Time-Lapse Bright-Field Microscopy and Image Acquisition of In-Vitro Neural Stem Cells.
This project involves development of a system for time-lapse image acquisition of neural stem cells. The purpose of analyzing time sequences is to find migrational patterns that are characteristic for different cell types. Neural stem/progenitor cells posses the capability to differentiate into three lineages: neuronal, astrocytic, and oligodendrocytic. Preliminary results show that glial progenitors migrate significantly more random, than for example astrocytes, that express a more directed movement.
Automated image acquisition constitutes the first part of a migration analysis system, followed by an image segmentation and tracking module. It is crucial for the analysis that images are properly acquired. Translation between subsequent images has to be compensated for, and background non-uniformity eliminated. Most important is that the focus level is selected optimally with good repeatability. Optimal in the sense of accurately locating cells, is a criteria difficult to formalize, especially since the bright-field microscope does not produce sufficient contrast for imaging cells in-focus. More contrast was bought at the cost of lower resolution by defocusing, and to find the optimal balance we moved on to work with cell segmentation methods.
Inspired by scale-space models of retinal receptive field sampling in human vision, we used multi-scale Laplace of Gaussian with automatic scale selection as a detection filter. The cell shape was computed by tracing the border using dynamic programming. This method show slightly better performance than the watershed segmentation and the multi-scale approach is a promising step towards finding optimal prerequisites for cell location.