基于机器学习的低信噪比细胞图像分割

doi:10.3969/j.issn.1674-7968.2021.01.019

摘要
图/表
参考文献
相关文章 (1)

全文: PDF (7400 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要荧光显微图像在生命科学研究中具有重要作用。利用计算机和机器学习可以有效处理大量图像数据,从而得到具有统计意义的结论。本研究针对低信噪比的细胞荧光图像,提出一种基于U-Net的细胞分割模型,并构建了单通道低信噪比细胞图像数据集用于模型训练和测试。本研究提出的模型使用不同尺度的卷积核提取特征,利用残差模块加深网络深度,并使用权重损失机制使机器学习过程更加关注于细胞边缘。相较于其他方法,在低信噪比荧光显微图像的细胞分割上,可以有效解决细胞与背景对比度低、胞内信号亮度分布不均的问题,其像素准确率、IoU(intersection-over-union)可分别达到87.6%和72.0%。本研究为细胞形态学研究、借助图像的高通量细胞筛选提供技术支持。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	阎彧萱
	卢金旺
	宋奕泓
	旷小宇
	田原
	傅静雁

关键词 ：细胞分割, 机器学习, 低信噪比, U-Net

Abstract：Fluorescence microscopy image plays an important role in the research of life sciences. A large number of image data can be processed effectively using computer and machine learning, so that we could obtain the conclusion with statistical significance. In this study, a cell segmentation model was proposed based on U-Net for low signal-to-noise ratio (SNR) fluorescence images, and a cell image dataset for training and validation was constructed. In this method, convolution kernels were adopted to extract features, residual modules to deepen network, and weighted loss function to make machine learning process pay more attention to cell edges. Compared to several other methods, this algorithm shows better performance in cell segmentation for low signal-to-noise ratio fluorescence images with 87.6% pixel accuracy and intersection-over-union (IOU) 72.0%. This study provides technical support for cell morphology research and image-based high-throughput cell screening.

Key words： Cell segmentation Machine learning Low signal-to-noise ratio U-Net

收稿日期: 2020-07-22

ZTFLH:

S-03

基金资助:国家自然科学基金项目(31722032; 31771494; 31970662); 国家级大学生创新创业训练计划项目(201910019040; 202010019014); 中国农业大学生命科学实验教学示范中心探索项目(20190102)

通讯作者: **jingyanfu@cau.edu.cn

作者简介: *同等贡献作者

引用本文:

阎彧萱, 卢金旺, 宋奕泓, 旷小宇, 田原, 傅静雁. 基于机器学习的低信噪比细胞图像分割[J]. 农业生物技术学报, 2021, 29(1): 198-206.
YAN Yu-Xuan, LU Jin-Wang, SONG Yi-Hong, KUANG Xiao-Yu, TIAN Yuan, FU Jing-Yan. Cell Segmentation in Low Signal-to-Noise Ratio Microscopy Images Based on Machine Learning. 农业生物技术学报, 2021, 29(1): 198-206.

链接本文:

http://journal05.magtech.org.cn/Jwk_ny/CN/10.3969/j.issn.1674-7968.2021.01.019 或 http://journal05.magtech.org.cn/Jwk_ny/CN/Y2021/V29/I1/198

[1] 古德费洛I, 本吉奥 Y, 库维尔 A, 主编, 赵申剑, 黎彧君, 符天凡, 李凯, 主译. 2017. 深度学习[M]. 北京: 人民邮电出版社, pp. 204. (Goodfellow I, Bengio Y, Courville A, eds., Zhao S J, Li Y J, Fu T F, Li K, trans. 2017. Deep Learning[M]. Posts & Telecom Press, Beijing, China, pp. 204.)
[2] Akram S U, Kannala J, Eklund Let al.2016. Cell Segmentation proposal network for microscopy image analysis[C]. // Deep Learning and Data Labeling for Medical Applications. Springer, Cham, pp. 21-29.
[3] Aly A A, Bin Deris S, Zaki N.2011. Research review for digital image segmentation techniques[J]. International Journal of Computer Science and Information Technology, 3(5): 99-106.
[4] Badrinarayanan V, Kendall A, Cipolla R.2017. SegNet: A deep convolutional encoder-decoder architecture for image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(12): 2481-2495.
[5] Dzhindzhev N S, Yu Q D, Weiskopf Ket al.2010. Asterless is a scaffold for the onset of centriole assembly[J]. Nature, 467(7316): 714-718.
[6] Falk T, Mai D, Bensch Ret al.2018. U-Net: Deep learning for cell counting, detection, and morphometry[J]. Nature Methods, 16(1): 67-70.
[7] Goldys E M.2009. Fluorescence Applications in Biotechnology and Life Sciences[M]. John Wiley & Sons, Hoboken, N.J., pp. 269-288.
[8] He K, Zhang X, Ren Set al.2016. Deep residual learning for image recognition[C]. // 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, Las Vegas, pp. 770-778.
[9] Ibtehaz N, Rahman M S.2020. Multiresunet: Rethinking the U-Net Architecture for multimodal biomedical image segmentation[J]. Neural Networks, 121: 74-87.
[10] Long J, Shelhamer E, Darrell T.2015. Fully Convolutional networks for semantic segmentation[C]. // 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, Boston, pp. 3431-3440.
[11] Quan T M, Hildebrand D G C, Jeong W-K.2016. FusionNet: A deep fully residual convolutional neural network for image segmentation in connectomics[J/OL]. arXiv:1612.05360.
[12] Rejintal A, Aswini N.2016. Image processing based leukemia cancer cell detection[C]. // 2016 IEEE International Conference on Recent Trends in Electronics, Information & Communication Technology (RTEICT). IEEE, Bangalore, pp. 471-474.
[13] Ronneberger O, Fischer P, Brox T.2015. U-Net: Convolutional networks for biomedical image segmentation[C]. // Lecture Notes in Computer Science. Springer, Cham, pp. 234-241.
[14] Szegedy C, Vanhoucke V, Ioffe Set al.2016. Rethinking the inception architecture for computer vision[C]. // 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, Las Vegas, pp. 2818-2826.
[15] Taha A A, Hanbury A.2015. Metrics for evaluating 3d medical image segmentation: Analysis, selection, and tool[J]. BMC Med Imaging, 15(1): 29.
[16] Thomas R M, John J.2017. A review on cell detection and segmentation in microscopic images[C]. // 2017 International Conference on Circuit, Power and Computing Technologies (ICCPCT). IEEE, Kollam, pp. 1-5.
[17] Wu J, Zeng P, Zhou Yet al.2006. A novel color image segmentation method and its application to white blood cell image analysis[C].// 2006 8th international Conference on Signal Processing. IEEE, Beijing, doi: 10.1109/ICOSP.2006.345700.
[18] Zhao H, Shi J, Qi Xet al.2017. Pyramid scene parsing network[C]. // 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, Honolulu, pp. 2881-2890.