摘要
显微细胞分割的精度直接影响疾病的判别诊断,特别在宫颈细胞的显微病理图像中,细胞核的形态大小、与细胞质之间的比例参数等对于病情的良恶诊断具有重大的意义.为提高宫颈细胞核质分割的精度,提出一种基于卷积神经网络的医学宫颈细胞图像的语义分割方法.标定宫颈细胞显微图像中的细胞核和细胞质轮廓,制作基于长沙市第二人民医院的基于新柏氏液基细胞学检测TCT(Thinprep cytologic test)制片技术的宫颈TCT细胞涂片的CCTCT数据集;通过卷积神经网络对核质分割模型进行训练,避免人工提取特征;通过反卷积达到图像的语义分割.实验结果表明,该算法在宫颈细胞的显微病理图像中的核质分割准确率高达94.7%,具有很高的鲁棒性和适应性.
The accuracy of microscopic cell segmentation directly affects the differential diagnosis of diseases.Especially in the microscopic pathological images of cervical cells,the shape and size of nucleus and the ratio parameters between cytoplasm are of great significance for the diagnosis of disease.In order to improve the accuracy of cervical nucleus segmentation,this paper proposed a semantic segmentation method of medical cervical cell images based on convolutional neural network.We calibrated the nuclear and cytoplasm contours of cervical cells in microscopic images,and produced CCTCT data set of cervical TCT cell smears based on TCT technology in Second People s Hospital of Changsha City.Then,the nuclear and cytoplasm segmentation model was trained by convolution neural network to avoid manual feature extraction.And the semantics of image was segmented by deconvolution.The experimental results show that the our algorithm has an accuracy of 94.7%in the microscopic pathological images of cervical cells,and the algorithm has high robustness and adaptability.
作者
李智能
刘任任
梁光明
Li Zhineng;Liu Renren;Liang Guangming(College of Information Engineering,Xiangtan University,Xiangtan 411105,Hunan,China;School of Computer,National University of Defense Technology,Changsha 410000,Hunan,China)
出处
《计算机应用与软件》
北大核心
2019年第11期152-156,共5页
Computer Applications and Software
基金
国家自然科学基金项目(60673193)
关键词
语义分割
卷积神经网络
核质分割
宫颈细胞显微图像
Semantic segmentation
Convolutional neural network
Nuclear and cytoplasm segmentation
Cervical cell microscopy
