Stem Cell Tracking And Division Detection Based On Image Sequence

Induced Pluripotent Stem Cells(IPS)is a kind of pluripotent stem cells,which is produced by introducing exogenous gene to induce somatic cells dedifferentiating.Induced pluripotent stem cells can differentiate in vitro,and could be used for developing various tissues and organs.It could offer diseases models in vitro,which can be used for treating intractable illnesses.However,cells are often in huge amount and distribute densely.The traditional way of cell research is mainly through manual observation.Therefore,many researchers are interested in the combination of computer science and biotechnology,which is aiming at taking advantage of computers to observe the process of stem cell growth and proliferation automatically.It may save research human resource and relatively help researchers.This thesis combined computer image processing technology and machine learning to research imagine segmentation,stem cell division detection and cell motion detection.Finally,based on the result of research,this thesis has developed a cell lineage map of the image sequence.It has directly reflected the process of cell motion and proliferation.Following are main results of this thesis:1.Based on phase-contrast microscope image restoration algorithm,it improved cell segmentation algorithm by combining different segmentation of light and dark cells.Firstly,it pretreated cell image sequence by removing disturbance information and correcting image shift,then segmented cells with improved algorithm and finally examined the performance of improved algorithm.2.It has studied various ways of cell division detection.It used multimodal fusion cell division detection method to solve the problem of high division detection missing rate in existing method when dealing with image sequence produced by long shooting interval.Compared with single detection algorithm,it could reduce missing rate to a large extent.3.It improved the Mean-Shift cells tracking algorithm based on multi-core combination.When using multi-core combination Mean-Shift algorithm to track long-shooting-interval cell sequence,problems of tracing missing could happen.Meanwhile,because of the dense distribution of cells,problems of mistracking may also happen.According to previous problems: 1)it proposed a loss-target-detection algorithm,which made use of ellipse feature matching to relocate target and introduced kalman filter to predict locations;2)it proposed a kernels weight adaptive algorithm,which will improve the tracking outcome whenever cells were dense or not.The experiment result proved that the improved algorithm had a higher accuracy than original method,and could find cell motion sequence more efficiently.4.It proposed a method of generating cell lineage map,which combined cell segmentation and cell tracking.Results of division detection were often in a low reliability when processing the long-shooting-interval cell image sequence.This could result in a low accuracy of cell lineage map based on split detection data.The generating method proposed in this thesis combined cell segmentation and tracking.This method no longer depends on division detection,which can improve accuracy significantly.It also built a cell lineage visualization framework,which could directly show the process of cell growth and proliferation over time,and eventually helped stem cell research.