2009 Project Week Breakout Session: 3D+T Microscopy Cell Dataset Segmentation

Back to 2009_Summer_Project_Week#Agenda

Following the presentations at the NAMIC project week winter 2009 and publications of the corresponding work in the Insight Journal (see below), we are presenting here our latest work on Cell Segmentation in Confocal Microscopy Datasets. Most of the algorithms we will present here are available in [ITK 3.14] as classes as well as fully integrated in a free, open-source application called [GoFigure2].

Because of the size of the datasets (terabytes), the number of cells to process(millions), the intended audience (biologists), and the fact that we cannot foresee what the requirements will be even one year from now, we designed GoFigure to be a layered application with a plug-in mechanism. The layered approach (IO, Processing, GUI) allows for very easy extension of GoFigure. For example, GoFigure right now handles two IO components: File System, and MySQL Database. Before the end of the year, we plan to have an Omero Component, and a component to support Harvard Medical School specific storage cluster. That IO layer mechanism, along with client-server mechanisms has been one of the key in ParaView or Slicer applications. The danger of an integrated application is to tightly integrate a predefine set of service and never being able to extend it. Fortunately, GoFigure implements a plug-in mechanisms based on Qt to allow extension while still taking advantage of better speed and integration than scripting languages allow. Slicer 3 is a good example of how successful such plug in mechanism can be.

Here is a detailed list of the features *currently* available that will be presented during this session:
< Kishore Mosaliganti>

• Chan and Vese level set
• Mulltiphase level set
• Coupled level set for non-overlapping objects segmentation (e.g. ... cells)
• Coupling Membrane channel and Nucleus channel for better segmentation
• Specific pre-processing algorithms for nuclei segmentation
• Cells Tracking using level set

<Arnaud Gelas>

• Region based level set
• New level set framework (kernel based solution)
• Compactly Supported Radial Basis Functions: Interpolation and Approximation
• 2D Manual segmentation

< Alexandre Gouaillard >

• Layered design
• Specific Streaming framework for Multiphase Level Set
• Multifiles Acquisition "Reader" (with support for Zeiss multi time macro naming convention)
• DataBase Support

< lydie Souhait >

• Qt Widgets
• Qt interface of Visualization layer
• Qt DB front end

Here is a list of features that are available but not mature enough to be transferred to ITK or integrated in GoFigure

• 3D Manual segmentation (PCA, Level set, ...)
• lineage reconstruction and extraction
• plug in mechanism (Qt based, cannot be transferred to ITK)

In a true community sharing spirit, our entire development process is open and free. Our svn server, road map and documentation is hosted on sourceforge and fully public. All our source code has been double-checked to be patent-free, and is distributed under the very permissive BSD licence. Finally, whenever possible, we bring our implementation of the features up to the quality level of the ITK toolkit and transfer them to best serve the community.

[1] NAMIC winter 2009 - Cell Segmentation Session
[2] NAMIC winter 2008 - GoFigure Presentation
[3] NAMIC winter 2009 - ITK level set solution for cell segmentation in microscopy datasets
[4] NAMIC winter 2009 - VTK widgets for manual segmentation and manual validation of segmentation

• Mosaliganti, Smith, Gelas, Gouaillard, Megason, "Level Set Segmentation: Active Contours without edge", [Insight Journal]
• Mosaliganti, Gelas, Gouaillard, Megason, "Microscopy Image Analysis: Blob Segmentation using Geodesic Active Contours", [Insight Journal]
• Mosaliganti, Smith, Gelas, Gouaillard, Megason, "Cell Tracking using Coupled Active Surfaces for Nuclei and Membranes", [Insight Journal]
• Mosaliganti, Smith, Gelas, Gouaillard, Megason, "level set segmentation using coupled active surfaces", [Insight Journal]
• Gouaillard , Brown, Bronner-Fraser, Fraser, Megason, "GoFigure and The Digital Fish Project: Open tools and open data for an imaging based approach to system biology", [Insight Journal]