Segmentation and Tracking Algorithms for in Vitro Cell Migration Analysis
Doctoral thesis, 2005
This thesis describes a system for automatic in vitro cell migration analysis. Image sequences of adult neural stem/progenitor cells were acquired using a time-lapse bright-field microscopy setup. The adult neural stem/progenitor cell has the ability to differentiate into three different neural lineages; neuronal, astrocytic and oligodendrocytic.
Different image analysis techniques were investigated for segmenting the cells in the images, such as watershed segmentation and boundary detection using dynamic programming. Some segmentation techniques required the positions of the cells to be detected first. This was done either using a multi-scale Laplace of Gaussian (LoG) filter, which detects blob-like objects in an image, or using the extended h-maxima transform. It was found that the performance of the multi-scale LoG-filter as a cell detector could be increased by using information about the cells-positions in the previous image.
To track the individual cells through the sequence, the segmented cells in two consecutive images were associated using Bertsekas modified auction algorithm. The association weights were calculated based on distance, correlation and size between possible matching cells. A comparison of three different segmentation methods, evaluated after completing the tracking step, showed that the best system was an algorithm consisting of a multi-scale LoG-filter, followed by cell border detection using dynamic programming. Using that system, 93 % of the cell-to-cell associations in the evaluated sequences were correct.
The obtained cell movement data was used for statistical modelling of the cell migration patterns. Using a Hidden Markov Model with two states, it was found that the motion of the glial progenitor cell was random 2/3 of the time, while the type-2 astrocyte showed a directed movement 2/3 of the time. This finding indicates possibilities for cell-type specific identification and cell sorting of live cells based on specific movement patterns in individual cell population, which would have a valuable application in neurobiological research.