. The analysis reveals clinical features of the lesion, required for diagnosis, and allows for calculating the probability of its malignant transformation. The method proposed was tested on 360 digital images of in vivo skin pigment lesions, including ordinary dysplastic nevi and melanoma. The sensitivity and specificity of the method proved to be 97 and 95%, respectively.