This example shows how to use the MESHER using a surface model .stl from a CAD application. This method was used to mesh the tanks in this paper, see Head Tanks for 3D printable versions. This requires MATLAB and iso2mesh to be installed. The script example_neonate.m walks through the process of converting to inr and calling the mesher all through MATLAB.
The input surface data will be converted into a volume mask using functions from iso2mesh. Each .stl file must contain only a single body, and each layer is voxelised on the same grid, and assigned a different binary value. The coordinate system will change, so the electrode positions are also transformed and stored in a new text file. The resolution of the output is specified by the pixel_scale option, which is then carried through to the parameters file.
elec_pos=dlmread('NNelecposorig.txt');
% run the conversion to INR. Plotting the figures to check the alignment of
% the electrodes and the quality of the output
[ full_mask,elec_pos_new_sc ] = stl2inr({'NNscalp.stl','NNskull_lowpoly.stl'},pixel_scale,elec_pos );
A slice through each of the cenre of the binary masks is plotted, which shows the alignment and domain values are correct.
As the INR file has a new coordinate system, it is sensible to check the newly created positions are correctly aligned on the new mesh
The initial parameter file was chosen to provide ~100k elements for demonstration and to check alignment and refinement settings, prior to running the final high resolution parameters.
The MESHER can be called using the command below, or using the MATLAB wrapper as shown in example_neonate.m.
../../bin/mesher -i NNscalp.inr -e NNscalp_elecINRpos.txt -p NNscalp_param.txt -d output/ -o NNexampleUsing paraview we can view the .vtu file we can check the output mesh as below. The regions of higher density around the electrodes are visible on the surface. Here we have changed the representation to surface with edges and colouring to subdomain index. Taking a slice/clip through the mesh we can see the boundaries of the skull (red) and scalp (blue) domains, as well as the depth of the electrode refinement.
Finally, we can check the output in MATLAB and ensure the electrode positions are maintained.
Mesh=loadmesh('output/NNexample');
figure
hold on;
trep = triangulation(Mesh.Tetra, Mesh.Nodes);
[Triangle_Boundary, Nodes_Boundary] = freeBoundary(trep);
h= trisurf(Triangle_Boundary, Nodes_Boundary(:,1), Nodes_Boundary(:,2), Nodes_Boundary(:,3));
set(h,'EdgeColor',[0.3,0.3,0.3],'FaceColor','w','FaceAlpha',1);
daspect([1,1,1]);
plot3(Mesh.elec_pos(:,1),Mesh.elec_pos(:,2),Mesh.elec_pos(:,3),'.','Markersize',40);
hold off
view(3)
title('Neonate Mesh Low Res')
Once the parameters are established, the final ~4mln element mesh can be created. First we increase pixel_scale to 4 and repeat the stl2inr process for a more accurate starting geometry. The cell_fine_size_mm, cell_coarse_size_mm and cell_size_electrodes_mm parameters were reduced to create a denser mesh. Finally, the optimisations are all turned on with sufficient time to let them finish. This process takes ~20 minutes.
../../bin/mesher -i NNscalp.inr -e NNscalp_elecINRpos.txt -p NNscalp_param_HR.txt -d output_HR/ -o NNexampleHighResThe final mesh is visibly smoother, and the thin skull is better represented due to the higher pixel_scale
