transforming non-uniform sized 3D mask to mesh or tet
Changran Geng
Qianqian Fang
--
You received this message because you are subscribed to the Google Groups "iso2mesh-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to iso2mesh-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/iso2mesh-users/f9b69390-75b5-43d6-82a9-03c6f898bd23n%40googlegroups.com.
Changran Geng
Qianqian Fang
hi Changran,
I ran your script, your v2s output contains very large number
nodes (80k for each layer, 160k nodes total). Because 80k node is
the default maximum node number for v2s, that means both surfaces
had encountered issues and could not finish surface extraction.
This is a combination of two things: 1) your opt.distbound
setting is too small, and 2) a known issue with CGAL where it gets
trapped at "non-manifold" nodes and hits an infinite loop for
refinement, see
https://github.com/fangq/iso2mesh/issues/5
I found that if you relax your distbound setting, v2s can complete with a much less node number. If I comment out your line opt(i).distbound=..., v2s out a surface with only 9136 nodes with good/comparable accuracy.
Aside from removing the distbound setting, another change I made is to insert a surface smoothing step before s2m, you can call
no=sms(no,fc(:,1:3),20);
to apply 20 iterations of smoothing, this makes both surfaces
quite nice.
in order to correctly label the two regions, I also added
seeds=surfseeds(no,fc);
[node,elem]=s2m(no,fc,1,100, 'tetgen', seeds);
to extract a seed point for each closed compartment, this can be
passed to s2m to correctly label each region
my updated script can be found below, along with the output mesh
screenshots. by the way, I felt your data aspect ratio is not
correct.
Qianqian
load('cleanimgfull.mat');
clear opt;
% set method for vol2mesh to 'simplify' to use
these option
for i=1:2
opt(i).side='lower'; %
opt(i).radbound=4; % set surface triangle
maximum size
%opt(i).distbound=0.2; % set
max distance that deviates from the level-set
opt(i).A=diag([0.9,0.7,3]);
opt(i).B=[0 0 0];
end
[no,fc]=v2s(cleanimgfull,[1,2]-0.5, opt);
no=sms(no,fc(:,1:3),20);
seeds=surfseeds(no,fc);
figure
plotmesh(no(:,[2 1 3]),fc,'facealpha',0.7);
[node,elem]=s2m(no,fc,1, 100, 'tetgen', seeds);
%[node,elem]=s2m(no,fc,1, 100, 'tetgen1.5'); % you
can also auto-label with 'tetgen1.5', but you don't know which
is which
figure
plotmesh(node,elem, 'y>160')
Dear Dr Fang,
Thank you very much for your kind reply. Yes, I also found some thread discussing this topic in the forum. It can be well solved.But I have another issue when using v2s and s2m, I attached the test code and mask that I would like to use. It actually have the skin and the brain, similar to the example you gave. It gives some error messages about the mesh intersection etc. But I couldnot found those errors.Do you have any suggestions on this kind of error? Is there any efficient tools you recommend in iso2mesh to solve this problem?
Thank you very much!Best regards,Changran
On Thu, Mar 16, 2023 at 9:48 PM Qianqian Fang <q.f...@neu.edu> wrote:
if the domain is simple, you can still call v2s() to extract the surface, and scale the x/y/z columns in the node output by the respective voxel size. This essentially stretches the surface mesh and creates a mesh of anisotropic density, however, when you call s2m() to produce tetrahedral meshes, as long as opt.radbound and opt.distbound is sufficiently small, the anisotropic triangles will largely remeshed without losing accuracy. below you can see an example of surface scaling and then s2m output.
alternatively, you can remesh the surface (remeshsurf) before passing it s2m.
if your domain is complex, you can still directly call v2m(.., 'cgalmesh') to create a tetrahedral mesh. then scale the node coordinates by voxel sizes, this creates an anisotropic tetrahedral mesh. You can then pass this stretched mesh to meshrefine() and play with the opt parameter. As long as opt imposes a more stringent quality or size constraint, the anisotropic elements will be refined and make the mesh more uniform.
the x/y/z scaling factors can be directly passed to v2s/vol2surf via the diagonal elements of the opt.A field, see
Qianqian
On 3/16/23 02:15, Changran Geng wrote:
Hi all,
I have a mask from the CT image, and trying to transform it to mesh. I have a following question.As we know the dimension of the voxel of the CT image is usually not uniform, for example to say the dimension of each voxel can be 1mm *1 mm* 3 mm.When I use the s2m or v2m, how could I tell the functions the above information?
Thank you very much for help.
Best,Changran--
You received this message because you are subscribed to the Google Groups "iso2mesh-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to iso2mesh-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/iso2mesh-users/f9b69390-75b5-43d6-82a9-03c6f898bd23n%40googlegroups.com.
--
You received this message because you are subscribed to the Google Groups "iso2mesh-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to iso2mesh-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/iso2mesh-users/CANMKZWfvCmXiMUwXOX2XMzdnTYheUVEi5X8Jqgeu8T4%3DXqRNJg%40mail.gmail.com.
Changran Geng
<head_mesh.png>