Tristan,

On Jan 17, 2011, at 6:56 AM, Tristan Chaplin wrote:

> Hi,
> 
> I'm working on constructing brain models from histological sections with 
> cortical areas demarcated.  Until now we've been writing our own progrms to 
> achieve this.  I'd like to start to use Caret for more of this work so 
> non-programmers can do it and our data is more compatible with other 
> datasets, but I've got a few questions.
> 
> Firstly, with regards to drawing a contour of the cortical mid-thickness, the 
> tutorial suggests that you just estimate this visually when drawing the 
> contours.  I was wondering, is there a standard procedure or set of tools for 
> calculating the mid-thickness given the pial and GM/WM boundaries?
Not for histological contours.  Unless the mapping between white and pial 
surfaces has already been determined (e.g., via Freesurfer automated 
segmentation) computing the midthickness is a very hard problem.

In practice, we generally achieve a reasonable outcome by drawing a contour 
along the estimated midpoint between the white matter and pial contours in a 
given section, modulated by the fact that histological layer 4 tends towards 
the pial side in gyri and near the white matter side near sulcal fundi.

Unless your histological sections are exceptionally well aligned (overcoming 
tissue distortions during histology), the inaccuracies associated with 
imperfect midthickness determination are likely to be small on average relative 
to the typical between-section misalignment.

Finally on this front, Erin Reid in my lab has been doing contour-based 
reconstructions on a couple of recent projects and has been helping to update 
the old tutorial.  If you have specific technical questions, you can contact 
her:
Erin Reid <[email protected]>

> 
> Secondly, with regards to marking cortical areas, normally we would represent 
> the area boundaries as points in space ("cells" in Caret I believe) and using 
> our own programs, project to the nearest mesh node.  Then, for a given 
> cortical area, we would use the Dijkstra algorithm to "cut out" the cortical 
> area by finding the shortest path between it's boundary points.  

Sounds nice.

> However it seems that in Caret the method is to view the projected cells on a 
> flat map, and draw a border around it to create a paint file.  
Yes, except that Caret now permits borders to be drawn on inflated or other 
closed surfaces, thereby obviating the problems along the cuts in flat maps.

> Is there a more precise way, perhaps similar to the one I described, of doing 
> it in Caret?

Not at present.  It would be nice to have this as an automated 
process/algorithm, but given other priorities it is not something we are likely 
to undertake in the near future.  If you have C++ code that can do it, let us 
know if you're willing to share it.

David VE

> 
> Thanks,
> Tristan
> 
> _______________________________________________
> caret-users mailing list
> [email protected]
> http://brainvis.wustl.edu/mailman/listinfo/caret-users

_______________________________________________
caret-users mailing list
[email protected]
http://brainvis.wustl.edu/mailman/listinfo/caret-users

Reply via email to