Abstract

Reflectance spectroscopy is utilized for investigation of pigmented skin lesions, making it possible to differentiate between melanoma and benign pigmented skin areas. However, normal skin spectra are presented in a wide region of variations because of differences related to the skin pigmentation. The reflectance spectra minima at 540 and 575 nm, related to haemoglobin absorption are clearly defined for the all skin types but for the more pigmented areas they are covered by epidennal melanin absorption.¹ The spectral deviations complicate the automation of the diagnostics based on spectral distributions. The automated diagnostic of the skin diseases existed uses analysis of computer images derived from the suspicious lesions, but it should be mentioned that the instruments work with interactive selection of the analyzed image segments. Development of instruments with a higher degree of automation supposes possibilities for object classification based on more than one independent methods and comparison between the results on different stages of processing. Therefore, the aim of this work is a development of a method for automatic estimation of the skin lesions using the calculated colorimetric data obtained from reflectance spectra measured.

© 2007 IEEE