超分辨成像中快速超高密度荧光分子定位及追踪

Localization-based super-resolution microscopy (or called localization microscopy) has overcome the diffraction limit, and has been an important breakthrough for microscopy imaging techniques. Localization microscopy addresses a wide range potential application in Life science, Medicine, Physics and Chemistry. However, in this technique, there are some defects such as low temporal resolution, low-capability of imaging the high densely labeled region, and difficulties in on-line monitor of imaging process, which lead to the challenges in fast detection of complex biological events and molecules tracking. Aimed at the defects and challenges as proposed above, this project will carry out the following research. Firstly, by increasing the information (the features of the photons-emission of molecules) and introducing some progress in imaging science, we will develop the method that can achieve fast localization of ultra-high density active molecules with weak signal in super resolution imaging. Secondly, based on the first part of research of this project, and combined the advantages of probability model and minimal energy model in molecule tracking methods, we will propose the method of fast tracking of high throughput high density active molecules. This project will boost the performance of localization microscopy and overcome the challenge, i.e., high density data analysis and process, in the field of molecules tracking. We firmly believe that this project will provide an efficient tool to observe the basic process of life activity, analyze the deep-level dynamics of molecules in living cell and carry out some relative researches.

基于荧光分子定位的超分辨成像(称为定位超分辨成像)技术突破了衍射极限，是显微成像领域的一个重要进展，在生命、医学、物理、化学等多个领域有极大的应用潜力。但是该技术目前存在时间分辨率低、分子标记过密区域无法成像、成像过程难以实时监控等不足，因此，用于复杂生物事件快速检测、分子追踪等领域存在较大挑战。针对上述不足与挑战，本项目拟开展如下工作：（1）通过增加信息量（荧光分子发光特性）和引入图像科学领域的工具，建立起适合于弱信号水平下，快速超高密度荧光分子定位方法；（2）在前一个工作基础上，融合分子追踪的概率模型和能量最小模型各自优势，发展快速的高通量高密度荧光分子追踪方法。本项目的研究将提升超分辨成像性能，拟解决分子追踪领域的一个公开问题：高密度数据分析与处理，从而为观测生命活动基本过程、分析活细胞中深层次分子动力学行为、以及相关研究提供有意义工具。

在超分辨成像中，用于分子定位及追踪的方法拥有坚实的理论知识，能够被一些数学模型描述。在这里，我们展开的工作主要集中在这些数学模型在生物图像信息领域的一些应用。首先，我们建立了一系列的快速的逆卷积方法用来提升光学图像的分辨率。这些方法依赖于高效率的稀疏矩阵乘积运算，避免了图像的频域分析，从而比传统的逆卷积方法有更强的抗噪声能力，同时，能融入图像的先验信息，逆卷积的图像更接近真实图像。所建立的逆卷积方法，原理上实现了快速超高密度的分子定位。我们扩展了超分辨分子定位模型的运用范围，将其模型应用于复杂形态的神经元定位及从低信噪比的钙成像信号重建神经元的放电活动。因所建立的模型可以融合重建信号（胞体位置分布、放电序列）先验信息，拥有比其它方法更好的重建结果。最后，受到分子追踪模型的启发，建立约束主曲线法，用于神经突起追踪。基于约束主曲线法，我们先后实现了大尺度快速重建不连续形态的神经元，及高密度分布纤维的神经群落形态重建。