Abnormally High Joint Contact Forces

[Edward Syrett]

Hi all,
I'm relatively new at OpenSim, so this may be a somewhat obvious question. I've been trying to run simulations that assess the axial joint forces in the knee during gait. I have been trying to run baseline normal walking trials to assess if my models are working correctly using the 2392 gait model. I first have been running inverse dynamics to determine if the moments generated are comparable to those calculated with my data collection and analysis software (Visual 3D). They appear to be similar to that gathered experimentally. However, after running through RRA, CMC, and Forward Dynamics, my axial joint forces come out 200-300% higher than I would expect. They are typically 5-600% body weight, whereas all the literature I can find states peaks out at 300-400% body weight. I do not have EMG data to compare muscle activations, so I am only using the default CMC setup files that have been given with the program.
I believe that I am using the CMC and RRA steps correctly, but I may be missing something. I have been going through iterations of RRA to reduce the residuals to acceptable levels before going through the CMC and Forward steps. I have tried altering tracking weights to see if that would change values, but it does not appear to be significantly changing the values I am receiving at the end result.
Has anyone done this before and could give me some ideas on some things to try?

Thanks!

Dan

[Lauren Sepp]

Hi Dan,

I would say the first step in understanding your results is to check if you CMC reserves are unusually high. I would also recommend that you check you IK export from Visual 3D - this can be tricky depending on if you are using their pre-programmed export methods, or building your own IK chain and solution and adjusting your model. Some of the body origins and orientations aren't the same in Visual 3D as OpenSim (namely the pelvis and the ankle angle) so if you are using your own IK solution, you must manually adjust for this.

Have you tried using the Analyze tool only without using the Forward Dynamics GUI?

Best of luck!

Lauren

[Ross Miller]

Hi Dan,

How are you calculating the joint contact forces?

The first thing I check in cases like this is if bodyweight is set to the subject's weight in the osim model and in the value the forces are ultimately scaled by (if done outside of OpenSim). The second thing is if the knee moments fed to CMC are unusually high in Nm/kg (although it sounds like you did this already). If neither of those things look unusual then I can at least be sure the issue is probably not the data.

Peaks of 5-6 bodyweights would be very high for an average over numerous subjects, but if that's a result from just one subject, there are modeling studies that include that range. For example Richards & Higginson (2010) used ~ the same method you described and their peaks were (mean+/-SD) ~ 4.4+/-0.8 BW. Winby et al. (2009) used a different approach but also had high peaks.

https://www.ncbi.nlm.nih.gov/pubmed/20627301
https://www.ncbi.nlm.nih.gov/pubmed/19647257

Hope this helps,
Ross

edit: typo

[Edward Syrett]

Thank you both for the replies. Sorry I have been not able to respond until this point. I have taken both of your suggestions, and managed to reduce my forces by about 25%.

I am using Open Sim to do all of the analysis. I have been using the provided MatLab code to split the Visual 3D files into .mot and .trc, then performing all of the IK, ID, RRA, and CMC through the GUI in OpenSim. I have made sure to scale the model with the subject's appropriate weight (as measured by the force plates in Visual 3D). I will run a moment analysis on the data through Visual 3D, then compare that to what is generated by the RRA in OpenSim, which generally agrees within +/- 5 nM. I am making sure that the residuals of both forces and moments are within the acceptable limits for both the RRA and CMC steps, then running the joint reaction analysis through the analysis tool rather than the forward dynamics tool. I am still getting high contact forces, bit I am currently in an iterative process of playing with tracking weights to see if I can affect those.

Thanks again,
Dan