A. Designing a study
Designing an fMRI study is a daunting task. Our fMRI experiment
design tool is available here. Note that with fMRI you have to tradeoff
predictability with statistical efficiency. Completely random event
related designs will inherently have low power, requiring long scanning
sessions. While block designs offer optimal statistical power, but the
particpant can anticipate the upcoming type of condition. Note that
this softwware only covers a few basic designs. The software uses the
canonical hemodynamic response function used by SPM to convolve events,
reporting efficiency as the mean variance accross conditions. This
efficiency is only relative - if you change the TR or number of
volumes, many efficiency parameters will alter. Also note that this
software generates a general efficiency value (the predictable variance
of a condition), but does not calculate whether different conditions
create independent signals. Finally, note that this software (like most
analysis packages) assumes completely linear addition of singal from
multiple events.
For all designs, use set the TR (time required for each 3D volume of
data) the number of volumes and the number of conditions. Then press
the icon that corresponds to the design you wish to create. If you like
a design, press the 'i' button to see the precise order of conditions
and trial onset times.
- Block Design: Attempts to create a nearly optimal fMRI
design. . The software will create blocks that are approximately
12 seconds in duration. Note that conditions do not repeat to
subsequent blocks (e.g. if you have three conditions, than a block of
condition 1 will either be followed by condition 2 or condition 3).
Adjustable values are:
- minISI: Time between trials
- Permuted Block Design: A block design is created, and then a subset of trials are randomly swapped.
- minISI: Time between trials
- Permutations: Low numbers of permutations will
essentially generate a raw block design (high power, but highly
predictable). A large number of permutations will cause random trial
order but lower statisstical power.
- Fixed Interstimulus Interval Event Related Design. This
generates a series of trials with a fixed time between trials, but
random trial order. Note: do not create Fixed ISI trials for studies
with only one condition - as the result will be very low power (use
variable ISI instead).
- meanISI: Time between trials
- Iterations: Computes multiple trial orders and reports
the order with the best statistical power. Low numbers result in random
orders, but low power. A very high number will essentially begin to
resemble a block design, becoming predictable.
- Variable Interstimulus Interval Event Related Design. Time
between trials follows an exponential distribution, with many rapid
trials and a few trials with long rests. This essentially creates
effects similar to blocks, without much predictablity. For experiments
with one condition this offers much better power than fixed-ISI
designs. However, for multiple condition experiments this design tends
to offer lower power (as fixed ISI designs can present stimuli as fast
as people can complete them, while the time lags in variable designs
reduce the number of trials. However, multi-condition variable ISI
events may be useful for extracting the true shape of the hemodynamic
response in a study.
- minISI: minimum amount of time between trials - essentially how fast the participant can execute the task.
- meanISI: Average time between trials
- Iterations:
Computes multiple trial orders and reports the order with the best
statistical power. Low numbers result in random orders, but low power.
A very high number will essentially begin to resemble a block design,
becoming predictable.
B. Presenting stimuli
Our presentation computer
runs WindowsXP and uses
optical response buttons, ceramic headphones and a computer projector
(optimal resolution is 1024x768) for behavioral presentation. Most
studies are run using EPrime running. A reverse-KVM allows
experimenters to run these studies from either the MRI console room or
from the research lab. A custom built circuit is able to synchronize
experiments with the scanner, as well as providing accurate data
logging.
Sample experiments are stored inside "My Documents\My Experiments". We
have created a sample block design as well as an example event related
task (a block design with 35% permutation rate). Both experiments
present an arrow flashing on the screen - the arrow points either left,
right or up. The task is to press the left index button in response to
the left arrow, the right index finger for the right arrow, and to do
nothing in response to the up arrow. You can make a copy of these
studies to modify.
- Details for having EPrime triggered by the fMRI scan are on my data logger web page.
- Details for having EPrime send event-onset data to our data logger are found on my data logger web page.
- Make sure that your EPrime events use cumulative, not
event timing. If you use event timing, trials will tend to take longer
than expected, and the duration of the study will be longer than
designed with a slightly random duration.
- We use the BrainLogics fiber optic interface console.
- We have special ceramic fMRI compatible headphones to present auditory responses.
|
 |
C. Data logging
Our Data logging system is
designed to record the onset of each 3D volume of fMRI data, as well as
recording when behavioral stimuli were presented. It automatically
saves files, so you just have to ensure that the datalogging computer
is running and that the ezLog program is running.
- Note that this program records the input from the
presentation computer (described above). Your experiment should send a
unique signal for each condition you wish to analyze. Possible values
are 0..15, with 0 being used to denote the offset of a signal.
- Data files are saved after 20 seconds of inactivity.
Files are named for the timing of the first fMRI pulse. For example, a
study that started at1:32pm on January 9, 2006 would be saved as 20060109_133248.txt.
|
|
D. Acquiring data
We have a 3T Siemens Trio with TIM and a 12-channel headcoil. We have tested protocols available, for most fMRI studies you will only need to adjust the number of volumes you wish to acquire.
- Always make sure to check health and safety issues.
- Ensure proper ear protection.
- Check for metal before entering the magnet hall.
|
|
E. Analyzing a study
By using our SSH server,
you can download and analyze your data on any computer you want. The
McCausland center includes two data analysis computers - one runs
Centos Linux (and also doubles as our RAID SSH server) and the other
runs Windows. Both sport dual-core Athlon-64 processors, so they are
excellent for data analysis. They include MRIcro and FSL, so you can
simply log on and process your data. Note that when you are finished
using the Windows machine you should probably back up your data to the
SSH server, as disk space on this machine must be shared by all users.
On the Linux machine, type 'FSL' and 'startmricro' from a command
prompt to launch the brain imaging software. On the Windows machine,
MRIcro and FSL icons appear on the desktop.
Here are a couple of tutorials to get you started:
|
|
|
