A rudimentary system for automatic discrimination among basic skin lesions on the basis of color analysis of video images.

Abstract

Detection of differences in color is necessary for an automatic diagnostic system for skin lesions. The purpose of this study was to examine whether color analysis of an image obtained with a videomicroscope can be used as a key to distinguish basic skin lesions. We carried out quantitative color analyses of nine kinds of natural or artificial skin lesions. After images of the lesion and adjacent normal skin were obtained, the difference in the mean quasiabsorbance (logarithm of the inverse reflectance) values between the two were examined in each color band of red, green, and blue by an image analytic method. We could determine a set of thresholds of quasiabsorbance values to distinguish between basic lesion colors. On the basis of these results we developed a rudimentary system that automatically offers possible diagnoses of a test site. The algorithm that we developed may be of value in an initial screening of skin lesions in an automatic diagnosis system.