一种新的用于宫颈癌粘连细胞图像分割的分水岭算法  

A new modified watershed algorithm for image segmentation in cervical cancer cell adhesion

作　　者：赵英红[1] 周中顺 孙存杰[1,2] 平杰[2] 

机构地区：[1]徐州医学院影像学院,江苏徐州221004 [2]徐州医学院附属医院影像科,江苏徐州221006

出　　处：《军事医学》2014年第12期972-975,共4页Military Medical Sciences

摘　　要：目的探讨一种能有效分割宫颈癌粘连细胞图像的算法,完成各粘连细胞边缘的准确分割。方法利用水平集算法从背景区域中提取目标细胞图像,应用极小值的距离变换算法将图像归一化后与感兴趣区域的梯度图像点乘来抑制无用的梯度信息,然后运用标记分水岭算法对粘连细胞图像进行分割。结果与结论实验结果表明,该算法实现了染色不均的粘连宫颈癌细胞的有效分割,在粘连细胞的边缘建立了较传统的分水岭分割方法更准确的分割线,具有显著的临床应用价值。Objective To investigate an effective algorithm for image segmentation in cervical cancer cell adhesion,which enables accurate segmentation of the contour of adherent cells.Methods The images of target cells were extracted from the background area using level set methods,normalized with minimum values of transformation algorithms,and multiplied by the gradient image points in the region of interest（ROI） to inhibit the undesired gradient information before the images of adherent cells were segmented using labeled watershed algorithm.Results and Conclusion Compared to conventional watershed segmentation methods,this algorithm is not only effective in image segmentation of adherent cervical cancer cells with uneven staining and more accurate segmentation lines established around the contours of adherent cells,but of high clinical value.

关 键 词：宫颈肿瘤 粘连细胞 图像分割 分水岭算法 水平集 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]