Science and pirates!
Case essay: machine vision
An image-processing tool for the quantitative study of the developing Drosophila retina
Supervisors: Dr. Lewis Griffin | Dr. Franck Pichaud
Download: PDF (~2.5 MB)
The Drosophila retina is a highly ordered tissue in which cells adopt specific patterns with regularity reminiscent of inorganic crystal structures. Through morphogenesis, a coordination of cell growth and shape changes, unorganized epithelial sheets give rise to biologically active tri-dimensional structures, such as the Drosophila eye. Morphogenesis is driven by forces resulting from the subtle interplay of genetics, biochemistry and cellular mechanics. Most of the current knowledge is derived from fluorescence microscopy techniques that allow the imaging of cell patterns formation processes. However, it is only by adopting a system approach by integrating notions and tools borrowed from other fields such a fluid mechanics, material physics and computer sciences that it will be possible to fully understand the mechanisms underlying the formation of the Drosophila retina. More particularly, there is a need to make a transition from qualitative observations to robust and reproducible quantitative data.
Size distribution of the cells as determined by the image processing software
The goal of this project was to develop an image-processing program that could be used to extract information about cells, borders and vertices from a previously delineated fluorescence microscopy picture of the developing Drosophila retina and organize this data in a query-able format, allowing various quantitative analyses to be performed.
The image-processing software was programmed using Mathematica and allowed the quantitative analysis of biologically relevant features through a user-friendly interface. The distribution of size and number of neighbors were determined in a region-specific manner. Cells in the patterned regions (morphogenetic furrow and post-furrow) tended to have narrow distributions both in size and in number of neighbors in opposition to the unorganized pre-furrow regions where both distributions were found to be much broader. This might indicate that these analyses could serve as a basis to quantify the degree of disorder of epithelial tissues. Moreover, it was shown the numbers of neighbors of adjacent cells were dependant and that this dependency varies from a region to another.
