FSL Evaluation and Example Data Suite
To accompany FSL version 1.3
FMRIB Image Analysis Group, Oxford University
|
|
INTRODUCTION -
DOWNLOAD -
RUNNING THE EVALUATION -
USING THE EXAMPLE DATA
INTRODUCTION
This package performs two functions - it tests whether your FSL tools
are working properly, and provides example data to try running FSL on.
The evaluation of FSL is carried out by a single script
which normally takes between 1 and 3 hours to run (depending on your
computer), testing every major tool in FSL on example data. It
compares the output of each test with output data which is supplied
with this package, and reports significant differences as failures.
The example data includes both example input and example
output data. There is a detailed description below of how to analyse
the input data - this is a quick way of starting to use FSL.
For up-to-date information regarding FSL see the FSL home page. For support
relating to FSL or related theory, email the FSL email list or, even better,
join the FSL email
list.
DOWNLOADING THE FSL EVALUATION AND EXAMPLE DATA SUITE
If you have not already downloaded the FSL Evaluation and Example
Data Suite then do
so now.
Now unpack the distribution (it doesn't matter where you do this)
by typing
gunzip fsl-feeds-1.3.tar.gz
tar xvf fsl-feeds-1.3.tar
RUNNING THE EVALUATION
- Make sure the FSLDIR environment variable points to the top level
fsl directory. This might be, for example, /usr/local/fsl or
/home/bart/fsl :
- bash users: export FSLDIR=/usr/local/fsl; export PATH=$PATH:$FSLDIR/bin
- tcsh users: setenv FSLDIR /usr/local/fsl; setenv PATH
${PATH}:$FSLDIR/bin; rehash
- cd into the top level directory feeds.
- Type time ./RUN
- (If you are interested in comparing how fast FSL runs on your
computer with other computers, the final
output - the time marked with a u - tells you how many CPU
seconds or minutes the test took to run.)
The script RUN uses input data from the data
subdirectory, and saved all output in the results
subdirectory. It then compares this output with the test output data
in the data subdirectory. (Note that test input and output data
is mixed together in data, as in some cases the output from one
tool was used as input to another one.) So - each tool's output is
compared with the example data ouput and a percentage error is
generated - this may not be exactly zero as different hardware
platforms can give slightly different results without this being
classified as an "error". Each tool's error is scaled so that a
reported error of 1% is considered a failure.
If you get any failures, you may want to send us the complete text
output from RUN, and also even the complete results
directory.
USING THE EXAMPLE DATA TO LEARN FSL
INTRODUCTION
We now explain how the different tools in FSL can be run on the
example data provided in the data subdirectory. We strongly
suggest that you work with a copy of data rather than the
original, so that you can always go back to the original data if you
need to. You can do this by typing (inside feeds)
cp -r
data examples cd examples
and theh work with the
files inside examples.
The instructions given below should produce the same output as
provided in data.
Now make sure the FSLDIR environment variable points to the top
level fsl directory. This might be, for example, /usr/local/fsl or
/home/bart/fsl :
- bash users: export FSLDIR=/usr/local/fsl; export PATH=$PATH:$FSLDIR/bin
- tcsh users: setenv FSLDIR /usr/local/fsl; setenv PATH
${PATH}:$FSLDIR/bin; rehash
To start the main FSL GUI, type fsl.
To view the output of each tool, either load the output image(s)
into your favourite Analyze image viewer (e.g. MEDx), or use the
simple (non-interactive) display program slices which is in the
fsl/bin directory (the setup commands above should have already placed
this within your path). Where links appear below, these mostly point
to 2D GIF images created by running slicer on the relevant
Analyze format 3D image. (slicer is a command-line program which
takes a 3D Analyze format image and produces a 2D GIF image with
various slices at various orientations from the input image;
slices is a script which calls slicer and then starts up a 2D
image viewer to show you the GIF image.)
BET
Set the Input image to be structural.hdr and
press OK. The output will be structural_brain.hdr. You
will see a message on your terminal when BET has finished.
SUSAN
Set the Input image to be structural.hdr. Set
the Brightness threshold to 2000 (this is greater than
the noise level and less than the grey-white contrast in the input
image). Set the Mask SD to 2 (this sets the mask half-width to
be 2mm). Press OK. The output will be structural_susan.hdr.
FAST
Set the Input image to be structural_brain.hdr
(i.e. it is important to have run BET first). Turn on the Partial
volume maps, Restored input and Bias field optional
output images. Press Go. The outputs will be structural_brain_seg.hdr,
structural_brain_bias.hdr,
structural_brain_restore.hdr,
structural_brain_pve_0.hdr,
structural_brain_pve_1.hdr
and structural_brain_pve_2.hdr.
FLIRT
Set the Input image to be structural_brain.hdr. Set
the Output image to be structural_brain2standard. Under
Advanced Options -> Interpolation select Sinc. Press
Go.
FUGUE
FUGUE does not yet have a GUI. On the command line type
prelude -c fieldmap -u unwrapped_phase
This runs the phase map unwrapping. Now type
fugue -i epi -p unwrapped_phase -d 0.295 -u unwarped_epi
This runs the unwarping of the input epi image.
SIENA
SIENA does not yet have a GUI. On the command line type
sienax
structural
This runs the SIENAX cross-sectional
(single-time-point) atrophy script, producing as colour-coded atrophy
output, structural_render.
At present no example data is included for running longitudinal
(two-time-point) atrophy measurement.
FEAT
- Press Select 4D data and select fmri.hdr.
- Press Full model setup to setup the GLM details.
- Change the Number of EVs to 2.
- Setup EV1 (the visual stimulation timing): set Off
to 30, On to 30 and Phase to
30. This describes a square wave of total period 60s, starting
with an ON period, hence the phase setting.
- Setup EV2 (the auditory stimulation timing): set Off
to 45, On to 45 and Phase to
45. This describes a square wave of total period 90s, starting
with an ON period, hence the phase setting.
- Now setup the Contrasts. Set the Number of contrasts
to 2, set the first (C1) to [1 0] and the second
(C2) to [0 1]. Thus the first output colour overlay
image produced will show visual activation as only EV1 is used, and
the second will show only auditory activation.
- Now setup the F-tests. Set the Number of F-tests to
1. Select both contrasts. Thus the third output colour overlay
image produced will show where either visual or auditory activation
occurs (i.e. will show both on a single image).
- Press Done to finish the model setup.
- Set the High pass filter cutoff to 100. Although it
is common to set this to 1.5 times the maximum stimulation period
(in this case 90*1.5=135), the nonlinear filter used in FEAT
has quite a slow roll-off above the cutoff frequency, so in fact
setting this to just over the 90s period time is fine.
- Press the triangle next to Advanced options, and press the
button next to Registration. Note that the default settings in
Pre-stats and Thresholding & rendering can be left as
they are.
- In the Registration section, select the Subject's high
resolution image. Set this to structural_brain.hdr.
- You are now ready to run FEAT. Press Go. As FEAT completes
the different stages of processing, you will see messages appear on
your terminal. When it has finished, the final messages will tell you
the file name of a web page which you can view with your web browser
to see the results.
MELODIC
Set the 4D input data to be fmri.hdr. Note that this is
the same raw data as was input to FEAT - normally you would ideally
want to have done some pre-processing to the data before running
MELODIC - see the MELODIC help page for more information on
this. Press Go. When MELODIC has finished, the final messages
will tell you the file name of a web page which you can view with your
web browser to see the results.
Steve Smith
FMRIB Analysis Group
Copyright © 2000, University of Oxford