Hi. This is a tough problem that I face, also. Many basal ganglia
"functional units" are smaller than current voxel sizes in fMRI
studies. A voxel size of 3.5 x 3.5 x3.5 mm is not small enough,
really. It's close, and adequate for some things like finger tapping.
I don't smooth to look at individuals. Then, I apply the "double
the voxel size" rule to reduce noise. I've seen larger smoothing
kernels work, though, in the caudate. In the few subjects I've
examined closely, the results are almost always the same with and
without smoothing. The smoothed image is prettier. You might try
doing a series of smoothing steps with your data--like the one shown in
the book Stephen Smith edited along with others. You'll see when
things drop out or appear. I think that "there isn't a set answer to
this..." In deoxyglucose autoradiography studies that I've done in
rodents, a different amount of smoothing was needed for the cortex and
the striatum to make the signal VERY clear.
Best wishes,
Lucy Brown
On Friday, January 24, 2003, at 04:33 AM, Stephen Smith wrote:
> Hi. I'm afraid there isn't a set answer to this - the question should
> take
> into account not just expected activation area size but also
> cross-subject
> spatial variability....
>
> Sorry to be vague! Maybe people with more experience in these brain
> areas
> might add something here.
>
> Thanks, Steve.
>
>
>
> On Thu, 23 Jan 2003, Goldfine, Andrew (NIH/NINDS) wrote:
>
>> Another question...In my study group I'm looking for activations in
>> areas
>> including basal ganglia and higher level motor areas. What is the
>> ideal
>> amount of smoothing to use in FEAT for these regions? Thanks again
>> for your
>> help.
>>
Lucy L. Brown Mail: 1410 Pelham Parkway south
Department of Neurology Room 810
Department of Neuroscience Bronx NY 10461
Albert Einstein College of Medicine Phone: 718-430-3728
Bronx, NY 10461
website: www.bette.aecom.yu.edu
|