Working with inflation for morphometry

[lscheef]

Dear listmembers,

I would like to compare the size of different cortical areas between two populations. The borders between these areas are defined by a couple of sulci.

1) I wonder, if it would a be a valid approach to use "Ana Inflate Cortical Surface" (wich -as I understand- does use the white matter mesh ) to address this question.

2) Additionally, I would like to ask if there is a method to directly access the cortical thickness.

3) Which module do I have to use in order to add the appropriate texture mapping (as curvature etc. ) on the inflated mesh to distinguish between gyri und sulci...

Many many thanks for your help in advance,


Lukas

[riviere]

1) Yes "Ana Inflate Cortical Surface" uses the white matter mesh. But measuring cortical areas on an inflated brain seems not a good idea to me because the inflation induces local deformations so areas are not preserved.

2) No we don't have yet a tool to calculate a cortical thickness map. There is a beginning of such a process but as far as I know it's not functional yet (?).

3) The curvature map is calculated during the "Ana Inflate Cortical Surface" process, you can keep the curvature texture and map it on the mesh (either the original white mesh or the inflated mesh). Just clicking the eye icon of BrainVisa should be OK to render it in Anatomist.

There is a gyral parcellation module in brainvisa (a part of Arnaud Cachia's PhD work). You can try it but it's still a bit experimental so you must be in "expert" user mode to access it: it is in "surface based analysis/white matter parcellation". It is based on the cortical sulci recognition results so you must first do the anatomical pipeline and the sulci recognition (a pass of manual correction of cortical identifications may be required). The parcellation result is a graph of the main gyri. Then you can go to the "anatomy/morphometry/morphometry statistics" to get some measures on these gyri (using the "gyrusmodel" model) - the area will be written in ASCII files.

Denis

[lscheef]

Hi Denis,

thanks a lot for the reply. May I ask you what the gyral parcellation tool does?
Additionally, I wonder if it is possible to draw regions of interests on inflated brains and map those back into the native (uninfalted) space (the best would be even not normalized). This would solve my problem, because by doing this I could define the borders of the areas of interest using inflation and calculate the volumes on the native or normalized T1...

Best wishes,

Lukas

[Olivier Coulon]

Dear Lukas,
concerning the cortical surface inflation, I don't think you can use it to compare cortical areas. The inflation process used by AnaInflateCorticalSurface has been developed for vizion of distances/areas, ualisation purpose only. It does not guarantee any sort of distance preservation. The method presented by Fischl and al. in Neuroimage (vol. 9, 195-207, 1999) would be more suitable since it does attempt to preserve relative distances. Nevertheless, to preserve areas you need to preserve both distances and angles, which is impossible. Therefore, I think you should do your measurements on the original mesh.
Concerning your second point, there is a module in "Surface based analysis -> Cortical thickness". I am not sure it is in the current Brainvisa public distribution but if not it will be in the next one...
Concerning your 3rd point, AnaInflateCorticalSurface generates a texture of the white matter surface curvature, you can look at it in Anatomist, by selecting both the texture and the corresponding inflated mesh and doing a "Fusion".
Good luck,

Olivier

[riviere]

The gyral parcellation tool automatically defines gyral regions delimited by the main sulci (there is a list of sulcus limits somewhere) and makes a "gyri graph" from them, which is quite similar to the sulci graphs (but for gyri).
For more information, read about Arnaud Cachia's work:

A. Cachia et al. A generic framework for parcellation of the cortical surface into gyri using geodesic Voronoï diagrams. Medical Image Analysis, 7(4):403--416, 2003

It is not possible to draw ROIs directly on meshes (we'd like to add this functionality one day, but we lack time and programmers). However there is a possible trick to do it using the regular "voxel" ROI module:
- load the MRI volume and the mesh in Anatomist (or any volume that contains the full mesh)
- put the volume (and possibly the mesh) in an Anatomist window
- use the ROI drawing module of Anatomist to draw on this volume, you can use the mesh in the same view(s) to guide your drawing task. To access the ROI module, type F1 to show the toolbox window, and create a new drawing session and a new region.
- when you are done drawing, export the region as a binary volume
- load this binary volume in Anatomist
- fusion the binary volume and the mesh (3D fusion)
- this fusion creates a textured object. You can then export this texture to a file (click on the fusion object then use the menu "object specific/file/export texture")
Then you have a texture file that you can map either on the inflated mesh or the original white mesh.

However I don't think we have a tool to directly measure a mesh area inside a region defined by a texture: I'm afraid you will have to write a program...

Denis

[Jean-Francois Mangin]

Denis

There is a gyral parcellation module in brainvisa (a part of Arnaud Cachia's PhD work). You can try it but it's still a bit experimental so you must be in "expert" user mode to access it: it is in "surface based analysis/white matter parcellation". It is based on the cortical sulci recognition results so you must first do the anatomical pipeline and the sulci recognition (a pass of manual correction of cortical identifications may be required). The parcellation result is a graph of the main gyri. Then you can go to the "anatomy/morphometry/morphometry statistics" to get some measures on these gyri (using the "gyrusmodel" model) - the area will be written in ASCII files.

Denis

Yes we still have to work on the user-friendly interface of this module
and on the possibility for the user to define his own gyri from pairs
of sulci (and one day from sets of sulci). I would strongly
advice to wait for this upgrade before playing with these
gyrus-based measurements. The sulcus-based measurements,
indeed, provide already a lot of very similar informations.
See for instance:

J.-F. Mangin, D. Rivière, A. Cachia, E. Duchesnay, Y. Cointepas, D. Papadopoulos-Orfanos, D. L. Collins, A. C. Evans, and J. Régis. Object-based morphometry of the cerebral cortex. IEEE Trans. Medical Imaging, 2004.
http://brainvisa.info/pdf/mangin-TMI04.pdf

[lscheef]

Thanks a lot for your answers. Actually, I intend to use the sulcal based maesurement as well... and I will look through the publications you have mentioned.
As far as I have understood - it is not a good idea to measure the areas on the inflated surfaces (actually the same seems to be valid for FreeSurfer...) . My motivation for using the infalted surfaces was that on those it is quite easy to deliniate certian cortical areas. If the direct measurement does not make sense, the deliniation of the borders on the inflated brains still does it.
However, I would like to come back on Denis reply. I am not quite sure if this provide the answer. Does your answer mean, that I can use the inflated mesh to guide the drawing on the MRI? That would be perfectly fine... but I guess only 'folded meshes can be combined with MRIs...


Lukas

[riviere]

Yes, it makes sense to delimitate the gyral regions on the inflated brain and to measure them on the un-inflated one.
The only problem is that, as I said, we have made no tool right now to measure the area of mesh inside a drawn region. It would be easy to do, but it's not done up to now...

What I explained to draw a region on a mesh is actually an ugly trick to do something that has not been implemented yet, not an official way of drawing ROIs on meshes.
The only drawing tool in Anatomist works on voxels, not on meshes. So I explained how to draw, in a volume, a region that will be afterwards mapped on the mesh. The volume needs not to contain actual data, it can be an empty volume with all values set to zero, with any voxel size, where the mesh fits in. We will only need it to provide a grid geometry to the region(s) we draw (number and size of voxels). And yes you can use the inflated brain to guide the ROI drawing in this volume (the next version of Anatomist will be more convenient for this). Then a binary volume can be made from this ROI, and mapped on the mesh as a texture. Here you get a texture that marks the region you have drawn on the mesh.
This texture can then be applied to the original (non-inflated) mesh.
The next step is to measure the area of mesh inside this region (determined by polygons whose texture have a non-nul value), and this is where something is missing in our tools...

Denis

[PeterK]

A quick note about gyral parcelation tool. I've tried it a number of times, but it only created an empty graph. It was looking for label 105 and said that it is not found. Seems like it is missing a label translation file or something

[PeterK]

A quick note about gyral parcelation tool. I've tried it a number of times, but it only created an empty graph. It was looking for label 105 and said that it is not found. Seems like it is missing a label translation file or something

[Jean-Francois Mangin]

I strongly advise everybody to be reasonably patient with the processes that are still in advanced user levels. It means they are still under test in the lab developping them. We do not completely hide them because this testing period can be performed across several betasites. Playing with them elsewhere, however, is bound to lead to some problems. Since we already have the head partly under water with the difficulties triggered by validated processes, we will usually not support such attempts.

Our current plan is a refinement of our main processes to improve their robustness for a wider number of MR sequences. This first period of diffusion, indeed, helped us to discover various situations we do not deal with correctly. Then we might find the time to push more processes towards lower user levels.

Jeff

[PeterK]

A related question.
Say I want to depict certain information (cortical thickness) on the inflated map.

I do have the thickness value for each node in the "un-inflated" state - how would I go about depicting them on "inflated" state. Do you save the deformation information anywhere I can read from?

[Jean-Francois Mangin]

A related question.
Say I want to depict certain information (cortical thickness) on the inflated map.

I do have the thickness value for each node in the "un-inflated" state - how would I go about depicting them on "inflated" state. Do you save the deformation information anywhere I can read from?

If you have such information in the texture format (the *.tex files), you can fusion it (from Anatomist) with the initial OR the inflated mesh. It works the same because the nodes do not change.

[PeterK]

A related question.
Say I want to depict certain information (cortical thickness) on the inflated map.

I do have the thickness value for each node in the "un-inflated" state - how would I go about depicting them on "inflated" state. Do you save the deformation information anywhere I can read from?

If you have such information in the texture format (the *.tex files), you can fusion it (from Anatomist) with the initial OR the inflated mesh. It works the same because the nodes do not change.

Wow, that makes it all simple! Thanks man.