fMRI data acquisition and storage
Data from the 1.5T Siemens MRI at the Royal University Hospital (RUH) will
make it into your hands for analysis in 5 steps:
- A volunteer (subject) will lie in the MRI and perform their
assigned tasks while the MRI collects image data.(Click
here for detailed procedures for operating the RUH MRI to acquire fMRI data.)
- The data will be automatically
transfered, via DICOM protocol, to the research DICOM linux server
named akasha located in the fMRI equipment room at RUH. DICOM is
an industry standard for medical images meant to allow the easy
transfer of medical images between computers. (Click
here for detailed procedures for transfering and converting DICOM images
to akasha.)
- Using some locally developed software, the DICOM format images will
be converted into spr/sdt (stimulate) and HEAD/BRIK (AFNI) formats and
stored in the directory:
/home/fMRI/subject
where "subject" is the subject number. If you login to akasha as user fMRI you will
only see the numbered directories. After the stimulate/AFNI format data is
created, the DICOM format data will be tarred and compressed.
- The stimulate/AFNI format data will next be copied from
akasha to an appropriate analysis computer - currently our analysis
computers are thalassa and wilder. Once the stimulate/AFNI format raw
data are on the analysis computers, the data will eventually be deleted from akasha's
disks. Jennifer will make a backup copy, on CD-ROM or DVD, of the stimulate/AFNI format data.
(Click
here for detailed instructions on burning the data to cd yourself.)
The data will be copied into the fMRI account into the directory:
researcher/project/subject
(see the discussion on File Naming Conventions below for the conventions on directory
naming)
- You may now analyze your data. Please keep all data under the relevant project directory.
This will make it easy to archive the data when you are through.
Note for BioE 898.3 students: The dataset for your project will
be stored on a Sun computer named triceratops. You will have to
telnet or rlogin to triceratops using procedures that will be
given in class. After you login, the data will be in a directory called
"project898".
In terms of fMRI data storage, a "subject" is one
imaging session with one volunteer. So if you scanned subject number
2403 on November 23, 2148, the data for that particular volunteer would
be in:
/home/fMRI/researcher/project/2403
on your analysis linux workstation.
Many investigators will have multiple
projects going at the same time. For example, investigator borowsky may
have one study on listening and another study on talking. In that case, the
project directories would be:
/home/fMRI/borowsky/listening
/home/fMRI/borowsky/talking
If subject 2501 was part of the listening project then the stimulate/AFNI format raw
data
woiuld be in /fMRI/home/borowsky/listening/2501. And so on. You
will end up with a number of subject number directories stored under each
project.
More abstractly, the following convention for directory naming for
setting up your fMRI data for analysis is suggested:
/home/fMRI/investigator/project/subject
where "investigator" is name of the researcher,
"project" is the name of the project and "subject" is the number of the
subject. Note that the data in subject directories will be named by
task. A task represents an individual imaging "run" of continuous imaging.
So the HEAD and BRIK AFNI files for a "run" will be stored abstractly as:
/home/fMRI/investigator/project/subject/task.HEAD
/home/fMRI/investigator/project/subject/task.BRIK
Note that in the AFNI documentation our term "subject" is referred to
as a "session". Further suggestions for subsequent directory
oragnization are made in the Step-by-step
pages.
Frequently, it will happen that one subject will have participated in
more than one project will they were in the MRI. In that case, Jennifer
will copy the raw data to each of the relevant project directories
and then you should manually delete irrelevant data from the subject directories.
For example, if subject 2501 did task Rhyme for the listening project
and task Guess for the talking project then a copy of /home/fMRI/2501 on
akasha will go to to both
/home/fMRI/borowsky/listening/2501 and /home/fMRI/borowsky/talking/2501
on the analysis computers. You should then go
into /home/fMRI/borowsky/listening/2501 and delete Guess.* and go into
/home/fMRI/borowsky/talking/2501 and delete Rhyme.*. This will help to
conserve disk space.
Note on data acquired on the old MRI sparc hardware prior to
February 2002.
The old Sun workstation formerly known as akasha has been
renamed triceratops and an Intel based linux machine has taken
over the name of akasha. The switch was necessary because the
CPU change in the MRI (from the February 2002 upgrade) makes the data
from the MRI incompatable with the Sun sparc data format (a "big
endian"/"little endian" hassle). You may continue to analyze your old
data on triceratops. The old data on triceratops are
available through the user account "fMRI" and, as need arises, will be stored in the
old directory organization of
/home/fMRI/investigator/date/project/subject where "date" reflects the
date of the data acquisition. All of the old data are archived
to CD-ROM and can be reloaded onto triceratops or onto a linux
machine of your choice - please contact Jennifer Hadley if you need data
reloaded.
Back to U of S fMRI web page.