Hi Lars,
I think the formula is:
fwhm = (2.3*sqrt(niter*2/pi));
Doug's in Florence, so may not reply right away. I think there's also a
correction for the average surface as it has less surface area than any of
the subjects in the average.
cheers,
Bruce
p.s. here's the code snippet from mris_surf2surf.c:
int MRISfwhm2niters(double fwhm, MRIS *surf)
{
double avgvtxarea, gstd;
int niters;
MRIScomputeMetricProperties(surf);
avgvtxarea = surf->total_area/surf->nvertices;
if(surf->group_avg_surface_area > 0){
// This should be ok even if metric properties have been scaled ??
printf("INFO: fwhm2niters: Fixing group surface area\n");
avgvtxarea *= (surf->group_avg_surface_area/surf->total_area);
}
gstd = fwhm/sqrt(log(256.0));
//1.14 is a fudge factor based on empirical fit of nearest neighbor
niters = floor(1.14*(4*PI*(gstd*gstd))/(7*avgvtxarea) + 0.5);
return(niters);
}
On Mon, 12 Jun 2006, Lars M. Rimol wrote:
Hi,
In the --help output of mri_surf2surf, the following explanation is given
for --fwhm-src fwhmsrc:
Smooth the source or target with a gaussian with the given fwhm (mm).
This is
actually an approximation done using iterative nearest neighbor
smoothing.
The number of iterations is computed based on the white surface. This
method is similar to heat kernel smoothing. This will give the same
results as --nsmooth-{in,out}, but automatically computes the the
number of iterations based on the desired fwhm.
What is the equation used to compute the number of iterations, and is it
valid for an average surface?
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
