[aop]

I'm new to ANSYS and I'm working on a project which involves modeling part of human thorax.

I have segmented the tissues of interest in ITK-snap and exported them as .stl files which I have cleaned up and simplified with MeshLab and NetFabb. After that I used SpaceClaim to convert the individual .stl parts into solids and did the required boolean operations to fir the lungs, bones and cartilage in the thorax with DesignModeler.

The bones, cartilage and lungs mesh just fine but the meshing of the soft tissues fails every time regardless of settings. I believe there are some geometry errors left by the boolean operations performed in DesignModeler as the meshing error points to few specific faces (see the attached images).

Any idea how to fix the geometry? Repair functions in DesignModeler don't find any issues.

[divergence]

I might not understand the situation correctly here but let's try to help anyway:

Is the red thing in the first figure one volume or several? Sometimes it helps to try to narrow done the problematic volume by trying to mesh only one volume at the time while the other volumes are suppressed. This could provide you with additional information about the troublesome part.

You could also try to change the "Physics Preference" in Mesh Details into CFD, since Mechanical is usually waste of time in CFD cases. Additionally, you probably should try to mesh the volume with tetrahedral cells, which adapt usually easier into difficult places.

One could also try to use the "Share Topology" tool in DesignModeler, which can point out some errors or problems in your model. You seem to have funneled the model through many programs, which could lead into some errors in the model - details might get lost in translation.

I'm afraid these points are very general but they might help you out in finding the reason for such an anomality. Personally I haven't faced such trouble with DM and Meshing.

Hope you get ahead with the task!

[Kapi]

post your geometry first and point out the areas whr you are getting error.

Also, reduce the size of model, use a section of whole model to see the effect.

the pictures you posted doesnt help understand the problem.

[aop]

Thanks for answer.

I managed to solve the issue by importing the model piece by piece and making sure the connections between individual bones and cartilage were good while the files were still in STL format. Reducing the polygon count of the original STL files exported from ITK-snap was the culprit, there was slight shrinkage of the parts and the connections between cartilage and bone weren't perfect anymore and this resulted in slivers of soft tissue being between them in the final model. These slivers were too thin to be meshed properly so it caused the errors.

I worked around this by extending the cartilage pieces by 0.5-1.5mm on each end and performing boolean operation where I subtracted the connecting bones from the cartilage piece.

I know this isn't ideal but I don't see an alternative when working with the STL files.

I wonder if there's a better way to import data that was originally in voxel format to ANSYS.

Currently the workflow is a bit messy:
1. Voxel data in ITK-snap, VTK-format.
2. Export function in ITK-snap converts the voxel data into STL files with marching cubes algorithm. This results in STL files 1-2 million polygons each for each tissue type.
3. I used MeshLab to start reducing the polygon count and to smooth the models a bit. I used MC edge collapse (ment for reduction of polycount for geometries created by marching cubes algorithm) to reduce the initial polycount and then Taubin smooth to smooth out the surfaces a bit (since the CT images I used to create the geometry had 0.5mmx0.5mmx0.7mm voxel size the model surfaces were pretty coarse).
4. I imported these files to Autodesk Netfabb since it has polygon reduction tool which allows me to specify maximum allowed deformation of the model while reaching minimal possible polygon count. I specified maximum allowed deformation to 0.5mm (from voxel size of the original CT-image, the models aren't going to be that accurate anyways).
5. After this I needed to match the bones and cartilages together again because simplification operations had created small (0.5mm-1mm) gaps between them due to model shrinkage. I did the first step in Netfabb by using the repair tool and extrude surfaces tool to lenghten the cartilage pieces from each end.
6. When I had finished with Netfabb I imported the STL files created there to ANSYS SpaceClaim which made them into solids. I also verified the geometries and checked for too small faces in SpaceClaim. If there were any problems I returned to Netfabb to fix them and then reimported the files to SpaceClaim to check them again.
7. When the STL files were converted into solids in SpaceClaim I imported them to DesignModeler started putting the model together. I started building the model by importing a block which contained all the tissue types. I imported each piece of cartilage and the bones they were connected to one by one and used the boolean subtract operation to fit them together (cartilage minus bones). Then I subtracted the bones and cartilage from the soft tissue block to fit them in.
8. After each inserted bone and cartilage piece I checked if it would mesh correctly and run a simulation in ANSYS Workbench. If there were any problems I had to go back all the way to step 4. or 5. to fix it in Netfabb.
9. If the inserted parts worked correctly I would go back to step 7. to fit the next part to the model.

I repeated this process until I had all the 18 bones, 16 costal cartilage pieces and the lungs fitted into the surrounding soft tissues and the simulation completing without any errors.

It was real pain in the ass to do it this way but I really don't know a better way with tools I have available. Some software like Simpleware FE+ could the voxel data ---> solid model conversion in one step but it costs about 10k€ and the University doesn't have licence for it.

I hope these steps can help someone else who has to convert medical imaging data for finite element modeling.

[divergence]

Glad to hear you succeeded in fixing your volumes!

The amount of money companies ask for seemingly simpe additional software for model handling (for instance .sat or .stl files for DesignModeler). I just checked that ParaView - which is a free program - seems to have support for both VTK files as well as stl files. Sure enough, ParaView is not drawing program but I thought you might be interested it out in case you might find some shortcut from the program.