Modified PiG markerset in OpenSim for static optimization

[giulia mantovani]

Hello,
I am an OpenSim new user.
I have data collected in Vicon according to a modified version of Plug-in-Gait (namely UOMAM) that has extra markers on the medial knee condyles and medial malleoli for the calculation of the knee and ankle joint centers. My purpose is estimating muscle forces. I was able to make all the appropriate conversions to have my c3d files properly running in OpenSim, by modifing the Matlab pipeline and the lower_limb.osim model developed by Glen Lichtwark (thanks!). The comparison between OpenSim and BodyBuilder IK and ID results is pretty scary ! I read a very useful post talking about the inverse kinematics data validation in https://simtk.org/forum/message.php?msg_id=1943 , and, as I thought, the main problem is probably the different joints definition between the model in OpenSim and the modified PiG in BodyBuilder.

So now I have a dilemma to solve! If I want to get muscle forces, what is the more appropriate way to proceed?
1) Rerunning all the IK and ID analyses in OpenSim and then using the results to estimate muscle forces with, for example, static optimization.
2) Using the IK and ID results from BodyBuilder properly converted into OpenSim compatible formats and, from here, continuing to estimate muscle forces.

Both options have several drawbacks. In the first case, I think that the employed markerset is not appropriate for OpenSim because PiG (and UOMAM) marker locations are not as good as a cluster distribution (such as the one used in Gait2354_Simbody.osim) for the identification of the segment movements.

In the second case, the IK and ID results from BodyBuilder would not be consistent with the OpenSim model used for the optimization. Therefore, IK/ID results would be based on different assumptions than the estimations of muscle forces.

Which one do you think is more appropriate? Do you have other suggestions or options how to solve this problem?

Thank you very much, any feedback will be appreciated!

[Ayman Habib]

Hi Giulia,

I'd strongly recommend your first approach because:
1. Do yo have a model that has been scaled to match the subject? If not that can be a big source of errors downstream in IK and later.
2. It would be really hard to convert ID results to OpenSim when the joint definitions in the two systems are not known or different.
3. The resulting muscle force estimates would change depending on residuals (forces at the base segment that should be minimal and that are usually due to modeling assumptions). I'm not sure if or how these are reported by BodyBuilder.

Hope this helps and best of luck,
-Ayman

[giulia mantovani]

Thank you for your suggestion!
Yes, I did scale the model but, probably, I still have some adjustments to make in terms of filtering and other IK parameters.
No, BodyBuilder does not give residuals.
I will implement the first solution, then.

Anyway, where can I find info about which joint definition is used in OpenSim?

Thank you again!

[Ayman Habib]

Hi Jiulia,

Please consult the user guide/documentation (link below) for details regarding what joint types are supported and how they relate to the joint specifications in .osim files and let us know if you have further questions or feedback so we can improve the documentation.

Best regards,
-Ayman

http://simtk-confluence.stanford.edu:80 ... els-Joints