Hello,
I have a question about whether on not to include subtalar joint moment in equality constraint in custom static optimization [moment-matching approach].
I see a huge variability in the subtalar kinetics in some of the subjects in my data. These are multiple trials of one subject during walking (same leg):
If I lock the subtalar joint in the model, I will see unusual movements in the proximal segments, particularly in transverse plane. So, I unlock it in order to improve inverse kinematics.
If I include subtalar moment in the static optimization, I see significant difference in ankle joint contact force. This is ankle contact force in one trial (both legs):
I have no idea about what the source of such variability in subtalar moment would be, e.g., measurement or modelling errors. But the subtalar kinematics is much less variable. In this case, I think it's better to unlock subtalar joint but exclude it in static optimization. So, the results would be more reliable. Please let me know your thoughts.
Any help is greatly appreciated.
Regards,
Mohammadreza
Include/exclude subtalar joint in custom static optimization practice
- Mohammadreza Rezaie
- Posts: 388
- Joined: Fri Nov 24, 2017 12:48 am
- Nicole Strah
- Posts: 9
- Joined: Mon Aug 29, 2016 2:40 pm
Re: Include/exclude subtalar joint in custom static optimization practice
Hello Mohammadreza,
I've never tried the moment-matching approach, but just to check, do you have any other joint moments locked?
Edit: An example of this is the MTP joint; when using large treadmill force plates, the force from in front of the MTP joint cannot be distinguished from the force applied behind the MTP joint, and even then I would leave the MTP joint unlocked, run inverse kinematics, but then omit it from the static optimization calculation. I would leave subtalar joint unlocked if there is marker data to capture what is happening around the joint, and I would include the results in the static optimization. (You might have already considered all this; I will need to take a look at the moment-matching approach.)
Thanks,
Nicole
I've never tried the moment-matching approach, but just to check, do you have any other joint moments locked?
Edit: An example of this is the MTP joint; when using large treadmill force plates, the force from in front of the MTP joint cannot be distinguished from the force applied behind the MTP joint, and even then I would leave the MTP joint unlocked, run inverse kinematics, but then omit it from the static optimization calculation. I would leave subtalar joint unlocked if there is marker data to capture what is happening around the joint, and I would include the results in the static optimization. (You might have already considered all this; I will need to take a look at the moment-matching approach.)
Thanks,
Nicole
- Mohammadreza Rezaie
- Posts: 388
- Joined: Fri Nov 24, 2017 12:48 am
Re: Include/exclude subtalar joint in custom static optimization practice
Hi Nicole, thank you so much for your reply.
What you said makes sense to me. In my case (the Rajagopal model), I have only markers attached on heel, 1st and 5th met head on the foot and nothing on the toes. So, I locked the mtp joint and due to the reason you mentioned, I excluded the mtp joint moment in static optimization (SO).
Actually, I'm not sure whether these three foot markers are able to capture the subtalar joint kinematics accurately, because we need the position of rearfoot relative to the shank. In reality, we have multiple joints between calcaneus and metatarsal segments (e.g. Chopart and Lisfranc joints) and the forefoot markers do not represent the rearfoot movement and hence, the subtalar joint kinematics. I guess adding more markers or cluster markers to the rearfoot in our simulation (constrained one-segment foot model) will improve the subtalar joint kinematics/kinetics.
Image Ref: http://dx.doi.org/10.1177/1071100714559727 Image Ref: http://dx.doi.org/10.7717/peerj.3298
The constrained knee movement in the frontal and transverse planes would be another reason for this issue, I think.
There is also a note in custom SO example:
https://simtk-confluence.stanford.edu:8 ... +in+MATLAB
I think you are using acceleration-matching approach rather than moment-matching. I had asked about it in the forum and it would be great if you let me know more about it and how to implement it in this post:
viewtopicPhpbb.php?f=91&t=14668&p=42427&start=0&view=
Thanks again,
Mohammadreza
What you said makes sense to me. In my case (the Rajagopal model), I have only markers attached on heel, 1st and 5th met head on the foot and nothing on the toes. So, I locked the mtp joint and due to the reason you mentioned, I excluded the mtp joint moment in static optimization (SO).
Actually, I'm not sure whether these three foot markers are able to capture the subtalar joint kinematics accurately, because we need the position of rearfoot relative to the shank. In reality, we have multiple joints between calcaneus and metatarsal segments (e.g. Chopart and Lisfranc joints) and the forefoot markers do not represent the rearfoot movement and hence, the subtalar joint kinematics. I guess adding more markers or cluster markers to the rearfoot in our simulation (constrained one-segment foot model) will improve the subtalar joint kinematics/kinetics.
Image Ref: http://dx.doi.org/10.1177/1071100714559727 Image Ref: http://dx.doi.org/10.7717/peerj.3298
The constrained knee movement in the frontal and transverse planes would be another reason for this issue, I think.
There is also a note in custom SO example:
https://simtk-confluence.stanford.edu:8 ... +in+MATLAB
Given these reasons, I think it's better to exclude it. This is also one of the other advantages of custom SO compared with OpenSim built-in SO which every unlocked coordinate will be included in the analysis.Your static optimization solution should provide a reasonable estimate of muscle activity for walking. Therefore, we should only solve the problem across degrees-of-freedom with reliable experimental measurements that also contribute to the muscle activity solution. ... In addition, you may consider removing some muscle-actuated degrees-of-freedom that may have poor inverse kinematics solutions.
I think you are using acceleration-matching approach rather than moment-matching. I had asked about it in the forum and it would be great if you let me know more about it and how to implement it in this post:
viewtopicPhpbb.php?f=91&t=14668&p=42427&start=0&view=
Thanks again,
Mohammadreza