Cell Shape Matching for Efficient Live Cell Imaging Experiments

Live cell imaging is commonly used to study cell division, in which the cross-sectional images of an embryo are obtained at different time points using a confocal microscope. Starting from the root cell (zygote), we extract and track each cell and its daughter cells during cell division. Through the mother and daughter relationships, we can identify cells in all successive generations, name them and create a lineage tree.If we start image collection at a later stage, for example, when there are 150 cells, we can avoid laser radiation before the 150-cell stage. Thereafter, we can use higher laser power and higher temporal resolution for better image quality. In this way, the total light dose from the laser radiation can be kept under the limit in order not to damage the embryo. The problem with this approach is that we do not have the ancestor information of these 150 cells and cannot identify them. To tackle this problem, we will develop cell shape matching techniques. The strategy is to match these 150 cells’ shapes with those from a standard experiment that started from the root cell and find the corresponding positions in the lineage tree.Cell shape matching is also useful for studying cell dynamics, such as the shape changes between two time points. Another important application is to match cells from standard and gene knockdown experiments to infer gene regulatory functions.In this project, we will extract robust cell features for effective shape matching. Invariant pattern matching techniques will be investigated to allow cell variations in scale, translation and rotation. A high-order compatibility formulation will be developed to deal with non-rigid deformations. Tensor CUR decomposition will be employed to reduce both computing time and storage space.In addition to matching individual cells’ shapes, we will also consider their dynamic morphological features, nucleus trajectories in the temporal direction, and the cells’ neighborhood relations to improve matching accuracy. We will develop a multi-step matching process to solve these problems, one for matching static cell shapes, one for their dynamics, and one for their positions and neighbors.Our project will make it possible to conduct live cell imaging experiments more efficiently and provide innovative techniques to investigate cell dynamics and identify gene functions. Our technology will open the door to developing a new generation of methods for live cell imaging and cell division data acquisition and analysis.