Left-Right Data Mismatch in Shoulder Range of Motion Test

[Catherine X]

I'm analyzing data from a shoulder range of motion test where participants simultaneously move their left and right arms in sync. The test includes a calibration move (an arm punch motion) followed by repeated movements of both arms.

The a graph of the dashboard kinematics is attached.

During the synchronized motion phase, the data aligns relatively well, which is expected. At the end, the two arm flexions show a mismatch, with the left arm flexion drifting to be higher than the right. This is not the case in the video, where the right and left arms roughly mirror one another.

Why might the left and right arm data not match up at the end? What steps can I take to eliminate this mismatch in post-processing? Any general recommendations for these cases of data drift?

Thank you in advance for all the help.

[Matt Petrucci]

HI Catherine,

Seems like it might be some weird wrapping that happened in IK (i.e., the purple goes around 0 when it hits around 90 degrees).

Which model did you use? Was it with the ISB shoulder?

You might want to download this session, and then use the .trc file and the model file to recompute IK in OpenSim to see if it happens again. https://opensimconfluence.atlassian.net ... Kinematics

Hope this helps,
Matt

[Catherine X]

Hi Matt,

Yes, it should be the ISB Shoulder (LaiUhlrich2022).

Below are the plots from the same video recomputed twice using IK in OpenSim for comparison:
While the left-right arm mismatch is not as drastic as on the OpenCap kinematics dashboard from my previous forum post, it is still significant. From the video, the individual's arm does not reach over 180 degrees, and thus the peak at around 20.5 seconds in the first recomputed IK doesn't match.

Screenshot 2024-12-07 at 10.12.15 AM.png (508.23 KiB) Viewed 370 times

As a side note, the total squared error in the mismatch region is higher than that of other time intervals:
Frame 1221 (t = 20.6949): total squared error = 0.0116654, marker error: RMS = 0.0164708, max = 0.0316536 (r_mwrist_study).

Any recommendations on how I should proceed? Thank you!

[Matt Petrucci]

Hi Catherine,

Quick question, were both of these IK runs done with the same settings?

Apologies that I forgot to mention that the IK setup in OpenCap for the ISB shoulder model is here: https://github.com/stanfordnmbl/opencap ... oulder.xml

[Catherine X]

Hi Matt,

Yes, both of them had the same settings. Thanks for pointing out the .xml file, as I had all the default weights set to 1.0 for the two plots in my previous reply.

Now importing the ISB IK shoulder .xml file and running it twice, I am getting the following:

Screenshot 2024-12-09 at 12.55.21 PM.png (546.5 KiB) Viewed 296 times

The settings, for reference:

Screenshot 2024-12-09 at 12.48.57 PM.png (201.91 KiB) Viewed 296 times

Screenshot 2024-12-09 at 12.48.32 PM.png (165.13 KiB) Viewed 296 times

[Matt Petrucci]

Hi Catherine,

Thanks for sending. What you may want to do is inspect the TRC file (or visualize in OpenSim) for any big jumps in the the shoulder or other arm markers around that time point (~19 s). That would throw off the IK solver.

Matt

[Catherine X]

Hi Matt,

There are some larger jumps of (~0.011 m across a time snapshot of ~0.017 s) on both the right and left shoulder values, would this be large enough to throw off the data?

Thank you!

[Matt Petrucci]

Hi Catherine,

That could certainly affect things. You could also verify in your other trials that don't have this problem to see if you observe similar jumps.

Hope this helps,
Matt