Sounds from the North

How to create a wireframe 3D model from the MNI brain template with Matlab

Posted on 16. Feb, 2010 by admin in How To

PROBLEM: How can we obtain a 3D wireframe model of the standard brain in Matlab?

Sometimes we need to project significant results (statistical maps, correlation maps, functional connectivity, etc…) in a three dimensional space. While there are good tools for doing this, when we build a prototype we want to access the 3D image at a very low level.

This simple method utilises the standard MNI 152 2mm T1 brain mask template to build the wireframe.

The steps go as follows:

1. Make sure you have FSL, in the folder /usr/local/fsl/data/standard/ you find the volume MNI152lin_T1_2mm_brain_mask.nii.gz
2. Gunzip the volume (from the terminal: gunzip filename.gz) and load it in Matlab. You need the Matlab NIFTI toolbox.

   img=load('brain.nii');

3. As you can see by inspecting img.img the loaded volume is made of 0 when we are outside the brain and 1 when we are inside the brain
4. To plot the wireframe we are using a function developed by Luigi Giaccari. See how it works here. As you see, we need to provide the point of our surface as a list of 3D coordinates (Nx3 matrix).
5. With the next snippet of code, we extract the surface out of the Nifti standard volume:

   img=load_nii('brain.nii');
   brain=double(img.img);
   prev=0;
   p=[];
   for x=1:91;
       for y=1:109
           for z=1:91;
               v=brain(x,y,z);
               if(prev~=v)
                   p=[p; x y z];
               end
               prev=v;
           end
       end
   end

6. the variable p now stores Nx3 points in space that identify our surface
7. to plot the surface as a 3D wireframe with Matlab you can then use:

   [t,tnorm]=MyRobustCrust(p);
   figure(1)
   hold on
   axis equal
   h=trisurf(t,p(:,1),p(:,2),p(:,3),'facealpha',0.4,...
   ...'facecolor',[1 1 1],'edgecolor',[0.8 0.8 0.8]);
   view(3);
   axis vis3d

8. Using all N values of p can result in a very dense surface. We can then downsample the p matrix and still obtain a decent plot (plot in the image above):
p=p(1:5:end,:);

Tags: 3D, brain, featured, matlab