simtk.org

Dear Experts,

My femur model of a segmented CT scan has a round-shaped head exactly like a sphere (First image).
In the "bone generation workflow", I clicked several times on "fit" to ensure that the RMS is relatively small.
But the generated model had an ellipsoid shape (not sphere) and it didn't look like the target (Second image).
I know that the geometry of bone is just for visualization in OpenSim and has no effect on the result.
But overall are there any configurations to increase the accuracy of gias2?

Your response would be greatly appreciated.
Cheers

Hi Mohammad,

In the Documents heading, under User Guides you can find a document called Tips and Tricks, this contains some general advice on the fitting methods and settings to get the best results. You can also see the following paper https://journals.plos.org/plosone/artic ... ne.0205628 which discusses the different combinations of fitting methods to get the best results.

I am unsure of the methods used in the bone mesh worklflow you refer to, but you can use three different fitting methods which will give you a better fit.

Second, I should mention the shape actually does have more of an effect rather than just visualisation in OpenSim, it has an effect on the joint position and orientation as well as the muscle fixed and via points.

Best
Bryce

Hi Bryce,

Thanks for your quick reply.

brycekillen05 wrote: ↑

Wed Feb 12, 2020 3:11 am

Second, I should mention the shape actually does have more of an effect rather than just visualisation in OpenSim, it has an effect on the joint position and orientation as well as the muscle fixed and via points.

I'm not sure about this. Maybe there is a misunderstanding.
Have you ever tried changing just the geometries and compare the results?

Let me tell you about my experiences:

1) I rotated just the femoral condyles 35 degrees internally to simulate a patient with the anteverted femur (just the femoral condyles). all analyses (for gait2392 model) including IK, ID, RRA, JRF were the same for pre and post-modification. this is the knee joint contact force graphs: 2) just an hour ago, I changed the "r_femur.vtp" to "humerus_rv.vtp" for Rajagopal model. Again all the results were the same. Here is a comment from an expert in OpenSim forum:
"The bone geometry is used exclusively for visualization rather than for any computation, so the quality of the meshes has no role to play in the modeling of muscles or their attachments (other than eye-balling where the muscles attach)."

Indeed all parameters including the center of mass, inertia, mass, muscle points and length (all muscle parameters), references, locations and orientations values were intact and that's why the results were the same.

In terms of my main query, I've read the"Tips and Tricks" but I was looking for special values for that workflow to enhance the accuracy.

I'm going to read those methods and I appreciate it.
I will test them and wonder if I could ask more about it in the future.

Cheers

RE comment about geometries

In a generic opensim model they are correct BUT if you generate a personalised model through the MAP Client it DOES have an effect relating to the definition of muscle points. Just changing the geometry file does not have an actual effect in the model in the way you described because the muscle and joint features are NOT different which is what would cause a difference.

" but I was looking for special values for that workflow to enhance the accuracy."

What do you mean by special values, the settings should be listed in the tips and tricks document

Hi again,

e.g. Cheers

Some settings, and the weightings I used are listed on the second page of the document I mentioned before.

Other general setting recommendations can be found on the MAPClient FAI workshop page here :
https://map-client-fai-workshop.readth ... hgen.html

Step 10 and 11

Thanks Bryce