Sri Sivasubramaniya Nadar College of Engineering, Anna University, Chennai, India.
Mepco Schlenk Engineering College, Anna University, Sivakasi, India.
J Digit Imaging. 2023 Feb;36(1):231-239. doi: 10.1007/s10278-022-00688-7. Epub 2022 Aug 2.
In this paper, we describe an algorithm for accurately segmenting multiple nudfclei from clumps of non-overlapping immuno-histochemically stained histological hepatic (liver) images. This problem is notoriously difficult because of the degree of presence of stains among the multi-nucleated cells, the poor contrast of cell cytoplasm, and the presence of mucus, blood, and inflammatory cells in the images. Hepatocellular carcinoma, characterized by cellular and nuclear enlargement, nuclear pleomorphism, and multi-nucleation, poses a prominent threat. Our proposed method addresses the aforementioned issues for an automated diagnosis system by judging the presence of multiple nuclei in a two-step process: the Quickhull algorithm defines the convex hull of each cell in the image and candidate nuclei regions are located with morphological operations. A combination of features containing local minima and shape-dependent features is extracted for the detection of single or multiple nuclei in each cell with a significant reduction in the number of false positives and false negatives providing an accuracy of 89.76%.
本文描述了一种从免疫组织化学染色的肝组织（肝脏）图像的非重叠团块中准确分割多个核的算法。由于多核细胞中染色程度、细胞质对比度差以及图像中存在粘液、血液和炎症细胞，因此这个问题非常棘手。肝细胞癌以细胞和核增大、核多形性和多核化为特征，构成了突出的威胁。我们提出的方法通过分两步过程判断图像中是否存在多个核，解决了上述问题，为自动诊断系统提供了帮助：Quickhull 算法定义了图像中每个细胞的凸包，然后通过形态学操作定位候选核区域。组合使用包含局部最小值和形状相关特征的特征，用于检测每个细胞中的单个或多个核，显著减少了假阳性和假阴性的数量，准确率达到 89.76%。
