Re: [HCP-Users] Two questions about surfaces & template spaces | MSMall & FS

[Timothy Coalson]

On Tue, Jul 25, 2017 at 2:18 AM, Claude Bajada 
wrote:

> Thanks Tim et al,
>
> I have some follow-up questions:
>
>- I am performing tractography by defining an individual coordinate
>(in mm not voxels). If I use an XYZ that corresponds to one of the vertices
>on the 'white.32k' surface, that should correspond to the same point in the
>DWI data. Correct?
>- Would the rotation affect this? ("just rotated and moved to the
>usual orientation"). I was under the impression that the DWI and T1 were
>also rotated into the usual orientation.
>
> You state:
>
> "The surfaces in T1w are generated first, as we run freesurfer on the
> structural scans to generate them."
> I assume that these are the ones living in T1w/Native/
>
> "We then apply the FNIRT warpfield to them to get the MNINonLinear
> versions."
> These live in MNINonLinear/Native
>
> Then you downsample them to 32k... in MNINonLinear/fsaverage_LR32k
>
> The minimal processing pipeline publication (Glasser et al 2013) then
> states:
> "Finally, this 32k_fs_LR mesh [I'm assuming the one mentioned in the point
> above] is transformed from MNI space back to the native volume space"
>
> This may have been a wording simplification in the paper, there is no need
to use the volume warpfield.  I think the MSMAll pipeline in fact resamples
the T1w space native mesh midthickness surface to obtain the T1w space
standard mesh midthickness surface, here:

https://github.com/Washington-University/Pipelines/blob/master/MSMAll/scripts/MSMAll.sh#L1113

I'm not sure where the pipelines resample the other surfaces, but I would
expect that we do not take these coordinates through a volume warpfield
twice - using the coordinates from the existing native mesh surfaces
already in that volume space is much simpler.

> The question is:
>
>- Is this now a correct interpretation?
>
> Regards,
>
> Claude
>
> On 24.07.2017 22:38, Timothy Coalson wrote:
>
> On Mon, Jul 24, 2017 at 8:40 AM, Claude Bajada 
> wrote:
>
>> Dear all,
>>
>> I have a two questions about the surfaces that live in the HCP data
>> folders:
>>
>>1. MNINonLinear/fsaverage/
>>2. T1w/fsaverageLR32k/
>>
>> I assume you mean the fsaverage_LR32k folder within each of those folders.
>
>> If I have understood the minimal pipeline correctly:
>>
>>1. all the surfaces in MNINonLinear/fsaverage32k/ are the
>>individual's surface that was warped into a template space (either FS or
>>MSMAll).
>>2. the surfaces in T1w/fsaverageLR32k/ are the surfaces mentioned in
>>the above point that are warped back into the individual's native space
>>
>> Not exactly.  Volume spaces are a completely orthogonal question to
> surface registration.  Except for some special files, everything in the
> MNINonLinear folder aligns to the MNI template via FNIRT registration.
> Everything in T1w similarly aligns to the subject's AC-PC rigid aligned
> space (original size and shape of the subject's structural scans, just
> rotated and moved to the usual orientation).
>
> MSMAll, freesurfer, MSMSulc, or any other surface registration does not
> cause any anatomical deformations of the individual's brain.  The standard
> mesh anatomical surfaces you get after surface registration and resampling
> align with the same volumes as the original surfaces.  There is no
> pre-defined anatomical coordinate template in our surface atlases, so we
> don't need to work backward from any template coordinate system.  Instead,
> we have standard spheres that define the standard topology (triangle
> relationships) - once we have the subject's sphere aligned to the standard
> sphere via template data (sulc for MSMSulc or FS; resting state networks,
> myelin, thickness, etc for MSMAll), we can combine the standard sphere's
> topology with the subject's anatomical coordinate data (by resampling the
> XYZ values), generating the subject's standard mesh surfaces.
>
> The surfaces in T1w are generated first, as we run freesurfer on the
> structural scans to generate them.  We then apply the FNIRT warpfield to
> them to get the MNINonLinear versions.
>
> As a side note, group average surfaces ARE significantly deformed from the
> individuals' brain shapes.  Most notably, they contain much less folding,
> and therefore surface area, as all the incompatible or non-corresponding
> folding patterns (when you align based on function, folding becomes less
> aligned, as function isn't always fixed with respect to folds) effectively
> smooth each other out when averaged.  This is why we have correction
> methods for this when doing spatial operations on group average surfaces.
>
>> The two questions are:
>>
>>1. Given the above, am I right in assuming that any data that
>>correspond to a vertex of a surface in T1w/fsaverageLR32k/ also 
>> corresponds
>>to the surface in MNINonLinear/fsaverageLR32k/?
>>
>>  Yes.
>
> > 2. Do you recommend that we the MSMAll surfaces 

Re: [HCP-Users] Two questions about surfaces & template spaces | MSMall & FS

[Glasser, Matthew]

For tractography of HCP data you would use the surfaces in 
${StudyFolder}/${Subject}/T1w/fsaverage_LR32k/ and the diffusion data in 
${StudyFolder}/${Subject}/T1w/Diffusion, which have the same ACPC rigid 
registration to MNI space.  If you want your tractography data to be aligned 
with MSMAll across subjects, use the surfaces with _MSMAll in the names.  If 
you want to use MSMSulc, use the surfaces without _MSMAll in the names.

Peace,

Matt.

From: 
<hcp-users-boun...@humanconnectome.org<mailto:hcp-users-boun...@humanconnectome.org>>
 on behalf of Claude Bajada 
<c.baj...@fz-juelich.de<mailto:c.baj...@fz-juelich.de>>
Date: Tuesday, July 25, 2017 at 2:18 AM
To: "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" 
<hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>>
Subject: Re: [HCP-Users] Two questions about surfaces & template spaces | 
MSMall & FS


Thanks Tim et al,

I have some follow-up questions:

  *   I am performing tractography by defining an individual coordinate (in mm 
not voxels). If I use an XYZ that corresponds to one of the vertices on the 
'white.32k' surface, that should correspond to the same point in the DWI data. 
Correct?
  *   Would the rotation affect this? ("just rotated and moved to the usual 
orientation"). I was under the impression that the DWI and T1 were also rotated 
into the usual orientation.

You state:

"The surfaces in T1w are generated first, as we run freesurfer on the 
structural scans to generate them."
I assume that these are the ones living in T1w/Native/

"We then apply the FNIRT warpfield to them to get the MNINonLinear versions."
These live in MNINonLinear/Native

Then you downsample them to 32k... in MNINonLinear/fsaverage_LR32k

The minimal processing pipeline publication (Glasser et al 2013) then states:
"Finally, this 32k_fs_LR mesh [I'm assuming the one mentioned in the point 
above] is transformed from MNI space back to the native volume space"

The question is:

  *   Is this now a correct interpretation?

Regards,

Claude

On 24.07.2017 22:38, Timothy Coalson wrote:
On Mon, Jul 24, 2017 at 8:40 AM, Claude Bajada 
<c.baj...@fz-juelich.de<mailto:c.baj...@fz-juelich.de>> wrote:

Dear all,

I have a two questions about the surfaces that live in the HCP data folders:

  1.  MNINonLinear/fsaverage/
  2.  T1w/fsaverageLR32k/

I assume you mean the fsaverage_LR32k folder within each of those folders.

If I have understood the minimal pipeline correctly:

  1.  all the surfaces in MNINonLinear/fsaverage32k/ are the individual's 
surface that was warped into a template space (either FS or MSMAll).
  2.  the surfaces in T1w/fsaverageLR32k/ are the surfaces mentioned in the 
above point that are warped back into the individual's native space

Not exactly.  Volume spaces are a completely orthogonal question to surface 
registration.  Except for some special files, everything in the MNINonLinear 
folder aligns to the MNI template via FNIRT registration.  Everything in T1w 
similarly aligns to the subject's AC-PC rigid aligned space (original size and 
shape of the subject's structural scans, just rotated and moved to the usual 
orientation).

MSMAll, freesurfer, MSMSulc, or any other surface registration does not cause 
any anatomical deformations of the individual's brain.  The standard mesh 
anatomical surfaces you get after surface registration and resampling align 
with the same volumes as the original surfaces.  There is no pre-defined 
anatomical coordinate template in our surface atlases, so we don't need to work 
backward from any template coordinate system.  Instead, we have standard 
spheres that define the standard topology (triangle relationships) - once we 
have the subject's sphere aligned to the standard sphere via template data 
(sulc for MSMSulc or FS; resting state networks, myelin, thickness, etc for 
MSMAll), we can combine the standard sphere's topology with the subject's 
anatomical coordinate data (by resampling the XYZ values), generating the 
subject's standard mesh surfaces.

The surfaces in T1w are generated first, as we run freesurfer on the structural 
scans to generate them.  We then apply the FNIRT warpfield to them to get the 
MNINonLinear versions.

As a side note, group average surfaces ARE significantly deformed from the 
individuals' brain shapes.  Most notably, they contain much less folding, and 
therefore surface area, as all the incompatible or non-corresponding folding 
patterns (when you align based on function, folding becomes less aligned, as 
function isn't always fixed with respect to folds) effectively smooth each 
other out when averaged.  This is why we have correction methods for this when 
doing spatial operations on group average surfaces.

The two questions are:

  1.  Given the above, am I right in assuming that any data that correspond to 
a vertex of a surf

Re: [HCP-Users] Two questions about surfaces & template spaces | MSMall & FS

[Claude Bajada]

Thanks Tim et al,

I have some follow-up questions:

 * I am performing tractography by defining an individual coordinate
   (in mm not voxels). If I use an XYZ that corresponds to one of the
   vertices on the 'white.32k' surface, that should correspond to the
   same point in the DWI data. Correct?
 * Would the rotation affect this? ("just rotated and moved to the
   usual orientation"). I was under the impression that the DWI and T1
   were also rotated into the usual orientation.

You state:

   "The surfaces in T1w are generated first, as we run freesurfer on
   the structural scans to generate them."
   I assume that these are the ones living in T1w/Native/

   "We then apply the FNIRT warpfield to them to get the MNINonLinear
   versions."
   These live in MNINonLinear/Native

   Then you downsample them to 32k... in MNINonLinear/fsaverage_LR32k

   The minimal processing pipeline publication (Glasser et al 2013)
   then states:
   "Finally, this 32k_fs_LR mesh [I'm assuming the one mentioned in the
   point above] is transformed from MNI space back to the native volume
   space"

The question is:

 * Is this now a correct interpretation?

Regards,

Claude


On 24.07.2017 22:38, Timothy Coalson wrote:
On Mon, Jul 24, 2017 at 8:40 AM, Claude Bajada > wrote:


Dear all,

I have a two questions about the surfaces that live in the HCP
data folders:

 1. MNINonLinear/fsaverage/
 2. T1w/fsaverageLR32k/

I assume you mean the fsaverage_LR32k folder within each of those folders.

If I have understood the minimal pipeline correctly:

 1. all the surfaces in MNINonLinear/fsaverage32k/ are the
individual's surface that was warped into a template space
(either FS or MSMAll).
 2. the surfaces in T1w/fsaverageLR32k/ are the surfaces mentioned
in the above point that are warped back into the individual's
native space

Not exactly.  Volume spaces are a completely orthogonal question to 
surface registration.  Except for some special files, everything in 
the MNINonLinear folder aligns to the MNI template via FNIRT 
registration.  Everything in T1w similarly aligns to the subject's 
AC-PC rigid aligned space (original size and shape of the subject's 
structural scans, just rotated and moved to the usual orientation).


MSMAll, freesurfer, MSMSulc, or any other surface registration does 
not cause any anatomical deformations of the individual's brain.  The 
standard mesh anatomical surfaces you get after surface registration 
and resampling align with the same volumes as the original surfaces. 
There is no pre-defined anatomical coordinate template in our surface 
atlases, so we don't need to work backward from any template 
coordinate system.  Instead, we have standard spheres that define the 
standard topology (triangle relationships) - once we have the 
subject's sphere aligned to the standard sphere via template data 
(sulc for MSMSulc or FS; resting state networks, myelin, thickness, 
etc for MSMAll), we can combine the standard sphere's topology with 
the subject's anatomical coordinate data (by resampling the XYZ 
values), generating the subject's standard mesh surfaces.


The surfaces in T1w are generated first, as we run freesurfer on the 
structural scans to generate them.  We then apply the FNIRT warpfield 
to them to get the MNINonLinear versions.


As a side note, group average surfaces ARE significantly deformed from 
the individuals' brain shapes.  Most notably, they contain much less 
folding, and therefore surface area, as all the incompatible or 
non-corresponding folding patterns (when you align based on function, 
folding becomes less aligned, as function isn't always fixed with 
respect to folds) effectively smooth each other out when averaged.  
This is why we have correction methods for this when doing spatial 
operations on group average surfaces.


The two questions are:

 1. Given the above, am I right in assuming that any data that
correspond to a vertex of a surface in T1w/fsaverageLR32k/
also corresponds to the surface in MNINonLinear/fsaverageLR32k/?

 Yes.

> 2. Do you recommend that we the MSMAll surfaces over the FS ones?

Yes, they have considerably better cross subject functional 
correspondence than freesurfer.  That is, if a given vertex index is 
in area MT in one subject, MSMAll will have a considerably higher 
percentage of other subjects that also have that vertex in MT.


Regards,

Claude








Forschungszentrum Juelich GmbH
52425 Juelich
Sitz der Gesellschaft: Juelich
Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher

Re: [HCP-Users] Two questions about surfaces & template spaces | MSMall & FS

[Glasser, Matthew]

It is also worth noting that the HCP Pipelines output MSMSulc and MSMAll 
surfaces only (FreeSurfer is worse at aligning function than MSMSulc with more 
distortion).

Matt.

From: 
<hcp-users-boun...@humanconnectome.org<mailto:hcp-users-boun...@humanconnectome.org>>
 on behalf of Timothy Coalson <tsc...@mst.edu<mailto:tsc...@mst.edu>>
Date: Monday, July 24, 2017 at 3:38 PM
To: Claude Bajada <c.baj...@fz-juelich.de<mailto:c.baj...@fz-juelich.de>>
Cc: "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" 
<hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>>
Subject: Re: [HCP-Users] Two questions about surfaces & template spaces | 
MSMall & FS

On Mon, Jul 24, 2017 at 8:40 AM, Claude Bajada 
<c.baj...@fz-juelich.de<mailto:c.baj...@fz-juelich.de>> wrote:

Dear all,

I have a two questions about the surfaces that live in the HCP data folders:

  1.  MNINonLinear/fsaverage/
  2.  T1w/fsaverageLR32k/

I assume you mean the fsaverage_LR32k folder within each of those folders.

If I have understood the minimal pipeline correctly:

  1.  all the surfaces in MNINonLinear/fsaverage32k/ are the individual's 
surface that was warped into a template space (either FS or MSMAll).
  2.  the surfaces in T1w/fsaverageLR32k/ are the surfaces mentioned in the 
above point that are warped back into the individual's native space

Not exactly.  Volume spaces are a completely orthogonal question to surface 
registration.  Except for some special files, everything in the MNINonLinear 
folder aligns to the MNI template via FNIRT registration.  Everything in T1w 
similarly aligns to the subject's AC-PC rigid aligned space (original size and 
shape of the subject's structural scans, just rotated and moved to the usual 
orientation).

MSMAll, freesurfer, MSMSulc, or any other surface registration does not cause 
any anatomical deformations of the individual's brain.  The standard mesh 
anatomical surfaces you get after surface registration and resampling align 
with the same volumes as the original surfaces.  There is no pre-defined 
anatomical coordinate template in our surface atlases, so we don't need to work 
backward from any template coordinate system.  Instead, we have standard 
spheres that define the standard topology (triangle relationships) - once we 
have the subject's sphere aligned to the standard sphere via template data 
(sulc for MSMSulc or FS; resting state networks, myelin, thickness, etc for 
MSMAll), we can combine the standard sphere's topology with the subject's 
anatomical coordinate data (by resampling the XYZ values), generating the 
subject's standard mesh surfaces.

The surfaces in T1w are generated first, as we run freesurfer on the structural 
scans to generate them.  We then apply the FNIRT warpfield to them to get the 
MNINonLinear versions.

As a side note, group average surfaces ARE significantly deformed from the 
individuals' brain shapes.  Most notably, they contain much less folding, and 
therefore surface area, as all the incompatible or non-corresponding folding 
patterns (when you align based on function, folding becomes less aligned, as 
function isn't always fixed with respect to folds) effectively smooth each 
other out when averaged.  This is why we have correction methods for this when 
doing spatial operations on group average surfaces.

The two questions are:

  1.  Given the above, am I right in assuming that any data that correspond to 
a vertex of a surface in T1w/fsaverageLR32k/ also corresponds to the surface in 
MNINonLinear/fsaverageLR32k/?

 Yes.

> 2. Do you recommend that we the MSMAll surfaces over the FS ones?

Yes, they have considerably better cross subject functional correspondence than 
freesurfer.  That is, if a given vertex index is in area MT in one subject, 
MSMAll will have a considerably higher percentage of other subjects that also 
have that vertex in MT.

Regards,

Claude





Forschungszentrum Juelich GmbH
52425 Juelich
Sitz der Gesellschaft: Juelich
Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher
Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender),
Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt,
Prof. Dr. Sebastian M. Schmidt




___
HCP-Users mailing list
HCP-Users@humanconnectome.org<mailto:HCP-Users@humanconnectome.org>
http://lists.humanconnectome.org/mailman/listinfo/hcp-users


___

Re: [HCP-Users] Two questions about surfaces & template spaces | MSMall & FS

[Timothy Coalson]

On Mon, Jul 24, 2017 at 8:40 AM, Claude Bajada 
wrote:

> Dear all,
>
> I have a two questions about the surfaces that live in the HCP data
> folders:
>
>1. MNINonLinear/fsaverage/
>2. T1w/fsaverageLR32k/
>
> I assume you mean the fsaverage_LR32k folder within each of those folders.

> If I have understood the minimal pipeline correctly:
>
>1. all the surfaces in MNINonLinear/fsaverage32k/ are the individual's
>surface that was warped into a template space (either FS or MSMAll).
>2. the surfaces in T1w/fsaverageLR32k/ are the surfaces mentioned in
>the above point that are warped back into the individual's native space
>
> Not exactly.  Volume spaces are a completely orthogonal question to
surface registration.  Except for some special files, everything in the
MNINonLinear folder aligns to the MNI template via FNIRT registration.
Everything in T1w similarly aligns to the subject's AC-PC rigid aligned
space (original size and shape of the subject's structural scans, just
rotated and moved to the usual orientation).

MSMAll, freesurfer, MSMSulc, or any other surface registration does not
cause any anatomical deformations of the individual's brain.  The standard
mesh anatomical surfaces you get after surface registration and resampling
align with the same volumes as the original surfaces.  There is no
pre-defined anatomical coordinate template in our surface atlases, so we
don't need to work backward from any template coordinate system.  Instead,
we have standard spheres that define the standard topology (triangle
relationships) - once we have the subject's sphere aligned to the standard
sphere via template data (sulc for MSMSulc or FS; resting state networks,
myelin, thickness, etc for MSMAll), we can combine the standard sphere's
topology with the subject's anatomical coordinate data (by resampling the
XYZ values), generating the subject's standard mesh surfaces.

The surfaces in T1w are generated first, as we run freesurfer on the
structural scans to generate them.  We then apply the FNIRT warpfield to
them to get the MNINonLinear versions.

As a side note, group average surfaces ARE significantly deformed from the
individuals' brain shapes.  Most notably, they contain much less folding,
and therefore surface area, as all the incompatible or non-corresponding
folding patterns (when you align based on function, folding becomes less
aligned, as function isn't always fixed with respect to folds) effectively
smooth each other out when averaged.  This is why we have correction
methods for this when doing spatial operations on group average surfaces.

> The two questions are:
>
>1. Given the above, am I right in assuming that any data that
>correspond to a vertex of a surface in T1w/fsaverageLR32k/ also corresponds
>to the surface in MNINonLinear/fsaverageLR32k/?
>
>  Yes.

> 2. Do you recommend that we the MSMAll surfaces over the FS ones?

Yes, they have considerably better cross subject functional correspondence
than freesurfer.  That is, if a given vertex index is in area MT in one
subject, MSMAll will have a considerably higher percentage of other
subjects that also have that vertex in MT.

> Regards,
>
> Claude
>
>
>
>
> 
> 
> 
> 
> Forschungszentrum Juelich GmbH
> 52425 Juelich
> Sitz der Gesellschaft: Juelich
> Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
> Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher
> Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender),
> Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt,
> Prof. Dr. Sebastian M. Schmidt
> 
> 
> 
> 
>
> ___
> HCP-Users mailing list
> HCP-Users@humanconnectome.org
> http://lists.humanconnectome.org/mailman/listinfo/hcp-users
>

___
HCP-Users mailing list
HCP-Users@humanconnectome.org
http://lists.humanconnectome.org/mailman/listinfo/hcp-users

[HCP-Users] Two questions about surfaces & template spaces | MSMall & FS

[Claude Bajada]

Dear all,

I have a two questions about the surfaces that live in the HCP data folders:

 1.  MNINonLinear/fsaverage/
 2.  T1w/fsaverageLR32k/

If I have understood the minimal pipeline correctly:

 1.  all the surfaces in MNINonLinear/fsaverage32k/ are the individual's 
surface that was warped into a template space (either FS or MSMAll).
 2.  the surfaces in T1w/fsaverageLR32k/ are the surfaces mentioned in the 
above point that are warped back into the individual's native space

The two questions are:

 1.  Given the above, am I right in assuming that any data that correspond to a 
vertex of a surface in T1w/fsaverageLR32k/ also corresponds to the surface in 
MNINonLinear/fsaverageLR32k/?
 2.  Do you recommend that we the MSMAll surfaces over the FS ones?

Regards,

Claude





Forschungszentrum Juelich GmbH
52425 Juelich
Sitz der Gesellschaft: Juelich
Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher
Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender),
Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt,
Prof. Dr. Sebastian M. Schmidt




___
HCP-Users mailing list
HCP-Users@humanconnectome.org
http://lists.humanconnectome.org/mailman/listinfo/hcp-users