March 17, 2018 at 10:55 am #1170
on my search for a software which is capable of roiwise brain analysis, I found BrainSuite.
For the implementation of my project I need, in a addition to the in *.roiwise.stats.txt containing features of cortical thickness, area and volume, additional features, which mostly base on the surface of single regions. I am trying to calculate those features with the implementation of segmentation in BrainSuite (i.e. with Matlab).
My question is, if it is possible to calcualte the surface model of single anatomical regions.
Alternative to this, I took a look at the complete surface model of *.hemi.mid.cortex.svreg.dfs, sadly when I look at individual anatomical structures I do not get closed surfaces.
Alternative to this, I tried to calcualte the surface based of the in *.svreg.label.nii.gz containing volumes with isosurfaces. Sadly my resulting surface is very inaccurate.
Is there any better solution for this, or is BrainSuite not suitable for this kind of task?
March 21, 2018 at 4:17 pm #1174sychoiParticipant
I think what you’ve tried so far is great.
It seems for a closed surface, rendering a surface using the svreg labels are your best bet so far.
Can you specify what was inaccurate about your surfaces?
Note also that your svreg labels will be subdelineated by GM and WM labels for each individual cortical label. ROI ID labels for GM starts with “1” and for WM “2”
example: 1120: R. Superior frontal gyrus (WM) 2120: L. Superior frontal gyrus (WM)
I’m not sure how we can help you more with the given information. What are you trying to calculate exactly?
March 22, 2018 at 1:54 am #1175
Thank you for your fast answer. I tried to combine GM and WM now and it looks a bit better. Nevertheless the result seems to be a little edgy for me. Furthermore there are some small disjointed parts, as you can see in the image.
Maybe I can handle this by smoothing the surface.
My goal is to compute some shape features like total curvature and maybe some others. When I look at the surface model in *.hemi.mid.cortex.svreg.dfs it looks smooth so I hoped to get such an accurate model as well to increase the accuracy of my features.
March 22, 2018 at 11:25 am #1176Anand JoshiModerator
Hi Louise, For some reason, I can’t see the image you posted.
You can smooth the surface in matlab using svreg code.
First, save the surface as a dfs file using BrainSuite.
1. Generate the surface in BrainSuite
2. Goto Surface->Surface Display properties, select the surface that you want to save and press save button under the surface properties bar.
3. Download svreg source code from downloads section on BrainSuite website. Add path to the code in matlab.
You can also save it back to dfs file using writedfs function.
Additionally, you can compute vertexwise curvature using patchcurvature function available here:
March 23, 2018 at 3:10 am #1180
Hi Anand Joshi,
thank you for your fast answer. And sorry for the missing image. I try to repost it now.
I am not sure if I get you right. In my understanding you told me to extract total brain surface like *.hemi.mid.cortex.svreg.dfs. And then smooth it and extract features with matlab. The problem is, that I try to get closed surfaces and if I extract different anatomical structures from the *.hemi.mid.cortex.svreg.dfs file they are not closed.
The smoothing problem only appears if if I use *.svreg.label.nii.gz volumes to calculate self generated surfaces. And this is the way the image I uploaded is generated.
Is there any possibility to extract closed surfaces from single anatomical structures by using the command line?
- This reply was modified 2 years, 10 months ago by Louise.
April 3, 2018 at 5:37 pm #1191Anand JoshiModerator
The source code mentioned above has functionality to remove tiny patches of voxels that you can seen in the surfaces. It can be done in Matlab as follows, using the svreg source code from the brainsuite website.
We don’t have a command line utility for doing this, but the above mentioned matlab code should work.
- You must be logged in to reply to this topic.