Marker registration - which markers should move?

[Cassi Henderson]

Hello,
I am trying to use OpenSim to determine ankle joint angles from mocap data using the Rizzoli lower body marker set.
In the tutorial video on the documentation site (https://www.youtube.com/watch?v=7STRMJefzpI&t=2s) it appears that all the virtual markers are moved to match the locations of their experimental counterparts as the last step in the marker registration. However in the webinar (https://www.youtube.com/watch?v=ZG7wzvQC6eU) it says that anatomic markers should be fixed and not moved during the marker registration process. Can you please clarify which is the correct procedure and/or why you might use one protocol over the other?
Thanks!

[Mohammadreza Rezaie]

Hi,

Marker registration is the process of positioning the model markers so that they match the locations of the experimental markers. Markers are typically registered using a graphical user interface (GUI), but this method is subjective and may introduce errors and uncertainty to the calculated joint angles and moments.

https://doi.org/10.1371/journal.pone.0252425

Typically joint center markers should be fixed to be always in the center of each joint. These are the only markers that we know their exact locations in the model. That's why they get the highest weights in MarkerPlacer (https://www.youtube.com/watch?v=ZG7wzvQC6eU) (26:40). Other anatomical markers have been registered manually in the model and I do not set them fixed.

Hope this helps.
MrR

[Cassi Henderson]

Hi MrR,

Thanks for your reply. To confirm my understanding, the only markers that should be fixed are the joint centre markers? In the video tutorial (32:09) it recommends against moving anatomical markers during the registration process, which I took to mean markers on bony landmarks (i.e. not tracking markers), however, it sounds like you leave all but the joint centres unfixed (so they move during registration).
On a related note, do you manually adjust the position of any markers on the generic unscaled model before you run the scaling process to minimise the marker error RMS?

Thanks!

[Mohammadreza Rezaie]

Hi,
Personally I only fix the joint center markers and let other markers move to match the experimental counterpart. As you mentioned there are different approaches. You might be interested in looking at the setup files of this project. The scaling of the walking trial is quite sophisticated and all anatomical markers were fixed, while in the running trial, none of them were fixed.
https://simtk.org/frs/download_confirm. ... oup_id=773
Sometimes you need to try different things to figure out the best approach that fit your data.
It may also depends on the weights of the markers in Inverse Kinematics (IK) tool. If you are going to allocate some weights to anatomical markers in IK, fixing them in scaling process may lead larger RMSE in IK. In this case, it doesn't make sense to me. If you are not interested in calculating any virtual markers (e.g joint centers or ground projected foot markers), that would be fine too. You can still achieve the realistic static position and optimal marker registration by adjusting scale factors, markers and coordinates weights.

do you manually adjust the position of any markers on the generic unscaled model before you run the scaling process to minimise the marker error RMS?

Yes, I do. They have been defined subjectively and its better to adjust them for your own experimental data. I adjust the marker set and use identical setup files for all cases.

All the Best,
Mohammadreza

[Cassi Henderson]

Hi Mohammadreza,

Thanks for your reply. In the example file you mention, do you know why markers are fixed in one case and not in the other? I also noticed in that example there seems to be 2 scale files for the walk model (ScaleOnly and scale_walk) and I was wondering what the purpose of having two files was?
When you say 'realistic static position' what exactly do you mean? Is there an accepted range of joint coordinate values for someone standing in the T-pose that you compare the coordinates of your scaled model against to check it's reasonable?
When you manually adjust your markers, how do you determine when they are in a 'good-enough' position? Do you use total marker error or joint coordinate values or some combination?

Thanks,
Cassi

[Mohammadreza Rezaie]

Hi,

what the purpose of having two files was?

"It is common to iterate through Scale and Inverse Kinematics to fine-tune segment dimensions and marker positions that yield low marker errors for the task of interest" (Ref: https://simtk-confluence.stanford.edu:8 ... ootingTips:).

When you say 'realistic static position' what exactly do you mean?

Is there an accepted range of joint coordinate values for someone standing in the T-pose that you compare the coordinates of your scaled model against to check it's reasonable?

Do you use total marker error or joint coordinate values or some combination?

Sometimes you get unrealistic/unusual static position compared to your observations of the participant's static position (It's recommended to capture multiple photos of that during data collection, Ref: https://simtk-confluence.stanford.edu:8 ... reparation:). In this case you can solve it by adjusting any of model's markers location, markers' weight or coordinates' weight. For example, if you get unusual pelvis list/rotation or torso bending/rotation, adjusting coordinates' weight is helpful. Or if there are unusual toe-in/out or pelvis tilt angles, adjusting markers' weight can fix them, and ...

When you manually adjust your markers

how do you determine when they are in a 'good-enough' position?

It is worth to check the "Preview static pose (no marker movement)" in the scale tool and use "Associate Motion Data" afterwards. You will see the location of model and experimental markers relative to each other. If they are too far apart, you can move the model's markers in the generic model. OpenSim recommends 1 and 2 cm for total RMSE and maximum error of markers, respectively (Ref: https://simtk-confluence.stanford.edu:8 ... ourResults:).

Sometimes it's not possible to get the errors in the recommended range. Personally I pay more attention to static position and markers' location on scaled bodies than the amount of markers' error, also, I do Scaling only one time. Regarding whether or not fixing the anatomical markers, I suggest you to try both approach on same experimental data to see which one fits your data. What I said in this topic was based on my own experiences, so I apologies for that and maybe it's better to wait for experts to respond.

Best,
Mohammadreza