Microtrack

From VISTA LAB WIKI

Jump to: navigation, search

The microtrack project is centered around methods for predicting diffusion data from a set of estimated tracts. The project also aims to use this core technology in several ways. In one application, the method can be used to estimate tract properties such as axon diameter, g-ratio, and axon volume from diffusion data. In a second application the method can be used to test hypotheses about the most likely solutions for tracts.

This computation is closes the loop between tractography and diffusion measurements. By understanding how well different tract estimates predict the diffusion data, we can compare the models and evaluate whether the tractography estimates and diffusion data are sufficient to derive fundamental tissue properties.

Contents

[edit] Introduction

At present the Microtrack code depends on vistasoft and the Camino toolkit from D. Alexander's group. Both of these packages are installed on the Stanford Linux boxes (but not Windows).

The tutorial and validation scripts use a fixed data set that is stored in the vistadata repository.

Sherbondy wrote a MICCAI paper with D. Alexander's group describing certain aspects of the Microtrack project.

The first analyses at Stanford using these tools is an analysis of TBSS (Tsang). In that analysis we used Microtrack to create data sets with controlled and biologically plausible diffusion changes. We then asked whether TBSS or VBM was capable of detecting these known signal changes.

The next projects we hope to undertake involve testing alternative models of specific tracts, such as the quality of the diffusion fit to data assuming the right arcuate is present or the right arcuate is absent (Yeatman).

[edit] Software and data install

We only run this code using the bash shell. We will try to figure out how to make it run under c-shell.

This code has been run successfully on bash with Matlab 2008a, 2009a, 2010a

It has been run on Mac OS 10.6+ and Linux (but not Windows).

You must have mrVista on your path to run the microtrack code.

You should also have vistadata checked out (in particular the mictDiffusion subdirectory). This should be on your path. These are used to run different tutorial scripts. The key function in there is mrvDataRootPath


[edit] Camino setup

Camino should be installed on the Linux system by somebody knowledgeable (at this point Sherbondy). If you are running bash, then you must edit your .bashrc file to include

[edit] .bashrc file entries

# setup stuff for microtrack scripts
export MICT_HOME=/home/sherbond
#
#Microtrack
export PATH="$PATH:/home/YOURHOMEandPATH/Matlab/VISTASOFT/vistaproj/microtrack/bin"
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:$MICT_HOME/lib"
#
#Camino
export MANPATH="$MICT_HOME/src/camino/man/:$MANPATH"
export PATH="$MICT_HOME/src/camino/bin:$PATH"
export PATH="$MICT_HOME/src/camino/SGE:$PATH"
export PATH="~/Matlab/VISTASOFT/vistasoft/trunk/mrDiffusion/camino:$PATH"
export CAMINO_HEAP_SIZE=4000

[edit] .cshrc file entries

# Camino
setenv MICT_HOME /home/sherbond
setenv PATH {$MICT_HOME}/src/microtrack/nfg/bin:{$MICT_HOME}/src/microtrack/matlab:{$PATH}
#setenv MANPATH {$MICT_HOME}/src/camino/man/
setenv PATH {$MICT_HOME}/src/camino/bin:{$PATH}
setenv PATH {$MICT_HOME}/src/camino/SGE:{$PATH}
setenv PATH {$VISTASOFT}/mrDiffusion/camino:{$PATH}
setenv CAMINO_HEAP_SIZE 4000

[edit] Vistaproj setup

[edit] Vistadata setup

You can use the vistadata that has been checked out onto the Linux system in /usr/local/vistadata.

If that data repository is not up to date, please do an svn update, or ask Bob, Brian, Michael Perry, or someone to help you with it.

[edit] Tutorial overview

[edit] Forward modeling

If you already have a segmentation and fibers (as are checked in to vistadata), then you can run the t_mictForwardModeling.m script. This will step you through the process of changing the diffusion data to test different ideas. We will explain applications of this process below (in Applications).

[edit] Creating a segmentation

% If you need a segmentation file, do this
% 
% pathways_filename  = fullfile(top_dir,subj_dir,'dti06trilinrt','fibers','CCocc_clean.mat');
% fg = dtiReadFibers(pathways_filename);
% ni = readFileNifti(mictImages.forceNiftiName(og_raw_name));
% fd = dtiComputeFiberDensityNoGUI(fg, ni.qto_xyz, ni.dim(1:3), 1, 1);
% fd(fd>0) = 1;
% dtiWriteNiftiWrapper(fd,ni.qto_xyz,mictImages.forceNiftiName(og_seg_filename));
% showMontage(fd)

[edit] Creating fibers

[edit] Connectome estimation

If you have a segmentation and you want to estimate a connectome (full set of connections plus microstructure), then you can follow the instructions here.

Here we explain the overview using some of the tutorial files in microTrack directories.

[edit] Directories and functions

[edit] bin

This contains microtrack_sim, which is a executable that runs the forward model code.

[edit] Scripts

Inside of scripts is a tutorial folder and various examples in which we run the microtrack code.

For an overview of the forward modeling process step through the script t_mictTutForwardModeling.m. There are many comments in the header of that script and throughout to explain the basic idea. We do not copy that here (though perhaps we should link to it) in order to make sure there are no differences.


The convention for new tutorials is to use t_XXX. Scripts that are not tutorials, are s_XXXX.

[edit] Applications

[edit] TBSS Example

Has its own page

[edit] Hypothesis testing about tracts and their properties

[edit] Visualizing how well a tract predicts the diffusion data

[edit] TODO

  • Make it run under c-shell, not just bash.
  • See if we want to make it run under windows

[edit] References

Personal tools