Hi Marcos, the subfields of the left and right hippocampi are defined on different voxel spaces. If you want to combine them, you first need to reslice them to a common voxel space (for instance using -rl in mri_convert) Cheers, /Eugenio
On Fri, April 19, 2013 9:57 pm, Marcos Martins da Silva wrote:
TYVM, Juan Eugenio and Bruce. I am a neurologist and Professor in Federal University of Rio de Janeiro. I am confortable with Linux and scripts but, as a physician, eigenvectors will be really hard. I will try to get some help inside University. But if I correctly understood, it seems an elegant way to make tool that will work with several hippocampi with greater consistency than manual approach. Your step by step directions worked great, Juan. But I had to make a Left and a Right file. When I tried to concat right and left subfields in teh same file I got an error saying that right and left Ca1 dimensions did not match. Thank you again. Em Qui, 2013-04-18 Ã s 11:29 -0400, Juan Eugenio Iglesias escreveu:
I agree with Bruce. You can use MRIread.m to read in the data.
Regarding the other question: first you need to combine the posteriors of all the subfields into a single posterior for the whole hippocampus. You can do that like this:
mri_concat posterior_fimbria.mgz posterior_CA1.mgz ... --sum --o posteriorWholeHippocampus.mgz
Then you can threshold the posterior at probability 0.5 to obtain a binary hippocampal mask:
mri_binarize --i posteriorWholeHippocampus.mgz --min 127 --o binaryHippocampalMask.mgz
Finally, you can generate a surface by tesselating the mask:
mri_tesselate binaryHippocampalMask.mgz 1 surfaceFile
You can visualize the output with
freeview -v posteriorWholeHippocampus.mgz -f surfaceFile
Cheers,
/Eugenio
On Thu, 2013-04-18 at 11:22 -0400, Bruce Fischl wrote:
Hi Marcos
I would do it in matlab. Compute the spatial eigenvectors then divide
the
hippocampus up along the primary eigenvector
cheers Bruce On Thu, 18 Apr 2013, Marcos Martins da Silva wrote:
Thank you for your answer, Juan.
Yes, I want a surface of the whole hippocampus and I thought I had
just obtained that using
freeview and adding all hippo-subfield files. That resulted in the
screenshot I previously
attached. I just do not know how I can save the resultant surface.
After I checked "show as isosurface in 3D view" the low and high
threshold controls appeared.
The default threshold values on Freeview for low was 127.5 and high
was 255. When I played
with these controls, the size of the surfaces changed on 3D view.
That is why I made the
second question. I was trying to select the correct values for low
and high threshoulds so I
get an image consistent with the volume values we get after running kvlQuantifyPosteriorProbabilityImages. After setting the correct
values I would like to save
the work as a surface so I could load the whole surface instead of
loading each individual
subfield file. But I did not find a "save as surface" option. So I
used the screenshot tool to
"save" the work. This way I could at least show you what I got but I
thought it was not the
best way to do it.
Relating to the last question about generating a new segmentation
(anterior, body and
posterior), you said I could use simple geometric rules to get it.
Could you tell me where I
can find instructions about how to do it?
Sorry, I am afraid I am a novice here and my questions are the best
proof of this, so I would
like to thank you for all your patience.
Em Qua, 2013-04-17 Ã s 21:19 -0400, Juan Eugenio Iglesias escreveu: Dear Marcos,
1- How do I save the combined 3D view I got as a surface?
Do you want a surface of the whole hippocampus? In that case I
would
suggest adding up the probability volumes from all the subfields
and
thresholding the resulting volume.
2-I know the hippo-subifields .mgz files are probability maps
and if I
guess right kvlQuantifyPosteriorProbabilityImages read these
files to
count how many voxels you have in each subfield. I also guess it
count
only the voxels with reasonable high probability to be part of
each
subfield. If so, what is the minimum value between 1 and 255 the program use to decide if a particular voxel is valid?
The estimates of the volumes are based on soft probabilities, so
all
voxels with non-zero probability contribute towards the estimate.
For
instance, a voxel with value 255 contributes to the volume 255
times as
much as a voxel with a value of 1.
3- Is it possible to merge hippo-subfield .mgz files data and to generate a different segmentation like anterior (anterior
quarter),
body (the2 intermediate quarters) and posterior (posterior
quarter).
There are currently no scripts in FreeSurfer to generate such a partition. But you could always take a mask of the whole
hippocampus
(either from aseg.mgz of from the combination of all the
subfields, like
I mentioned in 1) and define some simple geometric rules to create
it.
Kind regards,
Eugenio
TY in advance _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
------------------------------------------------------------------------------------- Juan Eugenio Iglesias, PhD http://www.jeiglesias.com iglesias@nmr.mgh.harvard.edu Athinoula A. Martinos Center for Biomedical Imaging 149 Thirteenth Street, Suite 2301 Charlestown, Massachusetts 2129 U.S.A.