# 基于区域生长的细胞生长汇合度计算Calculation of Cell Growth Confluence Based on Region growth

DOI: 10.12677/CSA.2020.102025, PDF, HTML, XML, 下载: 268  浏览: 854

Abstract: Cell growth confluence is the main parameter in the in vitro culture of cells, which has an im-portant impact on subsequent subculture. The main method for calculating the cell growth con-vergence degree by manual observation is judged according to the empirical value, and the result has a large error, which is not suitable for large-scale mass production. The development of digital image processing is largely based on the calculation of cell growth convergence to improve efficiency and objective accuracy. To this end, this paper proposes a method for calculating the cell growth convergence based on region growth. First, the improved Gauss-Laplacian operator is used to enhance the contrast of the image preprocessing; secondly, the mean shift algorithm is used to separate the cell region from the non-cell region; finally, the mathematical region is used to filter the blank region with smaller area. In turn, the effective separation of the cell region from the background is achieved, and the growth confluence of the cells is obtained accurately. In this paper, the algorithm is applied to the image of cells collected in the laboratory and the network, and the effectiveness of the algorithm is verified by experiments, which effectively replaces the traditional artificial method for cell growth convergence calculation.

1. 引言

2. 相关研究

3. 算法概述

1) 通过滤波先消除噪声影响，再锐化边缘，开始默认进行两次拉普拉斯锐化处理；

2) 采用均值漂移算法对预处理的图像进行粗分，将颜色相似的区域合并(对图像多维度数据颜色值(RGB)与空间位置 $\left(x,y\right)$ ，有颜色半径cr和空间半径sr。对于图像内任一点为中心，sr为半径区域内的所有点，先找到RGB值在cr范围内的点作为样本点，通过计算样本点与中心点的向量均值，作为新的中心，移动直至收敛。此时之前所有的样本点都会连通并且像素值等于该收敛点的值)，并用漫水填充算法填充各个类区域，分离色彩相近的不同区域；

3) 通过并查集的数据结构存取各个类区域，通过颜色、大小等属性提取出背景区域；

4) 形态学运算过滤较小面积区域；

5) 计算得到细胞区域面积占图像大小比例如果大于75%，输出结果，否则重新从步骤一开始计算，此时只需进行一次拉普拉斯变化，最后直接输出结果。

3.1. 图像增强预处理

Figure 1. The main flowchart of our proposed algorithm

(1) 用一个 $G\left(x,y\right)$ 取样的3 × 3的高斯低通滤波器对输入图像 $f\left(x,y\right)$ 进行滤波平滑；

(2) 用拉普拉斯滤波对步骤(1)得到的图像进行图像增强，本文针对细胞图像多次实验发现对于细胞区域占整幅图像面积75%及以上的图像需要进行两次拉普拉斯锐化计算，对于低于75%图像占比的只需进行一次锐化计算。

3.2. 分离非细胞区域

Figure 2. Comparison of cell images before and after image enhancement

1) 选定种子点；

2) 检查种子点颜色，如果该点颜色与边界色和填充色均不同，则用填充色填充，否则不填充；

3) 检查种子八邻域位置，重复步骤2，直到遍历所有像素点。

Figure 3. Preliminary segmentation of the cell background

1) 在RGB颜色空间下，设k为当前像素为图像的第几个像素，将预处理好的图像的每一个像素记录在一个二维维数组id中，赋value值，value表示当前像素的父节点像素值，此时value等于k，表示父节点是本身；

2) 对图像顺序扫描，对像素k，遍历其四邻域，若RGB值相同，则合并，在id中记录这两个像素点的值为k；

3) 遍历id，将结果记录在一维数组index，其中index[k]表示第k个位置的像素点作为父节点包含子节点的个数，也就是连通区域的面积；

4) 最后将连通区域面积大于阈值的区域设置为背景。

$par=\left\{\begin{array}{l}3,\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}rate<0.7\\ 5,\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}rate\ge 0.7\end{array}$ (1)

Figure 4. The resulting cell background image

4. 实验结果与分析

(a) 汇合度0%~25% (b) 汇合度25%~50% (c) 汇合度50%~75% (d) 汇合度75%~100%

Figure 5. Experimental data set of different confluence

Figure 6. Partial cell growth confluence annotation and estimated distribution map

(a) (b) (c) (d) (e)

Figure 7. Comparison results of Mask-RCNN and this algorithm

5. 结论

 [1] Otsu, N. (2007) A Threshold Selection Method from Gray-Level Histograms. IEEE Transactions on Systems Man & Cybernetics, 9, 62-66. https://doi.org/10.1109/TSMC.1979.4310076 [2] Ali, S. and Madabhushi, A. (2012) An Integrated Region-, Boundary-, Shape-Based Active Contour for Multiple Object Overlap Resolution in Histological Imagery. IEEE Transactions on Medical Imaging, 31, 1448-1460. https://doi.org/10.1109/TMI.2012.2190089 [3] Liu, B., Jia, X., Jiang, Q., et al. (2015) A Segmentation System Based on Clustering Method for Pediatric DTI Images. International Journal of Imaging Systems and Technology, 25, 102-113. https://doi.org/10.1002/ima.22126 . [4] Wu, H.S., Barba, J. and Gil, J. (1998) A Parametric Fitting Algorithm for Segmentation of Cell Images. IEEE Transactions on Biomedical Engineering, 45, 400-407. https://doi.org/10.1109/10.661165 [5] Hunter, I.A. and Soraghan, J.J. (1995) Neural Network Edge Detection-Successes and Failures. IEE Colloquium on Applications of Neural Networks to Signal Processing (Digest No. 1994/248), London, 15-15 December 1994, 2/1-2/3. [6] Zhang, Y. and Cheng, X. (2010) Medical Image Segmentation Based on Watershed and Graph Theory. 2010 3rd International Congress on Image and Signal Processing, Yantai, 16-18 October 2010, 1419-1422. https://doi.org/10.1109/CISP.2010.5646332 [7] 沙盛中, 翁桂荣, 秦树伟. 基于数学形态学的白细胞分割及提取[C]//中国自动化学会, 中南大学, 湖南省自动 化学会. 第26届中国控制会议, 2007: 497-499. [8] 刘锡民, 白丽华, 张庆刚, 许炳章, 邢岳林. 彩色切片细胞的识别与计数研究[J]. 山东师范大学学报(自然科学版), 18(2): 32-34. [9] 杨小青, 杨秋翔, 杨剑. 基于形态学的显微细胞图像处理与应用[J]. 计算机系统应用, 2016, 25(3): 222-226. [10] 蒋秋霖, 王昕. 提出了基于区域生长算法的脑肿瘤图像分割[J]. 长春工业大学学报, 2018(5): 490-493. [11] 潘家辉, 朱玲利, 鲍苏苏. 提出的基于区域生长算法的CT序列图像分割[J]. 洛阳师范学院学报, 2015(5): 58-61. [12] 刘应乾, 曹茂永. 基于Gabor滤波与区域生长的细胞分割[J]. 山东科技大学学报(自然科学版), 2012, 31(2): 99-103. [13] 白华, 张凤凤, 张成, 赵军发. 间充质干细胞生长汇合度自动分析算法研究[J]. 中国细胞生物学学报, 2017, 39(11): 72-79.