Print

Print
Hi,

To start with, you can apply an anisotropic smoothing by creating a custom kernel file.  This needs to be a nifti image and will be fairly small (the size of the non-zero kernel) and filled with whatever values you wish it to have).  We recommend that it has the same voxel size as the image you will apply it to.  You can either set-up the kernel in something like matlab, or you could set values by hand in FSLView, or any other means.

As for applying it to images - if it is just the same anisotropic kernel everywhere, then that is fine.  If you want to apply this within an ROI but not elsewhere (or vary it in any other way) then you will have non-stationary noise which causes problems for most multiple-comparison correction techniques that assume that each voxel has been treated equivalently.  That being said, there is known non-stationary noise in the images themselves, and so to small degrees it won't matter too much.  Just avoid really big changes - or avoid any at all and apply the same (anisotropic) smoothing everywhere.

I hope this helps.
All the best,
Mark

 
On 1 Oct 2012, at 22:20, Joshua Lee <[log in to unmask]> wrote:

Thanks.
Generally, I agree that is a good idea. However, I am concerned about using a typical scheme because in my hi-res fmri study, the adjacent regions I am interested in are smaller than the typical smoothing kernel. Using even 4mm smoothing kernel on the whole image would mean that activation in region X is due to adjacent sub-region y...Further, without smoothing, the data is noisy.
-
Josh

On Mon, Oct 1, 2012 at 1:52 PM, Marenco, Stefano (NIH/NIMH) [E] <[log in to unmask]> wrote:

Not sure ….I would apply the same procedure to the whole image…

 

Stefano Marenco, MD

NIMH/CBDB

10 Center Drive, Bldg 10 room 3C103

Bethesda MD 20892

Tel 301 435-8964

Fax 301 480-7795

Email: [log in to unmask]

 

From: Joshua Lee [mailto:[log in to unmask]]
Sent: Monday, October 01, 2012 4:15 PM
To: [log in to unmask]
Subject: Re: [FSL] Anistropic Spatial Smoothing Kernels

 

Hey thanks!
Might I ask another question?

If one were to apply a standard spatial smoothing kernel to just subsections of a functional image through the use of masks (e.g. use a mask to Gaussiansmooth just one subfield of the hippocampus), would the resulting smoothed image be problematic for use in analyses?

Joshua

 

On Mon, Oct 1, 2012 at 8:02 AM, Marenco, Stefano (NIH/NIMH) [E] <[log in to unmask]> wrote:

Afni should have something to that effect….

 

Stefano Marenco, MD

NIMH/CBDB

10 Center Drive, Bldg 10 room 3C103

Bethesda MD 20892

Tel 301 435-8964

Fax 301 480-7795

Email: [log in to unmask]

 

From: Joshua Lee [mailto:[log in to unmask]]
Sent: Friday, September 28, 2012 4:15 PM
To: [log in to unmask]
Subject: [FSL] Anistropic Spatial Smoothing Kernels

 

Hi all,
I am interested in using an anistropic kernel to spatially smooth high-resolution functional data.  For example I'd want a 8mm FWHM in the z direction, but only 2 mm in z and y directions. In the structure I am examining, this appears to be a good way of increasing SNR while retaining spatial localization of activations to anatomical structures.

Has anyone done something like this before, and if so might I profitably use fslmaths to do so?

-

Josh