Cells Tracking Algorithm Based On Cellular Partition Combined With Multi-Frames And Original Images Feedback Correction

Cells’ research has played a crucial role in biology, medicine and other related fields. Thecomplex phenotypic changes and dynamic behavior of the cells can be observed via theanalysis of cell image data, and reveal the implicit law of life. Thus the development ofbiomedical can be greatly promoted. Cells’ tracking as one of the important part of cells’research has become the research hotspot and difficulty in recent years.Topological constraint algorithm is a novel cells’ tracking method which have beenproposed in recent years. The method can effectively track the high-density cell imagesequences, and is not sensitive for the deformation of the cell. But the processing capability ofthis method is not strong for the image area with sparse cells and some special cells, whichinclude cluster, division, moving in, moving out and so on. And this tracking method stronglydepends on the effect of the segmented images. To solve these problems, the method in thispaper applies the cellular model and Delaunay triangulation to topological constraintalgorithm, and first proposes the feedback correction method of multi-frames and original,then establishes a multi-module tracking system. The tracking system firstly applies regionoverlapping to track the inert cells, and provides topological constraints algorithm with theinitial source vertices. Then the system partitions all segmented images into cellular region,and establishes the cellular information storage center, which provides cell information andsearch range for faster and more accurate extraction, transmission, matching and correction ofthe cell trajectory information of the entire sequence. After Delaunay triangulation is appliedto the topological constraint algorithm and establish cells’ neighborhood graph. All cells havebeen matched via the topological constraint algorithm, and the matching results are stored intrajectory temporary storage. After the cells in trajectory temporary storage have beenclassified, then the trajectories of the cells have been corrected. All cells have been classifiedinto seven categories which are under segmentation, over segmentation, cluster, division,moving in, moving out and obscure one in the trajectory correction part. According to thecharacteristics of the cells’ types, the system applies feedback of multi frames and originalimages to correct matching results and segmentation errors. The system updates cellularinformation storage center and cell information storage center finally. The method in this paper applies the global nature of Delaunay triangulation to establishcells’ neighborhood graph. And this method uses cellular model which has the advantages ofseamless coverage and non-overlapping to partition all segmented images into cellular region,which is neither missing nor overlapping. This method solves the problem of cells’ unevendensity distribution via combination of global and local. In order to solve the problems whichprocessing capability is not strong for the special cells and strongly depends on the effect ofthe segmented images, this method applies the feedback method of multi-frames and originalimage to correct matching results and segmentation errors.The three image sequences have been tested by the method. Experimental results showthat the method can effectively track the image sequences with cells’ uneven densitydistribution, seriously clustering and moving fast. The method not only has greatly correctedsegmentation error to reduce the tracking dependence of segmentation, but expanded theapplicability of the topological constraints. Therefore it has greatly improved the accuracy oftracking.