Masmoudi Hela, Hewitt Stephen M, Petrick Nicholas, Myers Kyle J, Gavrielides Marios A
Department of Electrical and Computer Engineering, The George Washington University, Washington, DC 20052, USA.
IEEE Trans Med Imaging. 2009 Jun;28(6):916-25. doi: 10.1109/TMI.2009.2012901. Epub 2009 Jan 19.
The expression of the HER-2/neu (HER2) gene, a member of the epidermal growth factor receptor family, has been shown to be a valuable prognostic indicator for breast cancer. However, interobserver variability has been reported in the evaluation of HER2 with immunohistochemistry. It has been suggested that automated computer-based evaluation can provide a consistent and objective evaluation of HER2 expression. In this manuscript, we present an automated method for the quantitative assessment of HER2 using digital microscopy. The method processes microscopy images from tissue slides with a multistage algorithm, including steps of color pixel classification, nuclei segmentation, and cell membrane modeling, and extracts quantitative, continuous measures of cell membrane staining intensity and completeness. A minimum cluster distance classifier merges the features to classify the slides into HER2 categories. An evaluation based on agreement analysis with pathologist-derived HER2 scores, showed good agreement with the provided truth. Agreement varied within the different classes with highest agreement (up to 90%) for positive (3+) slides, and lowest agreement (72%-78%) for equivocal (2+) slides which contained ambiguous scoring. The developed automated method has the potential to be used as a computer aid for the immunohistochemical evaluation of HER2 expression with the objective of increasing observer reproducibility.
HER-2/neu（HER2）基因是表皮生长因子受体家族的成员之一，其表达已被证明是乳腺癌一个有价值的预后指标。然而，据报道，在通过免疫组织化学评估HER2时存在观察者间的变异性。有人提出，基于计算机的自动化评估可以对HER2表达进行一致且客观的评估。在本论文中，我们提出了一种使用数字显微镜对HER2进行定量评估的自动化方法。该方法采用多阶段算法处理组织切片的显微镜图像，包括彩色像素分类、细胞核分割和细胞膜建模步骤，并提取细胞膜染色强度和完整性的定量、连续测量值。一个最小聚类距离分类器合并这些特征，将切片分类为HER2类别。基于与病理学家得出的HER2评分进行一致性分析的评估显示，与提供的真值具有良好的一致性。不同类别之间的一致性有所不同，阳性（3+）切片的一致性最高（高达90%），而包含模糊评分的 equivocal（2+）切片的一致性最低（72%-78%）。所开发的自动化方法有可能用作计算机辅助工具，用于HER2表达的免疫组织化学评估，目的是提高观察者的可重复性。
