Cell Organizer: Building Models of Cell Structure from Microscope Images and Using them for High-Content Screening and Cell Simulations
Gregory Johnson and Devin Sullivan (Carnegie Mellon University, Pittsburgh, PA, USA)
May 18, 2013
CellOrganizer is an open source software system that can learn models of the size, shape and spatial distribution of cellular components directly from images. These models are generative, which means that they can be used to synthesize new images of cells that are statistically similar to the ones they were trained on. Such images are useful for testing image analysis algorithms, and can be used as the basis for spatially-realistic cell simulations using systems such as Virtual Cell and MCell. Perhaps most importantly, CellOrganizer models represent a transportable means of representing the results of High Content Screening (HCS) assays that is not dependent on a specific instrument, assay or cell type. This tutorial will focus on how to use CellOrganizer and how to interface it with other software.
See the tutorial
The tutorial will begin with a brief overview of the conditional structure of the models within CellOrganizer and the system organization. The first part of the tutorial will focus on training generative models. Students are strongly encouraged (but not required) to bring a laptop. Attendees are also encouraged to bring a fluorescent cellular image dataset of their own to use for building a model, but datasets will be available at the tutorial for attendees who do not have one. Ideally, images should be two or three dimensional single cell images (i.e., already segmented) with different fluorescence channels for a fluorescently labeled target protein (ideally a protein showing a punctate or vescular pattern), a cell membrane or cytosolic-labeled marker, and a DNA marker (but these are not strict requirements). The second part will focus on synthesizing cell images from the models and importing the images or model parameters into other software systems. The last part will focus on adding new capabilities to the open source system, such as modules for building new types of components.
Students should leave this session with mastery of the principles behind building probabilistic models from images and practical experience with training and using them with CellOrganizer. They will be able to use them to compare results from different HCS assays using the generative model parameters, and import synthetic images into cell simulation systems.
>> See the tutorial