They are unweighted. In theory, you can get a better gaussian approximation if you weight, but I don't know of a weighting scheme that works all that well. I've implemented Moo's heat kernel smoothing and was not happy with the results (also see a NI paper by Don Hagler dealing with the HK shortcomings).

doug

DRC SPM wrote:

Hi Doug,

Thanks for your speedy answers to both my smoothing questions. Can I trouble you one more?!

Are the connected-neighbour smoothing iterations unweighted averages, or weighted in some way? I'd guess that the approximation to Gaussian smoothing would be better if there was something like the exp(-distance^2) weighting in there (with the distance just being the length of the edge between the neighbours) but maybe this complicates things? I think this is more or less what is done in MK Chung's heat-kernel smoothing, but I haven't looked into the details... http://www.stat.wisc.edu/~mchung/softwares/hk/hk_smooth.m

Many thanks, Ged

2009/8/11 Douglas N Greve greve@nmr.mgh.harvard.edu:

...

Your 2nd description is correct. Note that the value of vertices outside of the mask are NOT changed (ie, they are not set to 0).

doug

DRC SPM wrote:

...

Dear FreeSurfers,

Please can you explain how mri_surf2surf's --cortex option for smoothing is implemented?

I assumed it would be along the following lines:

• mask the data to zero-out the medial wall
• smooth this masked data
• smooth the mask itself
• divide the smoothed data by the smoothed mask
• (perhaps) re-mask the data

But I've tried doing that long-hand (i.e. calling mri_surf2surf without the --cortex flag for each smoothing, and using l load_mgh and save_mgh in MATLAB for the rest), and it doesn't seem to quite match the results of the --cortex flag.

Now I'm guessing that it instead might do something like the following. For each connected-neighbour iteration of the smoothing, only include the neighbour in the average if it is inside the cortex (not medial wall). However, I can't easily/lazily reimplement this to check, so could someone reveal the truth? ;-)

Many thanks, Ged _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422

In order to help us help you, please follow the steps in: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting