摘要
通过对糖尿病视网膜扫描显微镜病理切片的分析,可以发现细胞之间微小的差距难以用肉眼来辨别.纤维层的薄厚变化以及层内细胞核的数量和形态变化更难以用肉眼评估.需要从专业人员的主观分析发展到定量分析,才能够客观评价视网膜疾病早期组织结构的细微变化.细胞核的自动分割是计算机辅助病理图像分析的关键步骤,然而由于数据复杂性和可变性,细胞核的分割存在一定困难.近年来,深度学习在病理图像的分析应用中受到了越来越多的关注.提出了一种基于深度学习的全残差注意力网络对糖尿病视网膜病理图像进行细胞核分割.在全残差卷积网络上结合注意力模块,能捕获全局上下文,强调关键语义特征,增强特征表示,更好的分割细胞核.最后结合糖尿病视网膜病变的细胞核数据,对基于全残差注意力网络的视网膜细胞核分割方法进行了验证,实验结果表明该算法整体性能较好,能有效地从背景中分割出细胞核.为评估糖尿病视网膜病变时的各项病理变化提供条件,进行客观定量的评价.
Through the analysis of diabetic retinal scanning microscope pathological,cells can be found hard to identify with the naked eye.A tiny gap between fiber changes in its thickness and layers in the number of nuclei and morphology change is more difficult to evaluate with the naked eye.Need from professional personnel’s subjective analysis to quantitative analysis,objective evaluation to retinal disease early the subtle changes of organizational structure.The nucleus of the automatic segmentation is the key step in the computer aided pathology image analysis,due to data complexity and variability,however,of the division of the nucleus has certain difficulty.In recent years,Deep learning in pathological image analysis application is more and more attention.This paper proposes a deep learning all residual attention network based cell nuclei in diabetic retinal pathological image segmentation.Combining attention on all residual convolution network module,can capture the global context,emphasizes the key semantic features,enhanced features,said better segmentation nuclei.Finally,the paper combines the nuclei of diabetic retinopathy data,based on the residual attention network of retinal cell nucleus segmentation method is verified,and the experimental results show that the algorithm has good overall performance,It can effectively segment the nucleus from the background and provide conditions for the evaluation of various pathological changes in diabetic retinopathy for objective and quantitative evaluation.
作者
易茗
陈娜
黄卉
张欣
YI Ming;CHEN Na;HUANG Hui;ZHANG Xin(Faculty of Mathematics and Statistics,Hubei University,Wuhan 430062,China;Beijing Duan-Dian Pharmaceutical Research and Development Co.,Ltd,Beijing 100176,China;Institute of Basic and Interdisciplinary Sciences,Beijing Union University,Beijing 100101,China)
出处
《数学的实践与认识》
2021年第10期124-132,共9页
Mathematics in Practice and Theory
基金
国家自然科学基金(61673381)
北京联合大学人才强校优选计划(BPHR2020CZ06)。
