simtk.org

In the recent release of the E.Arnold Lower Limb Model 2010 there is a second, stronger version of the model that is described as CMC friendly. I've been doing some work trying to compare the CMC muscle activations to recorded EMG in normal gait and, similar to other folk I have spoken to, I have found that the models don't appear to be strong enough to track the recorded kinematics. However, I am concerned that simply increasing the strength of the actuators is a "quick fix" rather than a physiologically valid solution. I feel this problem is made worse by the specific tension value of 61N/cm2 that comes from Delp's PhD work. Most other publications give values about 3 times lower. Assuming these lower value are correct then using a 50% muscle strength augmented model will therefore give muscles 6 times stronger than would be physiologically valid.

Have others been carrying out CMC analyses and been having the same problems? Do anyone know of any other solutions rather than augmenting muscle strength? (eg altering tendon length/stiffness to ensure the muscles operate at their optimum length).

This is a very important issue. It is my experience (and the experience of many others) that if the maximum isometric joint torque v. angle curve of a model matches the curve obtained from maximum voluntary contraction experiments, the model will not be strong enough to produce the torques seen in a movement such as walking or jumping. If you do inverse dynamics on a maximum height jump or even walking, you'll see higher joint torques than the peak of the max iso curve.

When I ran CMC with this model, I started by including strong reserve actuators. This ensured that the simulation finished and I had some results to analyze. You can tell that the model's not strong enough if the muscles max out in activations and it STILL has to rely on these reserve actuators to track the motion. Check out the plot from my early CMC simulations that I've added to the downloads section under "Plots to Share". Tibialis anterior maxes out and there's high demand on the dorsiflexion reserve actuator.


So, if the model's not strong enough, what's a biomechanist to do?
You could change the tendon slack length to bring a muscle closer to optimal fiber length, but I think this is less physiologically valid than changing F_max. One goal for this model was to capture the joint angle-fiber length relationship that was measured in cadavers by Ward et al so that we could use it to examine muscle structure and function. Changing this relationship might produce stronger joint torques (though I sort of doubt it, I think it would just move the location of the peak) but the model would no longer be valid for looking at fiber lengths. Plus, if you changed slack length, what would you change it to?

I chose to augment strength by a consistent multiplier for all muscles, here's why:

-- We suspect that this problem has to do with the experimental measurement of torque. It's likely that subjects are not reaching their true maximum force generation when they're just pushing against the machine, with someone yelling "push push push!" but they CAN produce high torques in a coordinated movement.

-- It is, in my mind, the closest physiologically to the idea that the MVC measurements of torque are submaximally activated consistently across the entire range. This makes the boosted strength both a quick fix, and physiologically reasonable.

-- My modeling philosophy is to keep my assumptions as simple as possible unless I have data to support something more intricate. A single multiplier for all muscles was simple, and I played around with the value to make the model only as strong as it needed to be to walk.

-- There is precedent for this. The gait_2392 model is based on Delp's 1990 model, but the F_max values have been multiplied up to 3x to make it strong enough to walk. That modeler decided to give each muscle group a consistent factor, but I think the hips are the largest at 3 and the dorsiflexors closer to 1.5.

It is important to remember that the way this model is set up, tendon strain is scaled to F_max, so changing F_max does effectively change tendon stiffness (stronger model -> higher F_max -> stiffer tendon). This is why I boost the strength as little as possible. You could make the muscles 10x as strong, and the model could do anything, but the tendons would be unrealistically stiff. If you find the model needs to be a lot stronger for your application, you should consider modifying the tendon curve.