double shape bone (scapula)

I am currently working on a model of total shoulder arthroplasty and the next step would be to evaluate the stress related to the resultant contact force on the glenoid implant. I also would like to take into account the possible micromotions at the interfaces bone/ciment and ciment/implant so a segment representing the cortical and trabecular bones appears important.

I used the Anybody scapula segment and would like to do all the process with this one (rather one reconstructed from CTscan). But, the geometry does not integrate a double shape (to simulate cortical and trabecular bones)

So my first question would be : have you a double shape model corresponding to the scapula segment ?

If not, my idea would be to use the extern shape of the scapula to create an intern one. I was thinking to the possible use of a scale function to create a new anysurf segment. Maybe could I then export it into .stl and create, via a boolean function, a new stl with the extern and intern shapes ?

Would you have suggestions about this process of creating a double shape model based on the anybody model of the scapula ?

Thank you !

Hi Lauranne,

We don’t have inner cortical surfaces for the bones. So you are right - 2 options:
a) use a CT-derived model as a target and scale AnyBody bone into it
b) scale AnyBody STL by say 90% and make an inner surface, say in Meshlab, by combining these surfaces. (Or you can just copy faces in ASCII mode into each other). Please be aware that the uniform scaling may move the bone.

For the option A you would need to use an affine scaling (AnyFunTransform3DLin) or advanced scaling tools such as AnyFunTransform3DRBF and AnyFunTransform3DSTL. A comprehensive tutorial should be available starting from the version 5.2. But the morphing tools are available now.

Please also be aware of the FE converter that we developed to assist with transferring forces/moment into an FE model. You can see it in the FE tools section on the website.

Hope this helps,

Hi Pavel,
Thanks for your complete answer.

I would like to clarify the consequences of each option :
a) could the fact to use a CT model conduct to some difficulties to perform the inverse dynamic ? I mean that probably the insertion does not correponsd exactly to what it is in the Anybody model and so the muscular forces calculated does not correspond between the model and the CT-derived model. What do you think about the implication on the results, especially in terms of accuracy and limitation of the method ?

b) in fact I tried it right before writtent my thread. I indeed used meshlab, scaled to 90% and align the both meshes. But the problem comes from the specific shape of the scapula : the alignement does not allow an appropriate alignement due especially to the not constant cortical thickness.

Regarding these 2 inherent problems attributed to each method, do you have any suggestion ?

Thank you for your advice about the FE converter. In fact I wanted to use the Any2Ans software. Unfortunately I encountered a problem linked with my method : I use the FDK and AnyForceSurfaceContact to calculate the glenohumeral reaction force between the 2 sides of the implant. The Any2Ans software does not integrate this method and I am unable, at this time, to export such a result. What would have been interesting for me would have been to export resultant contact force, point of application and surface area easily to use it as constraint in my FE model. Do you think that it is something possible with the FE tools ?

Thank you again for your help !

Hi Lauranne,

a) No, it should not cause any problem. Although I have not tried to morph the scapula itself, but I have tried to morph pelvis and vertebrae (and those are quite complex bones). It worked quite well. I would not expect to be problematic, but it may not be easy.
To explain: the procedure requires registering objects to each other in a non-rigid manner, which can be done by means of classes AnyFunTransform3DRBF, AnyFunTransform3DSTL, and/or AnyFunTransform3DLin. If two of your scapulas are topologically similar - you may just try to perform an affine scaling using the latter one. What you would need to do is to morph the full-body using anthropometric scaling, but then replace the scapula segmental scaling function by this non-rigid registration transform.
The problem here is that we are planning to write a tutorial on morphing for the next release 5.2. It can also be done with 5.1, but it may be a little more complex. If you decide to go with this approach I will guide you.

Please also be aware that you may need a corresponding CT-reconstructed humerus for better results.

b) Yes, I understand. I guess it is still an assumption, so you may apply smoothening filters in Meshlab to your inner surface (90% one) a few times. This step will remove all sharp outstanding features and then the alignment may go smoother. But that leads to a loss of some accuracy. I would personally go with the a) option. (but probably because I partially implemented the workflow)

Converters: If you are just planning to do what you said - then you could simply output required kinematic and kinetic data for both segments from AnyBody and use these results to set up the boundary conditions. I think that can be done rather simply. The converters are good for transferring all present forces onto an FE model of a bone. I am not sure about AnyForceSurfaceContact (I don’t think we used FE converters with this class), but FDK should not cause any problems.

Best regards,