深度学习驱动的PAINT成像技术开发及活细胞超分辨成像应用研究

DNA点积累在纳米尺度地形成像（DNA-PAINT）是超分辨成像技术的一种，其基于单分子定位原理，可实现5nm的超高空间分辨率。但该方法时间分辨率有限，难以实现活细胞结构动态超分辨成像。深度学习与成像技术的融合，已根本改变研究人员获取、处理、分析和解析数据的方式。本项目拟建立深度学习驱动的PAINT超分辨成像新方法。通过设计基于框架核酸的新型PAINT成像探针，建立多种亚细胞结构的特异性标记与超分辨成像策略。训练深度神经网络，实现多种低于光学衍射极限的亚细胞结构超分辨成像预测，减少重建图像数据量并提升重建的速度。验证深度神经网络PAINT成像效果，实现活细胞中亚细胞结构的多色、快速超分辨率成像。本项目研究通过跨学科的技术融合，开发新型PAINT标记探针并建立深度学习驱动的超分辨成像新方法，有望突破现有超分辨成像技术的发展瓶颈，在未来生物医学科学问题的研究中有广泛应用。

DNA points accumulation for imaging in nanoscale topography (DNA-PAINT) is a super-resolution microscopy technique. This method is based on single-molecule-localization and can achieve 5 nm ultra-high spatial resolution. However, DNA-PAINT technique has low temporal resolution and is difficult to achieve dynamic super-resolution imaging for living cell subcellular structures. The integration of deep learning and imaging technology fundamentally transforming how researchers acquire, process, analyze and interpret data. This project intends to establish a new method about deep learning driven PAINT super-resolution imaging. A novel PAINT imaging probe based on DNA framework is designed to establish specific labeling and super-resolution imaging strategies for various subcellular structures. Deep neural network is trained to realize the super-resolution imaging prediction of various diffraction limited subcellular structures，reduce the amount of images for reconstruction and improve the reconstruction speed. Then the effect of deep neural network PAINT imaging will be verifyed and multiplex, fast super-resolution imaging of subcellular structure for living cells will be achieved . Through interdisciplinary technology integration, this project develops a novel PAINT probe and establishes a new method of deep learning driven super-resolution imaging, which is expected to break through the bottleneck of the development of super-resolution imaging technology and has a wide application potential in the future research of biomedical science.