抄録
(Introduction) Frequency of use of the endoscope increased significantly in medical field. Then this increasing amount of data have made burden on endoscopists. The new diagnostic equipment such as the capsule endoscope will make amount of data for diagnosis increase be more huge. (Purpose) The endoscopists diagnose by the information of morphological and color change in every two-dimensional endoscopic images. So we propose the support system diagnosis to reduce this burden. Concretely we have been trying to make the system which can automatically extract a disease area. (Method) As for the endoscopic image, the color information such as blood vessels is made superposition onto the main red mucosa color. And the special structure and color of disease are added to image. The color information is extracted from the virtual spectral image method and structural information from the statistical processing. (Results) It was an automatic diagnosis positive rate of approximately more than 70% in about 100 disease lesion images. We will study the image processing for automatic diagnosis positive rate of more than 80%.
