Page 1 of 1

Problem with scaled model

Posted: Wed Sep 09, 2020 4:59 am
by sheebadavis
Hello OpenSim experts!!!

I have tried to scale my model based on a walking dataset. In the GUI, my scaled model is in a horizontal position. It clearly shows that it has not been scaled properly. But I want to know where might be the problem. Following is what I have done,

1. The original model didn't have all the markers that are in the dataset and also the names were different. so I changed the marker names according to the dataset and also introduced new markers (placed them manually) to make sure that it has all the markers that are present in the dataset.
2. I already have a matlab code which worked well with a different dataset. so I used the same code and slightly changed according to the requirement for the new dataset. I have also removed the markers from the model which are not there in the dataset.
3. Now the problem comes with the settings file (measurementset and task files). Particularly the measurementset file, as I have removed some markers, I could not use some of the marker pairs and so made them 'false'. Will this affect my output? Do i need to keep those markers (which I removed from the model) even if they are not included in the dataset? Will these markers be required for the settings file? Similarly with the task file, I made the markers 'false' (the ones which are not included in the dataset.
4. I also have another doubt regarding rotating the markers. In the walking dataset, the y axis coordinate for COP had 0s. So I assumed that y is the vertical axis. As we know that in OpenSim, y axis is vertical and so that means I don't have to rotate the markers. Is my statement right?

Can someone please help me with my doubts? Thanks in advance.

Sheeba Davis

Re: Problem with scaled model

Posted: Fri Sep 18, 2020 11:39 am
by mjasmuss
Hi Sheeba,
1. The original model didn't have all the markers that are in the dataset and also the names were different. so I changed the marker names according to the dataset and also introduced new markers (placed them manually) to make sure that it has all the markers that are present in the dataset.
I think this is fine, but you could run into issues where you manually placed the markers and which reference frame you used for marker location (e.g., global, tibia, femur)
2. I already have a matlab code which worked well with a different dataset. so I used the same code and slightly changed according to the requirement for the new dataset. I have also removed the markers from the model which are not there in the dataset.
If the markers that you removed were essential for something like scaling a segment, it could introduce issues
3. Now the problem comes with the settings file (measurementset and task files). Particularly the measurementset file, as I have removed some markers, I could not use some of the marker pairs and so made them 'false'. Will this affect my output? Do i need to keep those markers (which I removed from the model) even if they are not included in the dataset? Will these markers be required for the settings file? Similarly with the task file, I made the markers 'false' (the ones which are not included in the dataset.
You can make the markers set to false, but as I mentioned above, this could create some issues with proper scaling if those markers you removed were essential. If these were tracking markers, you may be okay for the scaling procedure.
4. I also have another doubt regarding rotating the markers. In the walking dataset, the y axis coordinate for COP had 0s. So I assumed that y is the vertical axis. As we know that in OpenSim, y axis is vertical and so that means I don't have to rotate the markers. Is my statement right?
Could you explain this issue in a bit more detail? It might be best to post some sample data to get an idea of your issue.

I hope this is helpful!

Mike

Re: Problem with scaled model

Posted: Fri Sep 18, 2020 8:50 pm
by sheebadavis
Hello Mike

Thank you very much reply.
1. The original model didn't have all the markers that are in the dataset and also the names were different. so I changed the marker names according to the dataset and also introduced new markers (placed them manually) to make sure that it has all the markers that are present in the dataset.

I think this is fine, but you could run into issues where you manually placed the markers and which reference frame you used for marker location (e.g., global, tibia, femur)
Yes, you are right, manually placing them caused issues. so I understood that it was a bad idea. I think as you mentioned, there is a problem with reference frame which I understood from couple of previous forum posts that I checked. But I don't know how to check it, can you please let me know how to check the 'ref frame for marker location'?
2. I already have a matlab code which worked well with a different dataset. so I used the same code and slightly changed according to the requirement for the new dataset. I have also removed the markers from the model which are not there in the dataset.

If the markers that you removed were essential for something like scaling a segment, it could introduce issues
The problem is that experimental markers were only 20 and so I had to manage with what was available. Some of them I had to just rename (model markers) according to the experimental markers, whereas some of the markers like joint markers (Hip, knee) were not available in the expt marker set. so for ex: to scale femur, instead of using HJ-KJ marker pair I used GREATERTROCHANTER-L.KNEE marker pair. Simillarly, for pelvis height instead of ASIS-HJ marker pair I could only use ASIS-GREATERTROCHANTER marker pair. Will this be an issue?
3. Now the problem comes with the settings file (measurementset and task files). Particularly the measurementset file, as I have removed some markers, I could not use some of the marker pairs and so made them 'false'. Will this affect my output? Do i need to keep those markers (which I removed from the model) even if they are not included in the dataset? Will these markers be required for the settings file? Similarly with the task file, I made the markers 'false' (the ones which are not included in the dataset.

You can make the markers set to false, but as I mentioned above, this could create some issues with proper scaling if those markers you removed were essential. If these were tracking markers, you may be okay for the scaling procedure.
I actually misunderstood the function of TRUE and FALSE. Now I realized that FALSE will actually move the model marker according to expt marker. hence I set all the marker pairs which I want be used for scaling as FALSE. Will this be fine?
4. I also have another doubt regarding rotating the markers. In the walking dataset, the y axis coordinate for COP had 0s. So I assumed that y is the vertical axis. As we know that in OpenSim, y axis is vertical and so that means I don't have to rotate the markers. Is my statement right?

Could you explain this issue in a bit more detail? It might be best to post some sample data to get an idea of your issue.

GRW1
mm mm mm
X Y Z
2355.54 0 1120.33
2359.82 0 1117.11
2334.86 0 1127.02
2322.18 0 1144.71
2314.06 0 1147.24
2323.85 0 1133.92
2341.57 0 1126.62
2353.69 0 1122.13
2332.49 0 1130.01
2313.13 0 1142.39
2295.6 0 1171.09
2307.31 0 1167.01

What is shown above is a snippet of COP values in the gait trial file. so what I meant was, since Y has zeroes, it must be the vertical axis for this expt data. I then viewed the expt marker data in the GUI and transformed that data until it is oriented same as the model. then I used those angles of rotation to actually rotate the data in my matlab code. Will this be fine?

Sorry for such a long post. Thank you once again MIke for your reply. I did not expect a reply after 10 days :)

Regards
Sheeba