NCMIR - UCSD

Microscopy Image Analysis and Visualization

The goal of this collaboration with the UCSD National Center for Microscopy and Imaging Research (NCMIR) is to build tools for the analysis of electron microscope tomography data sets. NCMIR develops cutting edge three-dimensional imaging and analysis technologies to help biomedical researchers understand biological structure and function relationships in cells and tissues through a range of scales that encompass macromolecular complexes, organelles, and multi-component structures like synapses.

We have identified a set of strategies that will aid in the pursuit of this collaboration. The strategies pertain to the development of tools and the plans for using them to help in the NCMIR projects.

User-Assisted Segmentation: Extend our current slice-by-slice tools to include three-dimensional models and to take into account direct segmentation strategies.
Statistical Filtering and Reconstruction: Integrate filtering tools that rely on neighborhood statistics in the reconstruction and filtering of EMT data.
Volume Visualization: Integrate the use of neighborhood statistics into the volume visualization of difficult data sets such as gap junction and actin fiber EMT.
Data Analysis: Validate on specific data sets. We will work with investigators at NCMIR (Maryann Martone for the dendrite project and Gina Sosinsky for the GJ and actin project) to build data sets to pursue specific scientific questions. In the case of the dendrites, we will identify a particular set of EMT volumes and a specific hypothesis to validate the computational approaches and to refine a set of tools to help with that process. In the case of the GJ study we will work with NCMIR and incrementally refine the approach in order to produce pictures that capture the strutural phenomena they are studying.

A volume rendering shows the use of multidimensional transfer functions for visualizing EMT data of a spiny dendrite.

A surface rendering of a spiny dendrite extracted automatically from a EMT volume. The segmentation breaks down near the edges of the volume.

A snapshot of a SCIRun Power App that shows the use of interactive level sets to segment an EMT volume of mitochondria. The left panel shows one of a series of tabs that provide control panels for the components of the pipeline. Images clockwise from the upper left are: a slice of the segmentation, the speed function, the filtered image, and a volume rendering of the current three-dimensional segmentation.