生物显微投影图像的自动对位算法研究

The precise alignment of biological microscopic projection images is the prerequisite for the three-dimensional structure of biological macromolecules. However, current alignment technology either rely too much manual intervention, or the complexity of the algorithm is too high. Also, the technology can not handle effectively the high noise projection images..Facing the above problems, this proposal will study the alignment algorithms for golden marker alignment and markerless alignment. First, we will study the golden marker extraction algorithm and projection image feature extraction algorithm, to realize the accurate feature extraction in high noise conditions. Then we will study the matching and tracking technology for golden markers and image features, in order to accurately determine the mapping relationship between different images. Next，through the construction a projection model that can reflect the real imaging process, we will study the optimization strategies of alignment parameters, to obtain accurate results for image alignment. Based on above research, we will study the high performance parallel techniques for projection images alignment. Finally, a user friendly biological microscopic projection image alignment software will be developed, to provide an effective tools for processing Biological Projection images..Now China already has the most advanced electronic microscopic imaging equipment in the world, and a large amount of biological microscopy image data is produced every day. Therefore, Our research not only has important theoretical significance, but also has practical application value.

生物显微投影图像的精确对位是获得生物大分子三维结构的前提。当前的对位技术或依赖过多的人工干预，或算法复杂性过高，而且无法有效处理高噪声投影图像。.针对当前投影图像对位研究存在的问题，本课题将开展胶体金标记和无胶体金标记对位的算法研究。首先，我们将研究高效的胶体金标记和图像特征提取算法，实现高噪声下特征的准确提取。然后研究胶体金标记和图像特征的匹配和追踪技术，以准确确定不同图像间的点链关系。再次，通过构建能够真实反映成像过程的模型，研究对位空间参数的优化求解策略，获得准确的图像对位结果。在此基础上，研究高效的投影图像对位并行优化技术。最后，研制用户友好的生物显微投影图像对位软件，为生物科学家提供处理显微数据的有效技术工具。.目前我国已具备国际上最先进的电子显微成像设备，每天正在产生大量的生物显微图像数据。因此，本课题的研究不仅具有重要的理论意义，而且有着实际的应用价值。

生物显微投影图像的精确对位是获得生物大分子三维结构的前提。当前的对位技术或依赖过多的人工干预，或算法复杂性过高，而且无法有效处理高噪声投影图像。.针对当前投影图像对位研究存在的问题，本项目开展了胶体金标记和无胶体金标记对位的算法研究，并取得了如下进展：针对传统的胶体金匹配算法对人工干预依赖过大的问题，提出了一种全自动的胶体金电子断层对位算法Markerauto；并设计了一个基于高斯混合模型GMM的自动算法来加速基准标记的跟踪过程；利用特征识别算法的灵活性，设计了一种旋转不变特征冷冻电镜图像颗粒自动挑选算法；针对传统的几何参数确定方法精度不高，需要人工标注的问题，提出了AutoGDeterm算法以实现高精度全自动的几何参数确定；基于以上研究成果，设计并开发了面向生物显微投影图像的对位软件，并引入GPU并行技术对算法的运行速度进行了加速。.在项目执行过程中，主要参与人员在Cell Research、Bioinformatics、Journal of Structural Biology、Journal of Computational Biology、IEEE/ACM Transactions on Computational Biology and Bioinformatics、Tsinghua Science and Technology、IEEE International Conference on Bioinformatics & Biomedicine 2018等国际期刊和国际会议上发表论文10篇，其中SCI期刊论文9篇，EI会议论文1篇。本项目的研究可为生物信息处理领域提供自动对位技术，有实际的应用价值。