Trial for computer analysis of normal hepatic tissue and hepatocellular carcinoma tissue pictures
Kanno, T.; Yamamoto, S.; Shimazaki, M.
Osaka City Medical Journal 29(2): 105-143
1983
ISSN/ISBN: 0030-6096 PMID: 6091013 Document Number: 204058
To analyze the tissue pictures of the normal liver and hepatocellular carcinoma (HCC), the gray-tone on the surface of microscopic photographs prepared from hematoxylin-eosine staining specimens was numerically changed. Using this, several parameters expressing some brightness distribution patterns were decided and analyses with computer were tried. And these parameter values were comparatively examined among the 3 of the normal liver, well differentiated HCC and poorly differentiated HCC. Several of the selected parameters could distinguish the histologic image between normal and tumorous livers, and especially some of them, such as contrast and correlation, were useful even for discrimination between carcinomas of different degrees of differentiation. A graphic display in 2 dimensions of maltivariate image data sets, each representing histologic images, was undertaken using the 1st and 2nd factor scores to provide for a easy understanding of what groupings exist among these histologic pictures. Five groups of hepatocellular carcinomas, as noted in this factor plane have each some characteristic clinical features, including biological aggressiveness, which gives a good promise to these procedures for numeric evaluation of histologic images.