Topological Constraint In High Density Cells' Tracking Of Image Sequences

Cell movement plays a crucial role in many biological, and analysis of cell movement mainly includes tracking target cells, when it is an impossible task of identifying and tracking millions of cells in manual method, an auto-tracking system which can track and analysis thousands of cells became a research hotspot.The cell tracking method based on graph theory is newly developing. This cell tracking method is aim to multi-cellular situation, which take neighborhood topological relationship into account fully. The tracking method based on graph theory is better than previous method at high level data, and has great potential to achieve task of tracking a large number of cells. All of the merits make it to be a method worthy of further study.The multi-target tracking in cell image sequences is the main difficulty in cells'locomotion study. Aim to cells'complexity movement in high density cells' image, this paper has proposed a system of segmentation and tracking. In the part of cell image segmentation, we adopted different methods of segmentation base on the characteristics of cell image sequence in experiment. Due to the existence of overlapped and clustered cells in segmented image after basic cell image segmentation, we used the method of cell contour tracking coding based on Freeman, which can analysis cell contours morphology and identify the concave corner in overlapped cells according to encode characteristics of the chain code, then complete separation of overlapped cells by grouping and matching concave corner. In the part of tracking, the proposed algorithm of this paper combined overlapping and topological constraints to track inactive and active cells respectively according to cell motion characteristics. For tracking high density cells, we improved the original method with introducing area item to topological constraint conditions to solve the problem of false tracking for high density and cell division. The experiment result shows that the advanced algorithm solute this problem effectively, and help to improve the tracking efficiency, especially for the image of high cell density and frequent cell division. In addition, our algorithm sets map distance threshold according to cells'number, which increase the applicability of this algorithm. We tested two different image sequences in our tracking method, the experimental result shows that our method has higher tracking accuracy.