See inline responses
On Sat, Dec 1, 2012 at 1:14 PM, Michael Froelich <[log in to unmask]> wrote:
> Hello:
>
> I plan to perform a study to test activation in four regions of interest in the brain, two involved in the coordination of a motor response (M1-2) and two that are involved in the emotional response (E1-2) to the pain stimulus. I was thinking about using some sort of connectivity analysis to demonstrate that these regions of interest are connected to the primary and secondary (S1-2) somatosensory cortex. So my "connectivity hypotheses" are:
>
> A. Areas S1-2 affect both M1-2 (motor) and E1-2 (emotion)
> B. Areas S1-2 affect both M1-2 (motor) and E1-2 (emotion) + area E1-2 (emotion) modulates M1-2 (motor).
It is unclear what you mean by "modulate". DCM, PPI, etc. all test
whether there is differential connectivity between different
tasks/brain states. I can't tell if you are wanting to show task
modulations or simply regional connectivity. If its the latter, then
you should use resting state data and simple seed based connectivity
will answer your questions. If you want task based connectivity, I'd
suggest that you use DCM since it looks like your proposing that E1-2
and S1-2 both modulate M1-2. While some people are not huge fans of
DCM, I think there are some good applications, such as testing
hypothesis B. However,
> The only modulating influence might be the type of pain stimulus used (electrical versus cold versus heat) but typically I have done this in separate sessions.
>
> My questions are:
> 1.) Which type of connectivity analysis would be most suitable (DCM, coherence, Granger causality)?
A limited number of regions allows you to use DCM. PPI might also be
option if you want to see where else S1-2 is connected between
different conditions. Granger is slightly complicated as it uses the
BOLD response. In most cases, you can't infer that one region causes
another another region, but that one region is active later in the
task compared to the earlier region. This is mainly due to a time
scale issue though.
> 2.) Is a block design ok if I want to show connectivity or do i have to do an event related design?
Either will work. If you want to know what regions are connected, then
resting state is the preferred method. It is only differential
connectivity that requires a task. A block design would be fine for
PPI/DCM. You could use event-related data, but you'd need to make sure
you have a sufficient number of events (>32) and make sure that each
run has numerous events if you want to use DCM or PPI. One other
caveat with PPI and event-related designs, make sure that you specify
a duration. While you can treat the task as an instantaneous event,
using it in PPI is bad because you end up looking at the interaction
during the first 1/32 of TR (SPM default), rather than a more
reasonable time that would be the actual duration of the event. For
task activity, the duration up to 2seconds has very little impact on
the shape of the HRF.
Hope this helps.
>
> I appreciate any expert feedback.
>
>
> Michael
|