显微镜图像栈序列中植物细胞的鲁棒追踪算法研究

The shoot apical meristems (SAMs) also referred to as the stem-cell niche, is the most important part of the plant body. The SAM cells are imaged by Confocal Laser Scanning Microscopy and stored in image stack time series. Developing a computational platform capable of robustly tracking SAM cells in the 4D image stacks is very critical to obtaining quantitative spatiotemporal measurements of a range of cell behaviors. The cells in the SAM are tightly clustered in space and have very similar shapes and intensity distributions, and there are much noise in the deeper layer image slices because of the absorbtion of laser energy , and the images can be translated, rotated and scaled in the imaging process, thus how to segment and track all cells in image stack time series can be very challenging. In our research, we propose a local graph matching based method to track the cells both spatially and temporally,and identify cell divisions at the same time. The geometric structure and topology of the cells' relative positions are efficiently exploited as the basic feature to match the cells. After that, the local graphs are normalized and the translation, rotation and scaling components are removed before the matching procedure, so our local graph matching based tracking algorithm could resist the noise of translation, rotation and scaling nosie from the imaging process. Furthermore, we build a joint segmentation and tracking system, where the tracking output acts as an indicator of the quality of segmentation module and, in turn, the optimized segmentation can be improved to obtain better tracking results. Last, we associate the 4D tracklets in the space of image stack time series through the Bipartite graph matching algorithm, and then achieve the optimized tracking results.

植物顶端分生组织是植物器官中最重要的部分，植物学家通过共焦激光扫描显微镜来采集其细胞数据并存储为图像栈时间序列。开发鲁棒追踪四维（3D+时间）植物细胞及其分裂的算法对获取关于细胞行为模式的时空测量数据极其重要。由于植物细胞拥有相似的形状和灰度分布，且具有空间上紧密相连的特殊结构，在成像过程中又存在严重的噪声，图片还可能被错位、旋转或者放缩，这都给同时追踪所有细胞带来了巨大挑战。在本项目提出的追踪算法中，采用基于局部图匹配的方法来追踪细胞，每个细胞与其周围细胞组成的局部图的几何形状以及拓扑结构被用作匹配的基本特征。在此基础上，通过把局部图归一化去除位移、旋转以及放缩因子后再进行匹配，使本项目提出的追踪算法能成功对抗成像过程中带入的位移、旋转及放缩等噪声。而且，通过引入追踪算法的反馈来调节分割算法的参数而纠正局部分割错误。最后，通过二分图匹配算法实现四维追踪片段的重新连接，使追踪结果达到最优。

开发一种能鲁棒追踪植物细胞及检测其分裂的算法对获取植物细胞的行为模式极其重要。本项目提出了基于三角邻域结构局部图匹配算法，用图来描述细胞之间的拓扑结构，提取细胞的相关特征，包括细胞面积、邻居细胞方位分布和邻居细胞数目等信息。利用从构建的三角局部图中提取的特征向量来寻找匹配时的种子细胞对，再从种子细胞对开始利用邻居细胞扩散的方式进行更加全面的追踪。在此基础上，通过把三角邻域局部图归一化去除位移、旋转以及放缩因子后再进行匹配，使得我们的追踪算法能成功对抗成像过程中带入的位移、旋转及放缩等噪声。同时，通过引入追踪算法的反馈来调节分割算法的参数而纠正局部分割错误，对欠分割和过分割错误进行识别和局部修复。最后，基于匈牙利算法并开发了一种能重新连接被噪声和其他原因中断的细胞轨迹的算法。该项目成功探索了显微镜图像栈序列中植物细胞的鲁棒分割和追踪问题，基于该项目的研究成果，以第一作者和通讯作者发表SCI论文三篇和会议论文一篇，培养硕士研究生5名，项目研究的部分成果已经被国内外著名大学和研究机构采用。