深度学习在荧光显微镜图像增强中的应用  

Applications of Deep Learning in Fluorescence Microscopy Image Enhancement

作　　者：张彦丽 刘冰钰 孙梦依 李静 王文娟[1,2] ZHANG Yan-li;LIU Bing-yu;SUN Meng-yi;LI Jing;WANG Wen-juan(School of Life Sciences,Tsinghua University,Beijing,100084,China;Technology Center For Protein Sciences,Tsinghua University,Beijing,100084,China;Beijing Pharmaceutical and Health Technology Development Center,Bejing,100035,China)

机构地区：[1]清华大学生命科学学院,北京100084 [2]清华大学蛋白质研究技术中心,北京100084 [3]北京医药健康科技发展中心,北京100035

出　　处：《现代生物医学进展》2025年第2期394-400,共7页Progress in Modern Biomedicine

基　　金：清华大学实验室创新基金项目(53100103524)。

摘　　要：传统荧光显微镜在分辨亚细胞结构方面面临衍射极限的挑战,难以有效分辨小于200纳米的结构,同时图像信噪比也有限。近年来,深度学习技术在显微镜图像处理领域展现出巨大的潜力,通过自动特征提取,不仅可以提高图像分辨率,还能显著改善图像质量。本文系统综述了深度学习技术在荧光显微镜图像分析中的应用,尤其是在图像增强与超分辨率重建方面的最新进展。以卷积神经网络(CNN)、生成对抗网络(GAN)和U-Net等为代表的深度学习模型,已广泛用于分类、分割、目标跟踪及成像系统增强等多个领域。这些技术为低分辨率和低信噪比图像提供了高效的解决方案,使得活细胞和纳米尺度生物过程的动态观测成为可能。未来,深度学习技术有望进一步推动显微镜图像智能化处理,为生物医学研究提供更高效的工具和支持。Traditional fluorescence microscopy faces limitations in resolving subcellular structures smaller than 200 nm due to the diffraction limit and low signal-to-noise ratios.Recent advancements in deep learning have shown remarkable potential in addressing these challenges by automatically extracting features to enhance image resolution and quality.This paper provides a comprehensive review of the applications of deep learning in fluorescence microscopy image analysis,particularly focusing on recent developments in image enhancement and super-resolution reconstruction.Deep learning models,including convolutional neural networks(CNNs),generative adversarial networks(GANs),and U-Net,have been extensively utilized across various tasks such as classification,segmentation,object tracking,and imaging system enhancement.These technologies provide efficient solutions for low-resolution and low-signal-to-noise ratio images,enabling dynamic observations of live cells and nanoscale biological processes.In the future,deep learning is expected to further drive intelligent microscopy image processing,offering more efficient tools and support for biomedical research.

关 键 词：深度学习 荧光显微镜 图像增强 超分辨重建 神经网络 

分 类 号：R3[医药卫生—基础医学] TH742.64[机械工程—光学工程]