Average metabolic cost calculation using CMC

[Sebastian Neuberg]

Hello,

my question refers to the example Simulation-Based Design to Reduce Metabolic Cost.

I try to understand the output of metabolic probes and how to process adequately.
I've followed the guideline step by step. I haven't changed the CMC configuration which is provided in walk_Setup_CMC.xml, and the metabolic probes are as added by addMetabolicProbes.py. I've attached the MetabolicsReporter_probes.sto (renamed as csv) output, so that my data is traceable.

As I understand the API Doc of Umberger2010MuscleMetabolicsProbe the values stored in the file are muscle metabolic power in Watts. The column metabolics_TOTAL is the sum of all muscles.

I've calculated the average metabolic cost as follows:

Sum of all metabolics_TOTAL values = 961542.654 W
Amount of values = 1614
Average power per gait cycle = Sum of all metabolics_TOTAL values / Amount of values = 961542.654 W / 1614 = 595.75
Normalized by body weight = Average power per gait cycle / body weight = 595.75 / 72 = 8.27 W/kg

I am unsettled, as the value of 8.27 W/kg per gait cycle for walking at 1.2 m/s seems clearly too high to me.

Am I missing something or do I have a mistake in my calculation?

Furthermore the Question 3. in the Section Analyze the metabolic calculations for the walking simulation also states that common values might be between 4-6 W/kg. In fact, since the simulation only considers the lower extremities, I would have expected values between 2-4 W/kg only.

Most recently the sentence "You can compute the integral offline or re-run CMC with the probe_operation set to 'integrate'." confuses my. When I integrate the power I get the energy, which I am not interested in this context.

Hopefully someone can enlighten me.

[Thomas Uchida]

I've calculated the average metabolic cost as follows:
Sum of all metabolics_TOTAL values = 961542.654 W
Amount of values = 1614
Average power per gait cycle = Sum of all metabolics_TOTAL values / Amount of values = 961542.654 W / 1614 = 595.75

To calculate average power, you integrate the metabolic cost in Watts over time in seconds (giving Joules) and then divide by the duration in seconds (giving average Watts).

[Sebastian Neuberg]

Thank you very much for your answer. However, I am still puzzled.
Based on your answer I've executed CMC for unassisted gait again nut, I've set the probe operation of the metabolic probe to "integrate". Which seems to be the easiest way to integrate the metabolic cost over time. Hope that is right?
As I see in the column metabolics_TOTAL of the MetabolicsReporter_probes.sto the output adds up the energy for each line. I derive that the last value represents the whole energy which is consumed during the gait cycle. The gait cycle in the guide lats from 0.69 to 1.89 seconds. The last value in metabolics_TOTAL for time frame 1.89 is 707.299 (J),
When I divide that value by the duration of 1.2 seconds I got:
707.299 / 1.2 = 589.416 J/s
Normalized by body weight = 589.416 W / 72 kg = 8.19 W/kg
The value is hardly lower than the one with my previous calculation and does not seem plausible to me.
Am I still doing something wrong?

[Sebastian Neuberg]

Hi Thomas, can you relate to my irritation?
Do I still have an error in my consideration?

Could you maybe share the code you used to derive the data from your work "Simulating Ideal Assistive Devices to Reduce the Metabolic Cost of Running" from the CMC output?
That might help me build a better understanding.

[Thomas Uchida]

Unfortunately, I haven't run through the "Simulation-Based Design to Reduce Metabolic Cost" example in several years. However, I see that it uses the gait10dof18musc model, which is simple and therefore convenient but may not be suitable for research studies. Please see the "Musculoskeletal Models" page in the documentation (https://simtk-confluence.stanford.edu:8 ... tal+Models) for a description of the model and its limitations. If you plan to use metabolic cost in your research, you might want to use the Rajagopal model (https://simtk.org/projects/full_body/) or gait2392. Please also check that the properties of the metabolic calculator have been appropriately set; the properties are described in the API documentation (https://simtk.org/api_docs/opensim/api_ ... Probe.html). Finally, note that in past studies, the metabolic calculations have been less reliable for predicting absolute values and more useful for predicting increases or decreases from a baseline condition.

[Sebastian Neuberg]

Hello Thomas for my research I am using the gait2392 model. I used the gait10dof18musc model only because it is used in the example and I want to get a deeper understanding for the metabolic cost research.
I've read the API, the Umberger2010MuscleMetabolicsProbe Class and the related research topics carefully and think I've setup my research properly.

You wrote:

Unfortunately, I haven't run through the "Simulation-Based Design to Reduce Metabolic Cost" example in several years.

Does this mean you don't have the code available which derived your research data from the CMC output? Understanding how you refined the CMC output would be very helpful for my own research which is based on yours.

Lastly you wrote:

Finally, note that in past studies, the metabolic calculations have been less reliable for predicting absolute values and more useful for predicting increases or decreases from a baseline condition.

Can you maybe name one of these studies so I can take a closer look?