How reliable is musculoskeletal modelling in producing replicable outputs?

The study delves into the mechanics of muscle function in the hominin Australopithecus afarensis (specimen AL 288-1, affectionately nicknamed Lucy), aiming to understand how different methods of estimating muscle structure influence our interpretations of limb function.

The team digitally reconstructed the muscles and tested their ability to support an upright, single-support limb posture.

The force a muscle generates is dependent on muscle architectural parameters, in which fibre length, pennation angle and tendon elasticity all influence force production and subsequent body movement.

However, muscles are rarely preserved in the fossil record and instead we are left with the bare bones. A scientist cannot simulate the locomotion of an extinct individual without first reconstructing the missing soft tissues of the limbs.

Last summer, Wiseman (2023) published the first 3D muscle reconstruction of Lucy, but this study only reconstructed the muscle paths – that is, the space and path that a muscle occupies within a body.

To simulate movement, researchers next need to estimate the architectural parameters of each muscle. Yet, there are multiple ways to do this.

The team decided to test the influence of these different methods on subsequent model performance by creating seven identical musculoskeletal models of Lucy and by only changing the input muscle parameters.

Wiseman and colleagues focused on 36 muscles in the pelvis and lower limb. These models used different ways to estimate muscle parameters, such as fibre length, pennation angle, and tendon slack length.

Two types of muscle models were compared: a simpler 'static' model and a more complex 'dynamic' model that accounted for elastic tendons and variable force-length-velocity properties of fibres.

In the simulations, vertical and side-to-side (i.e., mediolateral) ground reaction forces were gradually increased until the limb collapsed (simulation failure).

Interestingly, all Lucy models showed similar muscle activation patterns, but the maximum vertical force the limb could withstand varied amongst the models. Static muscle models struggled to support the weight, whilst dynamic models performed better, suggesting that incorporating more realistic muscle properties improves the strength of the model – although the latter does come at a modelling computational expense.

Comparing our results with a human model, we found that both species required mostly similar muscle activations with just a few exceptions to sustain single-limb support under maximal forces. This study highlights the range of outcomes possible when modelling an extinct individual, emphasizing the importance of considering different factors in estimating muscle function, especially when assessing model strength.

What does this mean for future studies?

A researcher could create two identical models of an extinct individual. By simply adjusting the input architectural parameters and deciding whether to include or exclude an elastic tendon, they can produce one version of the model that is weak.

On the other hand, they can also create a second version capable of supporting the body with multiples of body weight on a single limb. The latter model could go on to simulate activities like running and jumping, whilst the former model may be unable to simulate a basic walk.

In one scenario, Lucy would be an apt biped, capable of running and jumping. In the other, Lucy would have high metabolic costs for a simple walk, and could lead to the inference that she was not a habitual upright walker.

The researcher must then be aware of the implications when selecting the best method to estimate muscle parameters - the choice of parameter estimation method should be carefully selected alongside research requirements.