一种综合有效的癌细胞图象分割方法

该文综合两种常用的分割算法,即阈值法和边缘检测法的优点,介绍了一种针对食管癌细胞的形状提取方法。对该方法提出一种基于直方图势函数的多阈值分割方法,分割时细胞及细胞核部分不变而将它们以外的部分设为背景,保留了核的区域性质同时去除了噪声,以便于边缘检测;采用Canny算子,结合预分割结果进行边缘检测,以获得细胞尤其是细胞核的边缘形状特征,为癌细胞的诊断工作提供好的依据。将该分割方法应用于实际的医学图象分割的实验中,证明了该方法的有效性。

A Synthetical and Effective Segmentation for Cancer Cell Image

Uniting the advantages of two simple segmentation methods,namely thresholding and edge detection,this paper introduces a new method to distill esophagus cancer cell shape character.This method proposes a multi-threshold method based on histogram potential information to realize initial segmentation,adopts Canny operator to detect edge of multi-threshold segmentation result.By this,people can obtain the shapes of nucleolus and cytoplasm that offer good gist for cancer cell diagnose.This paper applies this method into practical experiment of medical image segmentation and proves its validity.