you would change fsaverage5 to fsaverage4 (fewer vertices) or fsaverage6 (more vertices) On Thu, 16 May 2013, Ludovico Minati wrote:
Hi Thomas
many thanks. So, in the example you have sent me, how should I go about if I wished to vary the granularity? Are there different meshes to reshape or should I generate them somehow?
Ludovico
On 16 May 2013 14:47, Thomas Yeo ythomas@csail.mit.edu wrote:
Hi Ludovico,
The mri_vol2surf example I sent you directly resample to fsaverage5, which is around 4mm spacing, resulting in 18715 vertices for both hemispheres combined after excluding the medial wall in the surfaces.
I am also including the freesurfer mailing list in the email as suggested by Bruce.
--Thomas
On Thu, May 16, 2013 at 8:25 PM, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
Hi Ludovico
I wouldn't downsample before calling mri_surf2surf. We have tools for uniformly parcellating the cortex (mris_make_face_parcellation) that can be used for this purpose after sampling the timecourses to the cortex.
cheers Bruce On Thu, 16 May 2013, Ludovico Minati wrote:
Hi Bruce,
many thanks. Yes sorry I shall use the list from now on. But let me just ask you a clarification as we are already on these questions
For our connectivity analyses, at present we parcellated in 3D volume space the cortex into a pre-set number of roughly equal volume parcels, ranging in number between around 500 to 10,000. I would need to translate this approach to the surface space in Freesurfer, i.e. be able to smooth on the cortical surface and undersample it. What is the best approach that can be used?
Along this like: you wrote around 150k vertices/hemisphere. For our purposes that's an enormous number, much larger than the number of EPI voxels. How can we mesh downsample before calling mri_vol2surf?
Again thanks for your kind help and patience
Ludovico
On 16 May 2013 14:09, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
Hi Ludovico
can we move this discussion to the FreeSurfer list in the future so others can answer?
- Hard to say as sometimes one works better than the other, you'll just
have to try. The transform should be rigid (6DOF) since the scalings are read from the header, why would you want affine? I think Doug has worked on this and you can specify it in the bbregister command line, but I've never used it.
- About 150K vertices/hemisphere. They are distributed initially at 1mm
spacing (the surface of the voxels), and cannot be tuned. There are some tools around for mesh downsampling if you want, and this is even easier if you go to fsaverage space (with mri_surf2surf) where we have representations based on recursive icosahedral supersamplings.
- This is due to a bug in the initial nifti format that didn't allow
any one dimension greater than 64K. If you output in .mgz it should be nvertices x 1 x 1 x ntimepoints.
- Search for "coordinate systems" on our wiki. This is documented
extensively.
there are also matlab i/o functions for reading/writing all our formats (e.g. MRIread.m)
cheers Bruce
On Thu, 16 May 2013, Ludovico Minati wrote:
Hi Thomas,
thanks. A few quick questions, which hopefully will take little of your time but save me a lot of hassle:
- bbregister. Do you advise using FLIRT or SPM for initialization? Is
this a rigid-body transformation, or an affine one? Any way I can control the DOF? If I use SPM shall I feed the mean realigned volume here, as I expect?
- recon-all. What is the typical number of vertices for the cortex?
Any way I can control it, i.e. sample more or less finely? This would be particularly relevant for our task.
- mri_vol2surf. Perhaps naively I was expecting an output file simply
consisting of a list of values, i.e. one per vertex (assuming a single input volume that is). But the documentation says: "The output will be a data set with Nv/R colums, 1 row, R slices, and Nf frames..." What is the reshaping factor R for? How do I deal with this, i.e. how do I simply get a value per vertex? Or am I missing something here? Is mri_vol2surf averaging over neighbouring vertices?
- Where do I read the Euclinean co-ordinates of the vertices from?
Cheers
Ludovico
On 16 May 2013 04:30, Thomas Yeo ythomas@csail.mit.edu wrote: > > > Hi Ludovico, > > Just remember that you need to run "recon-all" on your T1 anatomical > first. > > For bbregister, the command goes something like this: bbregister > --bold --s $anat_s --init-fsl --mov $input --reg $reg. I find the > bbregister cost function value a good indication of whether the > registration has good well. Below 0.5 is good. Below 0.6 is probably > acceptable. > > For mri_vol2surf, the command goes something like this: mri_vol2surf > --mov $f --reg $reg --hemi $hemi --projfrac 0.5 --trgsubject > fsaverage5 --o $output --reshape --interp trilinear > > Cheers, > Thomas > > On Wed, May 15, 2013 at 8:57 PM, Bruce Fischl > fischl@nmr.mgh.harvard.edu wrote: >> >> >> Hi Ludovico >> >> you can get vertex-level timecourse by using bbregister to register >> the >> EPI >> data to the surface, then mri_vol2surf to sample onto the surface. >> >> cheers >> Bruce >> >> >> >> On Wed, 15 May 2013, Ludovico Minati wrote: >> >>> Hello Thomas, >>> >>> yes sure. Our area of interest is early dementia diagnosis, and we >>> are >>> exploring the predictive value of graph topological indices. As you >>> know >>> there are major volume changes in dementia, and from the literature I >>> have >>> convinced myself we may be better off resolving this confound through >>> working on the surface, rather than playing around with ad-hoc >>> atlases >>> for >>> volumetric normalization. >>> >>> I have got plenty of signal processing code to perform the analyses >>> from >>> BOLD timeseries; what I am looking for now is a way to apply >>> Freesurfer >>> normalization to obtain vertex-level time-courses. >>> >>> Hence, I am so far only interested in solving the following problem >>> with >>> Freesurfer: we have as input an individual T1 scan and a series of >>> EPI >>> volumes, we want to preprocess/coregister/normalize them with >>> Freesurfer and >>> extract the BOLD timecourses for each vertex. No fMRI analysis in >>> Freesurfer, no group analysis. >>> >>> Hopefully it is not too complex? I am dwarfed by the range of >>> commands >>> available! >>> >>> Thank you very much in advance >>> >>> Ludovico >>> >>> -----Original Message----- >>> From: Thomas Yeo ythomas@csail.mit.edu >>> Sender: yeoyeo02@gmail.com >>> Date: Wed, 15 May 2013 10:07:00 >>> To: Bruce Fischlfischl@nmr.mgh.harvard.edu >>> Cc: lminati@ieee.orglminati@ieee.org; Dennis >>> ChanD.Chan@bsms.ac.uk >>> Subject: Re: FMRI surface transformation for vertex-wise connectivity >>> analysis >>> >>> Hi Ludovico, >>> >>> Can you give me the context of what you need? >>> >>> Cheers, >>> Thomas >>> >>> On Wed, May 15, 2013 at 9:52 AM, Bruce Fischl >>> fischl@nmr.mgh.harvard.edu wrote: >>>> >>>> >>>> >>>> Hi Ludovico >>>> Thomas Yeo (ccd) has a set of scripts for this purpose >>>> Cheers >>>> Bruce >>>> >>>> >>>> >>>> On May 14, 2013, at 8:50 PM, "Ludovico Minati" lminati@ieee.org >>>> wrote: >>>> >>>>> Ludovico >>>> >>>> >>>> >>>> >>>> >>>> The information in this e-mail is intended only for the person to >>>> whom >>>> it >>>> is >>>> addressed. If you believe this e-mail was sent to you in error and >>>> the >>>> e-mail >>>> contains patient information, please contact the Partners Compliance >>>> HelpLine at >>>> http://www.partners.org/complianceline . If the e-mail was sent to >>>> you >>>> in >>>> error >>>> but does not contain patient information, please contact the sender >>>> and >>>> properly >>>> dispose of the e-mail. >>>> >>> >>> >>> >>