The Methode Of High Density Cells' Tracking Based On Topological Constraint Combined With Hungarian Algorithm

Cells'tracking plays a crucial role in many fields, such as the cells'behaviors, pharmaceutical research and treatment of diseases. For example, neuron stem cells have been proofed can treat the diseases of nervous system through transplantation which can greatly improve the diagnostic accuracy. However, the key techniques is to find out the specific neurons stem cells'progenitor cells which developt into specific nerve cells and then cultivated into the neural organs with specific functions by observing and studying neuron stem cells'natural separation, value-added rule and mobile process. Thus cells'tracking has a very important significance. In recent years cells'tracking system which can simultaneously track and analysis of thousands of cells automatically has become a research hot spot.Cell'tracking method mentioned in this paper is based on topological constraint theory and combination of the Hungarian algorithm, which is classified the tracking method based on image segmentation. Graph model is used to describe the topological relationship between cells, and cell tracking using topological constraint is to transform the tracking problem into the problem of vertex matching in the segmented image. Topological constraint method has been proved to track the high-density cell image effectively and not sensitive to the deformation of the cell. However, the applicability of this approach is not good when the cells were relatively sparse in the image and when cleavage has happened simultaneously among adjacent cells. For these problems, the cells'matching problem in two adjacent frames is regarded as the problem of a multi-objective optimization allocation, using the Hungarian algorithm which has macroscopic and overall characteristics. Ultimately achieve cell tracking with higher efficiency.Because of the tracking method proposed in this paper is based on segmentation, the cell image sequences should be segmented firstly, we adopted different methods of segmentation base on the characteristics of the two cell image sequences in our experiment. The cell image sequence 1 is a neuron stem cell image imaged by confocal microscope, which has low image contrast ratio with serious problems of cells'adhesion and cluster. Therefore, segmentation of this sequence is a tough job. The segmentation we used is based on level set algorithm without re-initialization combined with local gray threshold. The cell image sequenceⅡis fluorescence image sequence which is segmented by threshold segmentation method combined with sunken point matching. After segmentation the cells between two adjacent frames will be data associated according to the topological relationship between cells, which constitute the coefficient matrix. Then we use the Hungarian algorithm to implement optimize matching and achieve cells' tracking.Finally, the two cell image sequences have been tested using the method preoposed in this paper and the experimental results show that this proposed tracking method has better tracking ability and higher accuracy.