In this post I continue the presentation of our statistical running models from previous posts but focus more on what we might be able to compute that could directly benefit runners and our knowledge about running biomechanics in general.
To briefly recapitulate, let me mention that we have developed a parametric running model using machine learning techniques on a relatively large sample of real runners that have been measured with motion capture technology. You can read about these developments as far as movement goes in previous blog posts.
The new stuff is that we have now hooked the running model up with detailed biomechanical models that are able to simulate the loads on different tissues depending on the running style.
If you have read the previous posts, you will notice that the main difference is the addition of muscles, which are essential for a complete biomechanical description of the tissue loads.
Let us do an example: The patella ligament connects the knee cap with the bone in the lower leg (the tibia). This ligament transfers almost all the muscle force necessary to extend the knee, and it does so thousands of times during a normal run. It is one of the structures in our bodies that carries the highest loads. This can lead to injuries such inflammation in the tissues, also known as jumper’s knee. The condition is rather painful and very difficult to get rid of, once you have it. Some physio therapists believe that forefoot running will decrease the load on the patella ligament. So, should I take up forefoot running if I get in trouble with my patella ligament?
When scientists ponder how to go about such an investigation, they think about the concept of validity. In brief, validity distinguishes to which extent you are measuring the object of interest directly. The most valid way to investigate the effect of forefoot running on patella ligament loads would be to measure the ligament force inside the living human body. Unfortunately, this type of empirical investigation is very difficult to do. This load is inside the body and we cannot measure it on a living subject for technical, practical and ethical reasons. We could investigate the injury frequency in neutral versus forefoot runners, but they are different people and may differ in many other ways than their running style; we would probably need a very large cohort of test subjects and have to follow them closely for a long time. In cases like these, a computational model may in fact turn out to be more valid than most experiments we could do in practice.
The statistical running model has the advantage that it allows us to do an “all things equal” type of investigation. The model is only a model and not a real person (so it is less valid in that respect), but it allows us to get rid of all the other disturbing factors that can change in real test subjects, while we monitor them, and it allows us to gauge the ligament force directly and not through a derived property such as injury frequency or reported pain.
Let’s try it out. I start by defining a body similar to my own:
Then I specify my running speed at 5 min/km and a neutral foot strike pattern. I could also input step frequency, step length and even input from a running watch or another gadget if I had them.
Then I simulate my running style, and I get this handsome guy:
Notice that the center-of-pressure under the feet begins at the heel and moves forward in the stance phase as we would expect from a neutral foot strike pattern. I can now get the biomechanical parameters of interest from the simulation:
My neutral running pattern creates a patella ligament force of 6.31 times my body weight in the left knee and 6.47 times my body weight in the right knee. It is customary to report figures such as these in terms of body weight to enable comparison between individuals of different sizes.
Why are the data not symmetrical? This is because the movement, from which they are computed, is based on statistics from many runners, and most of these are right-dominant and therefore put a little bit more load on their right leg than on the left leg. I am also right-dominant, so this is all well and good.
I now move the Heel <-> Forefoot slider a bit to the right to simulate a forefoot strike pattern, and then I re-simulate the run and obtain this:
The center of pressure now begins at the forefoot as it strikes the ground as the first. Then it moves back as the foot flattens a little on the ground in mid stance, and forward again for to-off, just as we expected. But is this good or bad for the patella ligament forces? Well, the output will tell us:
The patella ligament force did indeed reduce in both legs, so the PT is onto something, and forefoot running might be a good idea for me if I have trouble with a jumper’s knee. However, if we look at the numbers for Achilles tendon force, then it goes up from 8.34BW to 8.76BW in the left leg and from 8.71BW to 9.49BW in the right leg. This means that, if Achilles tendonitis is my problem, then forefoot running is probably not the solution.
You might also notice that the peak ground reaction force increases a little bit from neutral to forefoot running. This might be counter-intuitive to some, because we think of forefoot running as having a cushioning effect. Cushioning might actually be real exactly at foot strike, but the foot rolls a bit faster off the ground in forefoot compared to neutral running, i.e. it is on the ground a bit less of the stride time. The shorter stance time means that you must apply larger ground reaction forces to keep up with the impulse of gravitation, which is the same regardless of running style.
The patella ligament and Achilles tendon forces we show here are just two among literally thousands of body loading parameters that we can derive from the model, and we can investigate their dependence on running style and body composition parameters to hopefully help runners steer clear of injuries.
More to come…