Analysis of spectral reflectance of mucous membrane for endoscopic diagnosis

摘要

Using color of mucous membrane is extremely important in endoscopic diagnosis. We have developed an endoscopic spectroscopy system (ESS) to clarify spectral characteristics of gastric and colon cancer, and a large amount of spectral reflectance data of mucous membrane has been measured by ESS. This paper proposes a method to find the spectral features of lesions considering the light scattering phenomena. We suppose that the influence of mucous membrane, i.e., tumor mainly appear in the spectral data of short wavelength, because the tumor of early cancer is usually present in superficial cell where short wavelength light is scattered more strongly than long wavelength light. To reveal the features in the short wavelength components, the spectral reflectance is divided by the same reflectance of a certain long wavelength. Then, the effect of individual variation on the mucosal color is reduced by subtracting the reflectance of normal region from that of adjacent tumor region. We apply principal component analysis and statistical discriminant analysis to the subtracted spectral reflectance data; the difference between the tumor and normal tissue is revealed.