摘要
目的针对肿瘤FISH图像细胞边缘模糊和粘连等问题,提出一种基于自动随机游走的细胞检出及HER2基因状态判别方法。方法首先,在RGB颜色空间中的分割细胞、信号点等区域;然后,改进极限腐蚀算法,自动获取细胞种子区域,并提取有效种子点进行自动随机游走分割,较好地实现了FISH细胞的检出;最后,依据细胞内红绿信号点数比值对HER2基因状态作出判别。结论实验结果表明,本文方法细胞检出效果理想,能较好地实现粘连、重叠细胞的分离,FISHHER2基因状态判别敏感度、特异性、准确率均达到90%以上。
Objective To solve the issues such as blur and adhesion of cell edges in FISH-stained tumor cell images, a random walk-based detection method of HER2 status from FISH-stained ceils is proposed. Methods First, segmentation of cells and signal points in the RGB color model was performed. Then, to improve the ultra-corrosion algorithm to automatically obtain cell seed region and extract the effective seeds for automatic random walk segmentation by fine FISH-stained cell detection. Finally, to detect the HER2 gene status according to ratio of red and green signal points in the cells. Results Experimental results show that detection results of the proposed method are satisfactory. This method can achieve better segmentation of adhered and overlapping ceils. The sensitivity, specificity and accuracy of FISH-HER2 gene status detection reach higher than 90% , respectively.
出处
《中国体视学与图像分析》
2013年第4期318-325,共8页
Chinese Journal of Stereology and Image Analysis
基金
福建省科技重点项目(2011Y0040)
福建省自然科学基金(2009J01283)
