Hiraki, Hideaki et al. (2009) International Worm Meeting "Semi-automatic system for the creation of cell shape models in C. elegans embryogenesis."

Kohara, Yuji, Ueta, Yumiko, Hiraki, Hideaki

[International Worm Meeting, 2009]

Cell to cell interactions play critical roles in early embryogenesis, therefore, it is very important to have information about the arrangements of cells, cell shapes and the contact among them. We have been developing a computer system to create cell shape models from a time series of confocal microscopic images of the embryo whose plasma membrane is stained with a vital fluorescent dye or is marked with a fluorescent protein. This system consists of two subsystems. In the first step, cell shapes are automatically calculated by a seeded region growing algorithm from a 3D image and a set of seed point coordinates. Manual editing of the seed coordinates is required and is assisted by its graphical user interface. The second step is an active balloon model under gradient vector flow executed on a PC cluster system. This step removes the bumps of cell shapes derived from image noises. Consequently, cell-to-cell contacting areas can be quantified. We recorded images of OD58, the strain expressing GFP targeted to the plasma membrane, with a multiphoton confocal microscope. We used glass bottom dishes to avoid variable pressure from coverslips. We could record for 1.5 hours from one cell stage on the condition that the embryos continued the development and hatched successfully after the recording. Ten normal embryos were recorded and their cell shape models were built. We analysed the time course of each cell-to-cell contacting area by fitting a line to its graph and found that some of them increased or decreased consistently from embryo to embryo and others were variable in this respect. This suggested that some part of cellular arrangement was intrinsically variable and that the consistency of cell-to-cell contacts was not necessary to be complete for normal embryogenesis. We are improving the system more, and plan to apply this system to compare the cellular arrangements and the cell-to-cell contacts among mutant embryos and the embryos from other species closely related to C. elegans.

Hideaki Hiraki et al. (2007) International Worm Meeting "Semi-automatic system for the creation of cell shape models in C. elegans embryogenesis."

Yuji KOHARA, Hideaki Hiraki

[International Worm Meeting, 2007]

Cell to cell interactions play critical roles in early embryogenesis, therefore, it is very important to have information about the arrangements of cells, cell shapes and the contact among them. We have been developing a computer system to create cell shape models from a time series of confocal microscopic images of the embryo whose plasma membrane is labeled with a vital fluorescent dye or a fluorescent protein. In this system, cell shapes are automatically calculated by a seeded region growing algorithm from a 3D image and a set of seed point coordinates. Manual editing of the seed coordinates is required and is assisted by its graphical user interface. We applied this system on the "dub" data set in WORM 4D CDs kindly provided by Bill Mohler, and obtained cell shape models of 24-200 cell stages successfully for the most part. This time, in collaboration with Jon Audhya, we performed live recording of OD58, the strain expressing the fusion of GFP with pleckstrin homology domain which is targeted to the plasma membrane. We successfully obtained data sets of 1-24 cell stages. We also developed a post-processing system to remove the bumps of cell shapes derived from image noises by implementing an active balloon model under gradient vector flow. As a result, the phenotype of par-2 RNAi embryos could be evaluated in terms of cell volumes and cell-to-cell contacting areas. We are improving the system more, and plan to apply this system to compare the cellular arrangements and the cell-to-cell contacts among mutant embryos and the embryos from other species closely related to C. elegans.

Hideaki Hiraki et al. (2003) International Worm Meeting "Semi-automatic system for creation of cell shape model in C. elegans embryogenesis"

Hideaki Hiraki, Yuji KOHARA

[International Worm Meeting, 2003]

Cells change their shapes and arrangements in embryogenesis. 4D microscopy of C. elegans is useful to observe the process. Time-lapse recording of multiple optical sections with Nomarski DIC microscope and tools to analyse the results have been developed and used successfully, especially for cell lineage analysis because cell nuclei of C. elegans are evident in DIC images. Though cell shape models of early embryo can be reconstructed from DIC images, delineating cell boundaries become harder as the cell number increases. On the other hand, cell shapes can be directly visualized by confocal fluorescent microscopy with vital stain of plasma membranes. We have been developing a computer system to create cell shape models from a time series of confocal images of plasma membranes. In the system, cell shapes are automatically calculated by a seeded region growing algorithm from a 3D image and a set of seed point coordinates. The algorithm segment the image volume to enclosed regions one per a seed. Here a cell boundary is detected as ridges of high fluorescent signal between two seed points. In many cases, the seed points can be automatically placed using results of another time point. But manual editing of the seed sets is necessary where signal to noise ratio is low. The system constructs a pseudo cell lineage from the overlay relationships among the segmented regions over the time course. The inconsistency in the pseudo lineage indicates where editing is required and the editing is assisted by the graphical user interface. We have tested the system on the "dub" data set in WORM 4D CDs, 6/17/97, Bill Mohler, UW-Madison, that is of N2 embryo in 24-200 cell stages. Changes in physical parameters of cell shapes and arrangements, such as cell volumes and cell-to-cell contacting areas, could be estimated from the resultant shape model of a whole embryo under gastrulation. We are planning to apply this system to compare mutant embryos and the embryos from other species closely related to C. elegans.

Hideaki Hiraki et al. (2005) International Worm Meeting "Semi-automatic system for the creation of cell shape models in C. elegans embryogenesis"

Yuji KOHARA, Hideaki Hiraki

[International Worm Meeting, 2005]

Cell to cell interactions play critical roles in early embryogenesis, therefore, it is very important to have information about the arrangements of cells, cell shapes and the contact among them. Early embryogenesis of C. elegans can be analyzed utilizing time-lapse recording of multiple optical sections with a Nomarski DIC microscope. Though cell shape models of early embryo can be reconstructed from DIC images, delineating cell boundaries becomes harder as the cell number increases. Conversely, cell shapes can be directly visualized by confocal fluorescent microscopy with vital stain of plasma membranes. We have been developing a computer system to create cell shape models from a time series of confocal images of plasma membranes. In this system, cell shapes are automatically calculated by a seeded region growing algorithm from a 3D image and a set of seed point coordinates. Manual editing of the seed coordinates is assisted by a graphical user interface. The region growing algorithm segments the image volume in such a way that one region contains one seed. A cell boundary is detected as ridges of high fluorescent signal between two seed points. We tested the system on the "dub" data set in WORM 4D CDs kindly provided by Bill Mohler, and obtained cell shape models successfully for the most part. However, we encountered a problem that some contours of cells weren't delineated correctly where the signal to noise ratio was low. This time, we reinforced the system with the following two methods to fix the problem. One is to merge some regions after the region growing step. By this method, a cell could be represented by a few regions and the noise-derived false region boundary became minimized. Another one is a correction method based on the comparisons of the segmentation results among neighboring timepoints. Even if the cell shape is delineated incorrectly, the images of its neighboring timepoints can be sometimes processed correctly. However, because cells deform during development, comparing cell shapes among the neighbor timepoints is not straightforward. Therefore, we devised a program to calculate a smooth deformation field between two timepoints assuming that cells deformed continuously. Some errors of the segmentation could be corrected by voting among the candidate shapes that were generated by deformation of the segmentation results from the neighbor timepoints. We are improving the system more, and plan to apply this system to compare the cellular arrangements and the cell-to-cell contacts among mutant embryos and the embryos from other species closely related to C. elegans.