Time-Lapse Microscopy Imaging and Image Analysis for Cell Migration Studies
This thesis considers a complete automatic system for in-vitro cell imaging and migration image analysis. The development of such can be divided into three main problems:
The first considered problem concerns how to build a robust time-lapse sequence acquisition system, for imaging adult hippocampal cells in bright-field microscopy. Due to low contrast the focusing problem becomes ill-defined, and image processing techniques, such as wavelet image fusion, is suggested for obtaining images with high contrast as well as resolution.
The second (and main) considered problem is cell image segmentation, i.e how to automatically locate and delineate cells in microscopic images using digital techniques. Cells cultured on glass plates have a tendency to congregate, which makes this task difficult. We study several different methods before arriving at our preferred choice; a modified version of the Chan-Vese level set segmentation. To improve cell segmentation performance, we apply an adaptive multi-scale Laplacian of Gaussian filter. Then, to solve the problem of undetected cells and merged cells (several cells detected as one), we add a new extension to the level set method that we refer to as specialized growing and pruning.
The third problem concerns the creation of a motion model for the cells, to aid the tracking and data association and to classify cells on the basis of their migration pattern. We take on several different approaches before concluding that cell migration can be modelled by a non-stationary two-state hidden Markov model. Using this model, we find that glial progenitor cells are moving randomly 2/3 of the time, while type-2 astrocytes show a directed movement 2/3 of the time. This finding indicates possibilities for cell-type specific identification and cell sorting of live migrating cells, which will have a valuable application in neurobiological research.
cell image segmentation
cell migration modelling