Step by step instructions for analyzing fMRI data with BOLDfold
and AFNI
The following step-by-step guide represents our current (2002)
standard approach to fMRI data analysis for data acquired in a repeated
event-related experimental design. As the fMRI research program grows
at the U of S, other methods for data analysis will be adopted.
The following guide was written by Ben Norris and Greg Kraushaar in the
Summer of 2002. For a fairly detailed idea of what you want to generate with
the analysis, see the Example Results web page
- Following the procedures given on the data acquisition and storage
page, make a copy of your data for analysis following the
project/subject directory structure organization.
- Make sure that the appropriate "*.std" files are in their
corresponding folder, as well as any anatomical BRIK and HEAD files, or
functional BRIK and HEAD files. If there are no anatomical BRIK or HEAD
files, you will need to make anatomical AFNI datasets from the
corresponding *.sdt files, using the program "image". Rename all your
files to include the participant's ID number (see the Example
Naming page for an example on how to rename your files for an
experimental design that is multi-factorial with respect to the given
tasks). It is also a good idea to include in your filenames the specific
condition that the participant was performing. For instance, if participant
23 completed a task that involved a High level of electric shock,
accompanied with a Happy mood, you would name the file something along
the lines of HighHappy23.sdt.
- Run Gord's super-cool imaging program, by typing "image" at the
command prompt. Use menu #23 (fMRI) and then the motion correction
option (#7) to align all functional volumes. It is recommeded that you
align your datasets to volume 5 or above; never align on volume 1 since
that volume's spin data are not yet in the steady state caused by the
imaging repetitions (TR). The file name to enter is the functional file prefix,
without extension (e.g. HighHappy23). Once
you are finished, exit the Image program. You should now have a new set
of sdt files, with an "F" suffix added.
These are the motion corrected files (e.g. HighHappy23F.sdt).
- Run the BOLDfold program, by typing "bfold" at the command prompt.
You must be in the directory where your files are contained to do this.
BOLDfold will ask you what file you want to run the calculations on.
You will want to run the BOLDfold transformations on both your motion
corrected, as well as your uncorrected data. BOLDfold will then prompt
you to enter certain design specifications, so make sure to have them
handy. Refer to the BOLDfold steps manual. The BOLDfold
program will produce an output of various "sdt" _BF maps, each with
different statistic values stored at each voxel, as well as one _BF map
in AFNI format (per task). It is your responsibility to ensure that all
tasks that each participant completed are BOLDfolded.
- Using the online guide to help you out, write all AFNI anatomical
and functional datasets to Talairach orientation (+tlrc). The online
guide can be found on the AFNI documentation
web page under the "How to make an Average Brain" section (Step 1 only).
- From here, you will need to prepare the BoldFold files for AFNI's
3dANOVA2 program if your experimental design requires an ANOVA
comparison of stimulus (task) conditions. To do this, you will need to
censor, blur, and clip your data. Refer to ANOVA
steps on how to go about preparing your data and running the
ANOVA program. For other experminetal designs you may want to do a t-test
or a simple subtraction or quantify the hemodynamic response in
pre-specified regions of interest. For those experimental designs, a perusal
of the AFNI documentation should help you to
set a method of analysis.
- Proceed to do cluster average time courses for each of the clusters
in the brain, at each eta cutoff. For how to do this, see Cluster Steps web page.
For guides on how to use AFNI,
see the AFNI documentation webpage.
Once you have saved an average time course of all the clusters in the
brain, now would be a good time to decide which are due to noise and
which are due to actual activation. Find the center of mass for each
cluster, and go back to AFNI's GUI to report the Anatomical and
Brodmann areas for each cluster. Be aware that the report is only
approximate, because when you Talairached your anatomical dataset, this
was only a "rough" anatomical dataset, and the "Where Am I?" program uses mm from origin as a
guide to where in the brain you are. Since each brain is highly
variable in its anatomy, when you report a cluster at, say 22mm from
the origin, this may not correspond exactly to the reported structure
in the Talairach atlas.
- Go get yourself a Coke or Mellow Yellow slush at the Cove for
all of your hard work. Well done!