Diagnosis of Skin Lesions from Diffuse Reflectance Spectra Using Computational Algorithms: a Preliminary Study

Abstract

The determination of diffuse reflection spectrum on human skin in the spectral range from 400nm-1000nm using an optical fiber spectrometers is a non-invasive technique widely used to study the optical parameters of this tissue, provides information about the absorption and scattering properties of light that can be employed to study the morphology and physiology of the tissue and to detect and diagnose skin diseases in early stages. In this paper a computational algorithm for the selection of the most important attributes of diffuse reflection spectra of human skin obtained with an experimental system that basically consists of a spectrometer, a white light source and bifurcated fiber optic probe that allows send and collect light. To classify the spectral signal was designed a Matlab2006© graphical interface which use support vector machines and algorithm for selecting attributes that allows to achieve a sensitivity and specificity exceeding 80% and 85% of accuracy in the classification.