基于UNet网络的乳腺癌肿瘤细胞图像分割

UNet⁃based image segmentation of breast cancer tumor cells

乳腺癌肿瘤细胞(MCF-7)的研究对乳腺癌的诊断和治疗具有重要意义。相比其他的语义分割模型,UNet网络在医学影像领域具有更加优秀的表现。为了将人工智能技术用于辅助诊断,该文结合深度学习和卷积神经网络理论,搭建了基于UNet卷积神经网络的乳腺癌肿瘤细胞分割模型。该文使用CMOS相机采集混有人体红细胞的乳腺癌肿瘤细胞图像,通过labelme软件对采集的细胞图像进行轮廓标注等处理,提取出细胞区域,使用UNet神经网络训练并测试。结果表明,肿瘤细胞图像分割准确率达到91%,精准率达到89%。

The study of breast cancer tumor cells(MCF-7)is of great significance for the diagnosis and treatment of breast cancer.Compared with the traditional semantic segmentation algorithm,UNet has a better performance in the field of medical image.In order to use artificial intelligence to assist diagnosis,this paper combined deep learning and convolutional neural network theory to build a breast cancer tumor cell segmentation model based on UNet convolutional neural network.CMOS camera was used to collect the images of breast cancer tumor cells mixed with human red blood cells.Labelme software was used to make contour marking on the collected cell images,and the cell regions were extracted.The UNet network was used for training and testing and the results show that the accuracy of tumor cell image segmentation is 91%and the precision is 89%.