Cell tracking method based on global and local optimum

Abstract

The invention discloses a cell tracking method based on global and local optimum. According to the method, independent regions are obtained through initial segmentation by means of ellipse fitting, and adherent cells in need of segmentation are identified; then a wavelet transformation method is used to carry out further segmentation; the center of mass of each cell subjected to segmentation is calculated and used as a unique characteristic of the cell; a bipartite graph matching method is used for inter-frame correlation of the cells, then cell shifting-in, shifting-out, division and shielding events are identified, matching is iterated and incomplete cell track segments are obtained; and through analysis, identification and correction of cell track segments, finally a complete cell track is generated based on local optimum. The method utilizes cell location information to achieve sub-cell relation of cell division and has good effect on different data sets.

Description

technical field [0001] The invention relates to a technology in the field of image processing, specifically a layered idea, which solves the problem of low accuracy of cell detection and tracking. First, the global image is processed, and the detection and tracking errors in the current stage are identified through automatic detection, and the accuracy is further improved through local optimization processing to obtain a complete trajectory. The specific application is: to model cell lineage by tracking large-scale cell movement under a microscope, and to analyze changes in cell cycle and cell activity of cells under different drugs and environments. Background technique [0002] The detection and tracking of cells under a microscope is an important means of studying cell life activities, including the development, proliferation, differentiation of a single cell, the interaction between multiple cells, and the migration of the cytoskeleton, thereby predicting the fate of the...

AI Technical Summary

Problems solved by technology

The shortcomings of this method are: due to the specificity of cell morphology, the extraction of cell features is specific in the process of tracking and matching, the scalability and transplantability are not high, and the trajectory error needs to be corrected manually in the later stage

The disadvantages of this method are: the first frame of cell outline needs to be manually initialized at the beginning of tracking, and the energy function needs to be reconstructed for the number of cells moving in and out of the field of view. At the same time, this method is very time-consuming in the iterative process. Cell scale limited by computational cost

The disadvantage of this method is that it needs pre-human-computer interaction or other initialization to obtain real data sets for training, and at the same time, as the scale of cell division increases, the computational complexity also increases exponentially.

However, compared with the present invention, this existing technology has unsolvable technical problems, including that as the number of tracking cells increases, the calculation scale brought by prediction becomes larger, and it cannot solve the detection error caused by cell occlusion and noise, and cannot solve the problem of cell occlusion. The situation that reappears after a short disappearance is not robust to judging the division by the area method after cell division, and is susceptible to noise interference

[0011] In summary, the existing cell detection and tracking methods have the following disadvantages: (1) The accuracy of detection is affected by cell adhesion, and at the same time, the uneven brightness of cell imaging brings over-segmentation problems and detection errors caused by noise; (2) Different The specificity of cell phenotype makes the algorithm not scalable for specific data sets; (3) Since the appearance of the same kind of cells is not much different, after cell division, the phenotypic characteristics are not well distinguished in the process of identifying and associating daughter cells (4) For various cell events during large-scale cell tracking: rapid and sudden movement, splitting, moving in and out, transient disappearance, occlusion, most of them are for a specific situation, and there is no good unified solution; (5) The scale of cell tracking is limited by the algorithm, and the calculation time increases rapidly; (6) The algorithm needs to manually set the initial state or parameters, which cannot realize fully automatic tracking, and the deviation needs to be manually corrected later

Images