复杂背景下的宫颈细胞核分割方法  

The Method of Segmentation of Cervical Nuclei in Complex Background

作　　者：赵晶[1] 梁隆恺 何勇军[1] 谢怡宁[1] ZHAO Jing;LIANG Long-kai;HE Yong-jun;XIE Yi-ning(Harbin University of Science and Technology,School of Computer Science and Technology,Harbin 150080,China)

机构地区：[1]哈尔滨理工大学计算机科学与技术学院

出　　处：《哈尔滨理工大学学报》2019年第3期22-28,共7页Journal of Harbin University of Science and Technology

基　　金：黑龙江省教育厅科学技术研究项目(12511096);国家自然科学基金面上项目(61673142);哈尔滨市杰出青年人才基金(2017RAYXJ0013)

摘　　要：自动阅片技术采用图像处理方法,在细胞核识别的基础上实现对细胞核DNA含量的准确测量,能够为医生诊断提供辅助。图像分割作为自动阅片系统关键直接影响系统性能。然而,显微镜下细胞核图片存在光照不均、背景阴影等情况。并且不可避免地存在一些血细胞、淋巴细胞、垃圾杂质、成团细胞,这严重降低了分割算法的性能。为解决上述问题,提出了复杂背景下的宫颈细胞核分割方法。该方法首先利用参数自适应的局部阈值法来分割图片,并利用自适应的分割参数调节来处理图片中的复杂情况。即利用局部阈值窗口大小和二值化后轮廓数量的函数关系来自动调节窗口大小。然后根据得到的二值图确定分水岭算法的标记图像,最后利用分水岭算法完成整张图片的分割。实验表明,此方法能适应复杂图像环境,并可以从成团细胞中分割出重叠程度较低的细胞,提高分类器识别细胞核的概率。Automatic screening technology developed in recent years.It applies image processing,and first recognizes nucleus and then measures the DNA contents accurately,so it can provide auxiliary for a doctor′s diagnosis.Image segmentation is the key technique of automatic screening system which directly determines the performance of the systems.However,the imaging environments under the microscope are complex.One the one hand,uneven illumination,background shading and uneven dyed nucleus exist.On the other hand,there are inevitably blood cells,lymphocytes,garbage,impurities and conglobation cells in cell images.These conditions degrade the performance of image segmentation.In order to solve these problems,we put forward a method to segment cervical nuclei in complex background.This method first employs the local threshold method to segment images.In this procedure we propose a parameter adapting method which adjusts its parameters automatically according to the function of local threshold window size and the binarized outline number.The local threshold method transforms an image into a binary image which is then passed to image corrosion operator to generate a marking image.With the binary image,the watershed algorithm was finally performed to segment the image.The experiment shows that the method can adapt to the complex image environment and separate the cells with lower overlapping nuclei images.

关 键 词：细胞核分割 参数自适应 分水岭算法 局部阈值法 

分 类 号：TP315[自动化与计算机技术—计算机软件与理论]