Quick Links
Overview Documentation
Visualisation Tools
Advanced Image Processing Tools
Pipelining Tools
Statistical Analysis Tools
Methodology Papers:
Programming at the BIC:
Mailing lists
|
Brain Imaging Software: An Overview
A quick note about the conventions used below: where possible I linked
the program being described to its documentation web page. Where
such a page does not exist, however, I simply rendered it in bold
font. To get help on these programs (as well as those where there is
a documentation page) try man [program] or [program]
-help, for example man mincmath or mincmath
-help. If you are unfamiliar with the computational environment
used at the BIC, you should look at the getting started page before continuing any
further.
For a more general overview of our toolkit, it might be worth checking
out some of the presentations give at the 2003 MINC workshop held in
Montreal. Those can be found here. It is
also very much worth your while signing up to the MINC-users
email list (or just browsing the archives).
There are five fundamental types of software programs available here at the BIC. They are, in no particular order:
- Visualisation Tools
- Interactive visualisation tools help you interact with
imaging data directly, manipulate slices, manually segment
images, superimpose different imaging modalities such as PET
and MRI, etc.
- Basic Tools
- And this is where we categorise the rest. They are
non-interactive, usually run quickly and only perform some
minor operations on images. Mincmath and volume_stats
especially come to mind as examples here
- Advanced Image Processing Tools
- The packages that fall under this category are
non-interactive, and perform one algorithm or another on an
input image in order to produce a modified output image. The
four most commonly used tools here are image registration,
non-uniformity correction, classification, and segmentation.
- Pipelining Tools
- These tools (or this tool, as it stands right now) are
designed to manipulate large data-sets, usually performing a
sequence of operations on each image in the data-set. This is
especially useful for clinical trials or large-scale studies.
- Statistical Analysis Tools
- Here you find everything (or at least a few things) related
to performing statistical analysis on medical images, whether
it be measuring atrophy using glim_image or rCBF using EMMA.
Displaying and visualising data is one of the main ways that most
researchers at the Brain Imaging Centre interact with their
data. There are five programs that are predominantly used for this
purpose, each of them useful for a slightly different task.
- Display
- Display is the most
complex of all of the visualisation tools available in the lab. It's
primary uses are the display of multiple imaging modalities (PET,
fMRI, etc.), as well as for manual segmentation of MRIs, aka "Brain
Painting."
- Register
- Register, on the other hand, can be used for
superimposing two volumes as well as performing manual
registrations.
- Postf and xdisp
- Postf and xdisp are
designed to quickly display a single volume, and also allow for the
definition (and gathering of statistics) of ROIs (Regions of
Interest).
- JIV
- Lastly, JIV is a register-like program
written in Java and usable as an applet within a web-browser, which
makes it an ideal tool for collaborations with other sites that are
minc-challenged.
- ray_trace.
- Ray_trace is not interactive; it is designed
for the creation of static images from the command-line. Some of the
advantages of this approach are reproducability (i.e. it is much
easier to specify the angle you want the picture taken from then to
try to maneuver the volume into that position, in display) as well as
cleaner image quality.
There are many utilities available at the BIC which don't nicely fit
into one of the above categories. The list that I will give here is
also by no means comprehensive, nor is it in any particular order.
- mincmath
- The most useful of the basic tools is
in all likelihood mincmath, which allows for most common
mathematical operations on minc volumes (such as addition,
multiplication, thresholding, etc.).
- mincinfo
- mincinfo by default gives you generic information about a minc file, such as it's type, sign, range of data, plus a brief descriptions of dimensions and their order. You can also specify detailed options about what information you would like about the file you are examining, however, which makes it very useful in scripts that need only one bit of information about a file.
- mincheader
- mincheader prints out the entire header of a minc file onto the console - which is very useful for getting a quick glimpse of what non-image information is contained in a file.
- mincedit
- mincedit is a program that allows you to change header
information.
- smooth_matter
- smooth_matter is an application for extracting blurred tissue matter maps from a classified image,
- mincaverage
- ... is for creating an average volume out of several minc files.
- xfminvert, xfmconcat, and xfmtool
- are all programs for manipulating transformation matrices.
The various image processing applications can be considered the heart
of the software at the Brain Imaging Centre. Their dominant goal is
the application of various algorithms to medical images. Below is a
small list of the most important of these programs and their
applications. To reiterate, if the program is in bold and not
hyperlinked then no online documentation exists for it.
- INSECT
- INSECT is the algorithm to separate a structural MRI into it's three tissue types: white matter, gray matter, and CSF. There is some documentation for it under the ANIMAL section.
- ANIMAL
- ANIMAL is designed to label the major anatomical regions (the different lobes, corpus callosum, etc.) of a MRI.
- SEAL
- SEAL stands for "Sulcal Extraction and Labelling" - which pretty much explains its use as well.
- N3
- N3 is the intensity non-uniformity correction algorithm in place here, and its use tends to be an essential first step in any processing sequence.
- cortex extraction
- The cortex extraction algorithm comes in two flavours: a straightforward creation of a cortex useful for tasks such as the masking of the skull and scalp, and a much more compute intensive double-hull algorithm.
- mritotal, mritoself, and other MNI AutoReg tools.
- This family of algorithms linearly register two images to each other. Mritotal registers an MRI to standard Talairach space, and is the most used of them. They are all part of the MNI AutoReg package.
- mincresample
- Resamples an image using a given transform and optionally like another volume.
A piece of software that has started being used with increasing
frequency is the pipelining system, also known as PCS or RPPL. Essentially it is designed to perform a series of
processing steps on a large database of images. At its base it uses
the UCSF Batch system.
There are three main packages available for statistical analysis of
medical images. The first is
- glim_image
- glim_image is designed for performing Voxel Based Morphometry - in other words, it is useful for
examining the changes in segmented tissue matter (white matter, gray
matter, CSF) as it relates to a linear model.
- DOT
- For PET analysis, DOT is the tool of
choice ...
- EMMA
- ... though matlab with the EMMA
interface can also be handy.
- fMRI statistical analysis
- And lastly fMRI
statistical analysis is also done using matlab.
|