Stem Cell Image Segmentation And Cell Tracking

Since the first discovery of induced pluripotent stem cells(iPS)by biological researchers in 2006,more and more technologies of iPS are emerging,and has been already applied in practice.It not only can provide the exoteric disease models to medical researchers,but also can be used for the differentiation of cells to produce organs for all kinds of difficult diseases.Digital image processing and high speed information processing technology has been an effective means to quantify the processing of the growth and differentiation of iPS and help researchers to improve the technology of inducing stem cells to benefit mankind.Based on the time-lapse phase contrast microscopy image sequences of stem cell populations,this thesis are trying to make the use of computer vision and machine learning to solve these three specific application problems of iPS: how to segment the stem cells from the background effectively,how to extract mitotic events in cell image sequences completely,and how to track the stem cell image with a large number of cell mitosis.The main exploratory work in this thesis is as follows:(1)A method of preprocessing technology and cell segmentation: Due to the optical principle,phase contrast microscopy images contain artifacts,such as the halo and shade-off,which hinder image segmentation and cannot get the ideal result.We studied the optical properties of the phase contrast microscope to model its image formation process.Based on this,we have succeeded in restoring artifact-free phase contrast images.According to the morphological characteristics of stem cells in the growth process,the stem cells were divided into "dark" cells,and "bright" cells.And through calculating the cell size,we remove the region of segmentation error.(2)A method of mitotic detection of stem cells based on multimodal fusion: Due to equipment and photography technology,it is difficult to obtain a sequence of images less than 15 minutes interval even in the top biological laboratories of our country,which caused a large number of cell growth information is lost.We have tried two mature technologies,HMM and CNN.But both of them are difficult to detect the mitosis of this kind of time-lapse.Therefore,this thesis proposes a new method based on multimodal fusion for stem cell mitosis detection,fusing the two phases of cell division periods then doing mitotic detection.The experimental results show this method can significantly improve the accuracy of mitosis detection.(3)A method of fusing stem cell mitosis detection and cell tracking: There are a large number of cell mitosis and proliferation in stem cell growth,causing the traditional multi-target tracking method is difficult to accurately track the growth of stem cells.Therefore,this thesis proposes a stem cell tracking method based on cell mitosis detection.The results show that this method can significantly reduce the computational load of stem cells tracking.