Print

Print
Hey Matt,

I finally got around to playing with this a little more, but I'm still experiencing the same hang-up on the smoothing step. film_gls seems to just be repeating the smoothing process over and over . For example, I see this in the output every 20 minutes or so:

Smoothing at vertex 163500 ...
Smoothing at vertex 163600 ...
Smoothing at vertex 163700 ...
Smoothing at vertex 163800 ...
.do smooth
done distances
Smoothing at vertex 0 ...
Smoothing at vertex 100 ...
Smoothing at vertex 200 ...
Smoothing at vertex 300 ...
Smoothing at vertex 400 ...

Even after about 7 hours it still hadn't broke out of the cycle. Any idea what might be going on? The commands I ran were Freesurfer's mris_convert to convert my files to gifti, and then film_gls as follows:

film_gls --rn=bsyn_01.lh.feat/surfstats --sa --ms=15 --epith=5 --in2=lh.white.gii --in=lh.bsyn_01.gii --pd=bsyn_01.lh.feat/design.mat --mode=surface

Is it possible that this behavior is due to the large number of vertices? I know you mentioned that it's best to carry out the stats at a lower resolution, but I haven't worked out a process for mapping the functional data to a lower res mesh yet.

My film_gls help message says the build is 506, if that's informative.

Thanks in advance!
Connor




On Thu, Mar 27, 2014 at 11:56 PM, Matt Glasser <[log in to unmask]> wrote:
Yes you can do this (with GIFTI files).  Here is an example call to film_gls:

  ${HCPPIPEDIR_Bin}/film_gls --rn=${FEATDir}/"$Hemisphere"_SurfaceStats --sa --ms=15 --epith=5 --in2="$DownSampleFolder"/"$Subject"."$Hemisphere".midthickness."$LowResMesh"k_fs_LR.surf.gii --in=${FEATDir}/${LevelOnefMRIName}${TemporalFilterString}${SmoothingString}.atlasroi_dil."$Hemisphere"."$LowResMesh"k_fs_LR.func.gii --pd="$DesignMatrix" --mode=surface

To use contrast_mgr or flameo you need to convert the GIFTI data to NIFTI first (and then back again afterward), which you can do with Connectome Workbench’s wb_command -metric-convert.  

It is best to do this analysis on a low resolution mesh (in the HCP we use 32k, which is 2mm average spacing on the surface).

Peace,

Matt.

From: Connor Lane <[log in to unmask]>
Reply-To: FSL - FMRIB's Software Library <[log in to unmask]>
Date: Thursday, March 27, 2014 at 10:48 PM
To: <[log in to unmask]>
Subject: [FSL] film_gls surface mode

Dear FSL experts,

I noticed in the film_gls help message that there's now a surface mode, but I haven't been able to find much documentation on this new option. Is it possible to use the surface mode to analyze functional data in a Freesurfer format? I tried just converting the Freesurfer files to gifti and running the command, but it seemed to stall on the autocorrelation smoothing step.

I've been looking for a way to incorporate Freesurfer's smoothing and inter-subject registration methods into my lab's processing stream without having to fully convert to Freesurfer. If this surface mode can work, I think most of my troubles would be solved!

Thank you in advance for the help,
Connor