Arihiro Koji, Oda Miyo, Ogawa Katsunari, Tominaga Kenshi, Kaneko Yoshie, Shimizu Tomomi, Ohnishi Shiho, Oda Megumi, Kurita Yuki, Taira Yuko, Fujii Masayoshi, Tanaka Maiko
Department of Anatomical Pathology, Hiroshima University Hospital, Hiroshima 734-8551, Japan.
Exp Ther Med. 2010 Nov;1(6):927-932. doi: 10.3892/etm.2010.142. Epub 2010 Aug 26.
The aim of this study was to compare the results of immunohistochemistry (IHC) assays evaluated by human examiners with the results evaluated by computerized image analysis, and to compare the computerized image analysis results among three automated IHC assays, namely the BioGenex, Dako and Ventana assays. All slides were semiquantitatively evaluated according to the Allred score and J-score by human examiners. The images were analyzed using MacSCOPE version 2.6 for Macintosh according to the H-score and the percentage of positive-stained nuclei per area of carcinoma cells (PP) irrespective of the intensity of the stained nuclei. The H-score for the estrogen receptor (ER) was significantly correlated with the Allred score (P<0.0001) and the PP for the ER was significantly correlated with the J-score (P<0.0001), suggesting that the image analysis used in the present study is a useful method for the evaluation of ER status. Several discrepancies were identified between the Allred score and H-score and between the PP and J-score due to the positive-stained cytoplasm area of carcinoma cells and/ or the positive-stained nuclei area of non-carcinoma cells, including benign epithelial cells, lymphocytes and stromal cells. Accordingly, advances in the algorithm of the digitized analyzing system is necessary.
本研究的目的是比较由人工检查者评估的免疫组织化学（IHC）检测结果与由计算机图像分析评估的结果，并比较三种自动化 IHC 检测方法（即 BioGenex、Dako 和 Ventana 检测方法）的计算机图像分析结果。所有切片均由人工检查者根据 Allred 评分和 J 评分进行半定量评估。使用适用于 Macintosh 的 MacSCOPE 2.6 版本，根据 H 评分和每癌细胞面积的阳性染色细胞核百分比（PP）对图像进行分析，而不考虑染色细胞核的强度。雌激素受体（ER）的 H 评分与 Allred 评分显著相关（P<0.0001），ER 的 PP 与 J 评分显著相关（P<0.0001），这表明本研究中使用的图像分析是评估 ER 状态的一种有用方法。由于癌细胞的阳性染色细胞质区域和/或非癌细胞（包括良性上皮细胞、淋巴细胞和基质细胞）的阳性染色细胞核区域，在 Allred 评分与 H 评分之间以及 PP 与 J 评分之间发现了一些差异。因此，数字化分析系统算法的改进是必要的。
