We have been developing a prototype system for the automatic detection of hepatocellular carcinoma (HCC) from whole slide images (WSIs) of liver biopsy based on image analysis techniques. In this paper, we present two color-related topics of this system: color correction and the calculation of color-related features of cytoplasm. A WSI-basis color correction method was implemented for the prototype system. We tested the color correction using more than 300 WISs, and it was confirmed that the color correction works well and stably. In addition, it was found that the success rate of nuclei detection significantly increaseed due to color correction. As for color-related features, we propose a method to calculate the representative color of cytoplasm and the clearness of cytoplasm. It was found that there was slight difference between the distributions of the representative colors of HCC and non-cancer tissues. In addition, there was slight correlation between the clearness of cytoplasm and nuclei density, which implies a promising role of the clearness index in a nuclei-based HCC detection.
Yuri Murakami, Tokiya Abe, Yoshiko Yamashita, Masahiro Yamaguchi, Masahiro Ishikawa, Akinori Hashiguchi, Kiyuna Tomoharu, Akira Saito, Michiie Sakamoto, "Color processing in pathology image analysis system for liver biopsy" in Proc. IS&T 22nd Color and Imaging Conf., 2014, pp 184 - 188, https://doi.org/10.2352/CIC.2014.22.1.art00032