general-discussion > Regress out Visual activations
Showing 1-2 of 2 posts
Results per page:
Jan 27, 2012  09:01 AM | Halleh Ghaderi
Regress out Visual activations
Hi everyone,
In my fMRI design, I provide a visual stimulus (STOP/Go sign) and I ask the subjects to perform a motor task (finger tapping) . However, after the statistical analysis, I realized that the activation due to visual stimulus is so wide that gets extended to the motor areas of thalamus and cerebellum. Is there any way that I can eliminate the visual signal from my fMRI post processing.
I mean, Is there anyway to isolate the signal due to visual stimulus signal and regress out that signal in fMRI post processing such that I don't see any effect (activation)  from Visual cues.
I will greatly appreciate any help !
I have attached a file that shows the huge activation I get from Visual and only small activations from my motor task
Jan 31, 2012  06:01 PM | Pierre Bellec
RE: Regress out Visual activations
Dear Halleh,

By visualizing the t-stats map you captured, my main concern would be the large t values around the brain, located in the sagital sinus and the meninges. Was there a large motion correlated with the task here ? I guess it's not too uncommon in motor paradigms but scary anyway. A collaborator of mine (Dr Pierre Orban) suggested that the coregistration between the EPI and the T1 maybe completely off, in which case the "motion artefacts" would fall in the sensorimotor cortex and the visual activation would be in the cerebellum. Certainly worth checking on the mean image ...

Regarding the visual component of the task, you can enter it in the model with a different duration than the motor task, and then regress out the visual regressor from the motor one. With this approach you will loose a lot of motor activation that cannot be disambiguated from the visual one, but you will clean-up the visual component of the task. However, the most classical way to adress your problem is to ignore it, and not to interpret the activations in the visual cortex. Usually, the contrasts of interest lie between a task and a control, and the visual component should be identical in both conditions. In this case, the visual activation will cancel itself out.

I hope this helps, good luck with your analysis,

Pierre Bellec