Elyse on torsional deformities in children and adolescents

My colleague Elyse Passmore will present her PhD completion seminar at 11.00am on Thursday 12 January 2017 in the Mechanical Engineering Seminar Room, Level 3, Building 170 at the University of Melbourne, Parkville. All are welcome!

Effect of lower limb torsional deformities on muscle and joint function in children and adolescents

Patients with torsional deformities of the femora and tibiae are commonly referred to our clinical pediatric gait analysis service prior to surgical correction. The goal of this dissertation is to understand the effects of torsional deformities on muscle and joint function during gait to improve surgical decision making. Musculoskeletal models were created from low-dose biplanar radiography incorporating patient-specific joint definitions and lower-limb torsion. The effects of torsional deformities were investigated in two groups of children; typically developing and those with spastic diplegia cerebral palsy.

Tagged , , , , ,

Me on knee-joint forces in osteoarthritic gait

Some things don’t get better with age. This sorta did.

I will present my PhD completion seminar (finally!) at 11.00am on Friday 15 January 2016 in the Mechanical Engineering Seminar Room, Level 3, Building 170 at the University of Melbourne, Parkville. All are welcome!

Individual contributions to tibiofemoral compartment loads in healthy and osteoarthritic gait

Increased cyclic compressive loading in the medial compartment of the knee is associated with the progression of knee osteoarthritis (OA). However as medial knee OA is accompanied by a range of neuromuscular, morphological and structural changes throughout the lower limb, a clear consensus on the etiology of the disease is yet to emerge. Of particular recent interest are the roles of non-knee-spanning muscles and the lower-limb kinematic chain in the pathomechanics of knee OA.

Joint loads during gait occur due to contributions from all muscles, gravity and inertia. As such the aim of this work was to utilise experimental gait data in conjunction with musculoskeletal modelling to undertake a novel and detailed examination of these individual constituent factors influencing loading in the osteoarthritic knee. Specifically, the objective was to decompose the time-histories of forces in the medial and lateral tibiofemoral compartments of osteoarthritic knees into contributions by individual muscles, gravity and inertia, and to explain the results in the light of the known neuro-musculoskeletal changes associated with OA.

This work would potentially be valuable in the design of non-pharmacologic interventions to mitigate medial knee OA progression, and may facilitate the development of more robust surrogate measures of knee-joint loads to improve clinical assessment of OA patients.

Tagged , , , , ,

Katie on Knee Bracing and Injury

My friend and colleague Katie Ewing will present her PhD completion seminar at 11.00am on Monday 26 October 2015 in the Mechanical Engineering Seminar Room, Level 3, Building 170 at the University of Melbourne, Parkville. All are welcome!

Understanding the biomechanics of prophylactic knee bracing in preventing knee injury during landing

Prophylactic knee braces are designed to prevent knee injuries during athletic activities, including anterior cruciate ligament (ACL) rupture, which results in painful, costly, and long-term effects. Non-contact ACL injuries commonly occur during high-risk manoeuvres, such as rapid changing of direction or landing from a jump, and have rapidly increased over the past decade. However, previous epidemiological studies have provided conflicting results on the use of prophylactic knee braces for preventing knee injuries. The overall objective of this dissertation was to better understand the biomechanics of prophylactic knee bracing during landing using: (i) experimental data recorded from athletes performing various landing tasks, (ii) rigid body musculoskeletal modeling software to simulate these landing experiments, and (iii) a validated finite element knee model for a detailed analysis of ACL loading.

This research demonstrated that recreational athletes changed their lower-extremity kinematics and kinetics when wearing a knee brace and adopted an energy absorption strategy that could help protect the knee joint and reduce the risk of ACL injury during landing. Significant changes in the magnitude of peak muscle forces were observed, suggesting that prophylactic knee bracing alters muscle function. However, simulating these changes with the finite element model of the knee joint revealed that the peak ACL force was not different in braced and unbraced knees. These findings provided further insight into the effectiveness of prophylactic knee bracing and advanced the use of computational modeling to study a clinical problem. Further research is needed so that the sports medicine community can better assess the benefits of prophylactic knee bracing.

Tagged , , , , , , ,

Under the Skin

This post is loosely based on my presentation entitled “Under the Skin: Revolutionising Musculoskeletal Health Through Computational Biomechanics” at the Universitas 21 Graduate Research Conference 2015 held at Shanghai Jiao Tong University from 9-12 June 2015. Figures are my own unless credited where appropriate.

(Photo credit: Gjon Milli)

(Photo credit: Gjon Milli)

In the late 1800s a cranky eccentric Englishman by the name of Eadweard Muybridge was making a name for himself as a photographer in on the west coast of North America. He had become acquainted with one Leland Stanford, a wealthy American industrialist who would later found that eponymous and reknowned university, who tasked him with settling a bet – Stanford was of the opinion that there occurred a moment in the gait of a trotting horse whereby all four hooves were in the air. After much difficulty, Muybridge developed a new photographic technique – a crude early form of motion capture which prompty settled the debt. There is indeed a time instance, albeit brief, during a trot whereby all four hooves were in the air. It was during this time that Muybridge turned his attention to human gait, and published well over tens of thousands of images of humans undertaking various dynamic tasks, including stair climbing, playing cricket, throwing, and ballet dancing. While initially undertaken for the artistic study of movement, he soon started photographing his subjects in front of a grid from which he could measure certain aspects of movement – kinematics – and thereby instantly metamorphosing the study of movement from an art into a quantitative applied science. While the technology has come a long way, the techniques used are not too far from Muybridge’s original methods.

(Photo credit: Eadweard Muybridge)

(Photo credit: Eadweard Muybridge)

Today we still utilise motion capture-based methods to record and quantitatively analyse gait patterns. This process is known as quantitative gait analysis (I described this very briefly in a previous post) and is typically performed in a gait laboratory. Quantitative gait analysis uses: (1) spatial data obtained from motion capture (the same kind of techniques used to animate characters in CGI film using real actors); (2) kinetic data from force plates embedded in the ground (think Wii Fit, but much more accurate and able to measure ground forces in all three dimensions) and (3) the processes of inverse kinematics and inverse dynamics, to investigate gait patterns, and in particular, musculoskeletal conditions which cause gait disorders. Used together with electromyography (EMG), in many places, this is a routine tool used to support clinical interventions, such as for children with spastic cerebral palsy.

Typical gait experiment

Typical gait experiment

Quantitative gait analysis enables us to evaluate the pattern of gait, i.e. the external observable and measurable characteristics of movement. However, it cannot readily identify what happens under the skin – particularly the action of muscles in generating the forces necessary to produce the movement. To some extent, EMG recordings provide an idea of the timing and coordination, and in some cases, also the relative activity levels of muscles. However, the surface EMG methods typically used are limited to a small set of superficial muscles (i.e. muscles close to the skin). It therefore relies on the skill and experience of the clinicians to interpret quantitative gait analysis results and infer internal musculoskeletal function from these external measures. For example, a child with cerebral palsy who is a potential candidate for muscle transfer surgery needs careful observation of movement patterns for a variety of common tasks over an extended period of time to understand and determine: (1) what muscles could be transferred to provide improvements in the functional performance of those muscles; and (2) where the origins and/or insertions of those muscles should be relocated for maximum functional benefit in a growing body. Surgical outcomes also need to be evaluated using quantitative gait analysis. However, despite the general success of this method, revision surgeries are not uncommon and, in children especially, this is of course highly undesirable.

OpenSim model

A musculoskeletal model developed in OpenSim

Can we therefore improve this process? In other words, can we peer under the skin to provide new information to clinicians to help them make more informed decisions? The answer may lie in musculoskeletal modelling, an engineering-based approach which applies the principles of engineering mechanics, dynamic control and optimisation, computer science, signal processing, systems biology, and others, to understand movement. The computational modelling of human movement has been around for decades however we are only now beginning to develop models sophisticated enough – yet user-friendly enough – to have application in quantifying clinically-significant gait disorders.

In future, models may be utilised in a variety of ways, but which mainly fit into two broad categories: (1) inverse modelling; and (2) forward simulation. These two broad categories form a rough framework for the utilisation of musculoskeletal modelling in clinical biomechanics. The inverse method is a form of reverse engineering. It extends current quantitative gait analysis techniques by inputting kinematics and kinetics, and even EMG, into robust algorithms, typically optimisation-based, to calculate muscle forces and activations. From this it is possible to further extend the analyses to determine joint reaction forces, energetics, metabolic cost and to evaluate the function performance of muscles – all of which could potentially find extensive use in the clinical evaluation of musculoskeletal disorders, joint implants, non-surgical interventions, rehabilitation, etc.

A modelling framework for clinical biomechanics

A modelling framework for clinical biomechanics

The forward method simulates movement from scratch. It is a powerful method which utilises sophisticated algorithms, often control- and optimisation-based, to determine the neural coordination of muscles required to achieve a specific goal. Usually the task being analysed is not explicitly defined. Instead some performance criteria is specified and the model left to its own devices to determine how to satisfy that criteria. For example, Anderson and Pandy specified a performance criteria of minimal metabolic cost in their ground-breaking simulation of one gait cycle of level walking, in which they assumed that walking is an optimal process. They did not explicitly tell the model to walk – instead the model found the best muscle coordination pattern to achieve the goal, which turned out to be a roughly akin to a walking gait. Forward simulation can be used to study novel movement patterns and could be used to optimise gait to achieve a clinical goal, such as minimising medial compartment force in knee osteoarthritis patients. However, forward simulations are very complex and difficult to use, and a lot more work is necessary to make them feasible for routine clinical application.

Both forward and inverse methods have other pressing limitations associated with achieving accurate patient-specific models which need to be addressed, including accuracy of model anthropometry, muscle origins and insertions, muscle moment arms, muscle paths, the excitation-activation-contraction behaviours of individual muscles, tendon and ligament properties, etc. The presence of a musculoskeletal condition can significantly alter these properties, many of which cannot currently be directly measured. Therefore in its early stages, musculoskeletal models will likely be just one tool in a clinician’s arsenal, albeit a powerful one.

Altered gait in cerebral palsy

Altered gait in cerebral palsy

Yet the future potential makes the all the current research into musculoskeletal models worthwhile. A possible application is virtual surgery in cerebral palsy. The inverse modelling method could be used to determine muscle forces, joint forces, and even internal bone forces in a child with crouch gait associated cerebral palsy. This information could be used by a clinician to perform a virtual muscle transfer on the model, and a forward simulation performed to investigate the potential outcome. Through multiple simulations testing various possible muscle transfer options, the clinician could find the surgical design which provides the best potential benefit. Post-surgery, the inverse method would again be used to evaluate the surgical outcomes. A similar pipeline could be used for designing gait modifications for patients with osteoarthritis, improving biomechanics in injury-prone athletes or even the design and performance evaluations of custom neuroprostheses.

When Eadweard Muybridge began his studies on human and animal movement, little did he know the future benefits his work would ultimately bring. Today a vibrant community of dedicated researchers continue his work to understand movement for the ultimate well-being of the wider community. And I’m sure young Eadweard wouldn’t mind that at all.

(Photo credit: Etienne-Jules Marey)

(Photo credit: Etienne-Jules Marey)

Tagged , , , , , , , , ,

Depression and Running

It is Mental Health Week in Victoria from 5 October 2014 to 11 October 2014. This is a great little short film by Joel Wolpert which gets inside ultrarunner Rob Krar‘s head, and his experiences with depression. In Krar’s words, depression and running – it goes hand in hand for so many.

I too have recently discovered ultrarunning. Although I have been fortunate so far to have never experienced depression, I find that running out on the trails has this incredible ability to clear the mind and restore the spirit. It is both wonderful and exhausting at the same time.

I believe we are responsible for our own health and well-being, which means recognising when we are in difficulty and seeking help. And, as endurance athletes, we have this one additional amazing tool to help us deal with it that most people don’t – our sport.

So what are you waiting for? Get out there and run, ride, swim or do whatever it is you do!

Tagged , , , , ,

Spring Coupling

10628378_10152791867586535_2014292172395400579_nIt’s spring, which means its finals season here in Australia and there’s a lot of footy around. Unfortunately my beloved Melbourne Storm were bundled out of the NRL finals in frankly ignominious fashion, an embarrassment of their own making about which I am still fuming, and my AFL team, the injury-ridden Collingwood, were so woeful towards the end of the regular season they were lucky to finish where they did. But with all this football about – pick your code – and all the spare hours I now have thanks to both my teams failures, it is of course only natural (and expected) for a biomechanist to use the time to ponder the dynamics of human movement.

We use our muscles to produce movement which in turn performs a particular task, like walking, running and kicking, but how do we define the specific function of a muscle in performing a particular task? One typical definition is the manner in which a muscle moves the joint or joints that it spans. For example, the soleus muscle spans the ankle joint and acts to acceleration the ankle into plantarflexion. The vasti muscles span the knee and act to accelerate the knee into extension. But as Zajac and Gordon described in a 1989 paper which is now compulsory reading for all budding computational biomechanists, determining the action of biarticular muscles based on this definition can be difficult and troublesome.

For example, take the other great ankle plantarflexor, the gastrocnemius. This is a biarticular muscle which also spans the knee. By our current definition of muscle function, we would describe the gastrocnemius as an ankle plantarflexor and a knee flexor. However, depending on the configuration of the lower limb, and if and how the foot is pushes against the ground, the gastrocnemius can perform other, seemingly counter-intuitive actions. For example, it can also act to dorsiflex the ankle while flexing the knee, or even act to extend the knee while the plantarflexing the ankle. But how? First some theory.

Muscles, or more correctly, muscle-tendon units, are linear actuators which act at a distance from one or more joint centres. When a muscle contracts to produce force, this creates torques around the joints they span. This causes the joint to accelerate. However in order for a muscle to move the rest of the body, it must somehow transmit the effects of this torque through to the centre of mass. If we consider the body as a set of linked segments, muscles which span a particular joint can accelerate all other linked joints through a mechanism called dynamic coupling. We can describe this mathematically: the expression for a muscle’s contribution to the angular acceleration of a joint it spans appears in the equations describing the accelerations of joints not spanned by that muscle. The precise effect of a muscle, uniarticular or biarticular, depends on several factors, including the type of task performed, e.g. walking or running, the configuration of the joints and the effect of any interactions with the environment, e.g. the ground, and the presence of any musculoskeletal conditions, such as cerebral palsy. A muscle’s effect can also change during the difference phases of a task.


How the soleus supports the body centre of mass

So let’s think about soleus action taking dynamic coupling into account. Think about standing with knees slightly bent and with both feet flat on the ground. Activating the soleus of course tries to accelerate the ankle into plantarflexion. But the ground prevents the foot from actually going anywhere. Instead, the shank rotates away from the foot – this is still a plantarflexion movement. However, the shank is linked to the thigh via the knee joint – shank motion sets up a type of joint force, known as an intersegmental force, which pulls the thigh along with it, straightening the leg. This in turn pushes the pelvis and torso vertically upwards. In this way the soleus supports the body’s centre of mass against gravity.

So for tasks which move the body centre of mass, perhaps a more informed way of defining muscle function would be to describe the effect of a muscle on the centre of mass. In fact a muscle’s function be defined in terms of three actions: (1) how it holds the body upright against gravity – vertical support; (2) how it powers the body forwards – progression; and (3) how it prevents sideways collapse – mediolateral balance. The big question for computational biomechanists is: can we quantify a muscle’s contribution to support, progression and balance during a specified movement? Indeed we can using musculoskeletal models – and there are 2 main techniques to do so. In one technique, muscle forces are perturbed by a small amount at every modelled time step and the resulting effect on the centre of mass is calculated.

Another technique analyses the ground reaction force. But what exactly is the ground reaction force? Here’s a question I ask my computational biomechanics class: when you stand still on scales, the force recorded is of course equivalent to one body weight. But what causes this one body weight ground reaction force in the first place? Did I hear you say gravity?… …Nope. It’s muscles. Well mostly muscles. Here’s a rough analogy: think about a box resting on scales which contains a fly. The fly is initially sitting on the bottom of the box. The scales read, say, 5 Newtons. Now the fly begins to hover steadily in mid-air in the middle of the box. How much do the scales read now?

Five Newtons of course. The action of the wings to produce upward lift on the fly must produce an equal and opposite downward thrust on the air in the box. This column of air pushes downwards on the box – so the scales will always read 5 Newtons. It’s just Newton’s Third Law.


How much do a box and a fly weigh?

A similar process occurs in the body. The human skeleton is effectively a highly unstable inverted pendulum. If we could turn off all our muscles, we would collapse in a heap. Then the scales would read 1 body weight – due entirely to gravity. But standing requires muscles, through dynamic coupling, to keep the body upright i.e. to support the body centre of mass. Recall the soleus action in standing: the plantarflexion action which straightens the leg, also pushes the foot into the ground. Hence the upwards thrust on the body centre of mass is matched by an equal and opposite push on the ground. In this way, all the muscles together produce an upwards thrust on the body centre of mass almost equal to the downwards action of gravity. It is almost equal because there is a small amount of the body weight which is supported directly by the skeleton. Nevertheless, the muscles match the upward thrust with an equivalent downward push on the ground. In this way muscles, through dynamic coupling, are responsible for most of the 1 body weight of force measured on scales during standing.

Also in this way, the ground reaction force represents (mostly) the net effect of muscles on vertical support, progression and mediolateral balance, which is just a superposition of contributions by the individual muscles. By manipulating the equations of motion of the whole musculoskeletal system, we can decompose the ground reaction force into the constituent contributions by each individual muscle. And this is how we can quantify the contribution by each muscle to centre of mass accelerations.

In human walking, a decomposition of the ground reaction forces shows that five major muscle groups accelerate the body centre of mass. The gluteus maximus and the vasti provide support and decelerate the body in early stance. But in late stance, the soleus and gastrocnemius support and accelerate the body forwards. In the frontal plane, gravity tends to accelerate the body laterally. The vasti, soleus and gastrocnemius also accelerate the body laterally. However, the gluteus medius is responsible for countering these effects by providing a large medially-directed force. Thus we can say that the gluteus medius is largely responsible for mediolateral balance.

In running, the situation is quite different. Only three major muscle groups are responsible for support, progression and balance. The vasti decelerate the body in early stance, but the soleus and gastrocnemius power running – they propel the body forward in late stance. Support is generated almost exclusively by soleus, with some help from the vasti. However, in mediolateral balance, the vasti take over from gluteus medius to provide a medially-directed force to counter a large laterally-directed force from the soleus. Note how the function of the vasti in mediolateral balance is completely different in between running and walking. The great importance of the soleus contribution in running leads to further theories about the biological limits of running speed – a fascinating and controversial topic unto itself.


Contributions by the major lower limb muscles to the ground reaction force in walking and running. (Figure courtesy of Dr Tim Dorn. Used with permission.)

The definition of a muscle’s function is of course task-specific. Hence our definition with regards to the effect of a muscle on the body’s centre of mass is only appropriate if the goal of the task being performed is to move the body’s centre of mass in the first place. In the process of kicking, from preparation, ball contact and follow-through, every muscle in the swing leg has a specific function related to the goal of propelling the ball. In this way, we can, perhaps, relate muscle function to the linear accelerations of the foot during each phase of kicking. In fact, during contact, if we know the time history of the foot-to-ball contact force, we can even use the same decomposition technique to understand how individual muscles contribute to the forces on the ball.

Anyway, that’s where I will leave this post. It’s less than a week to the AFL grand final. One of the downsides of being of Sri Lankan Tamil stock and growing up in Melbourne in the 80s is that there are far too many Hawthorn supporters amongst family and friends. And the only thing that annoys me more than a gloating Hawk is an entire contingent of them. And I think a few have just turned up at my door. Looks like I have to go dig out some red and white. Go Swannies???

Tagged , , , , , ,

Me & Dr David Gonsalvez @ The Laborastory (Updated)

Update: Thanks to all who came to The Laborastory to hear David and me talk about our science heroes! It was an awesomely fun night. If you missed it, you can listen to my talk on Eadweard Muybridge and Dave’s talk on Mondino de Liuzzi on The Laborastory website: http://thelaborastory.com/stories/


It’s been a while since I’ve posted a new article – thanks to moving to Samoa for the first half of the year to join my wife who’s been there for a couple of years now. But I’m back and will have a new post up soon.

In the meantime, my friend Dr David Gonsalvez and I will be presenting at the next edition of The Laborastory, a monthly event where scientists talk about their science heroes. Please do come down! Tickets need to be purchased on the website.

The Laborastory (http://thelaborastory.com/)
Tuesday 7.30pm, 3 June 2014
Cider House, 386-388 Brunswick St, Fitzroy, VIC

If you can’t make it, you can follow the evening’s proceedings on Twitter via @TheLaborastory, and audio clips of the night will be available on the website afterwards.

Tagged , , , ,

The Walk-Run Enigma

6a00e5538b84f388330105369f5158970c-800wiI was at my favourite Melbourne coffee haunt (yes I am an unabashed coffee snob) watching a mother and her child walk down the back alleyway which leads into one of the entrances to the cafe. The mother was walking at roughly constant speed, however, the child was alternately walking and running to try and keep up. It reminded me of a fascinating area of biomechanics research – the walk-run transition.

Walking is of course the usual human-powered method of getting from A to B. Adult humans walk at a variety of “preferred” speeds, but with the average at around the 1.3 m/s. As walking speed increases however, there approaches a limit at which the body starts to prefer to run. We’ve all experienced it. Maybe we’re at the swimming pool and we want to get back to the water slide as quick as possible for another go but the lifeguard yells at us to stop running. Maybe we’re late for the bus but we don’t want to risk crumpling our nice suit or dress. We’re walking so fast we’re not sure if we want to run. Then we start running but we’re not sure about that either so we go back to walking. This alternating walk-run behaviour is known as the walk-run transition.

The walk-run transition typically occurs around 2.0 m/s. The question we have yet to answer is why? There have been several theories, summarised quite well in a 1995 paper by Frederick Diedrich and William Warren, Jr. Some possible explanations include (1) a maximum walking speed related to the accelerations of the body’s centre of mass, suggesting that gravity imposes a limit on walking speed – but in the simplest modelling scenario this predicts a maximum speed of about 3.0 m/s, much higher than the observed transition speed; (2) the transition to running is somehow metabolically optimal at speeds around 2.0 m/s – however subsequent studies have shown that the metabolic cost of running is most cases higher than walking at those transition speeds; (3) muscle and joint forces are somehow reduced or optimised by switching to a running strategy – however, modelling studies show that muscle and joint forces are always higher in running, due to the ballistic nature of the task; and (4) an idea proposed by Diedrich and Warren themselves – that the walk-run transition is related to phase transition from unstable to stable states.

Recently, however, a new theory has emerged related to the behaviour of muscles in the leg. But before we explore this, we need to understand muscles. Muscles of course work by means of complex series of electrical and biochemical processes and interactions. In computational biomechanics however, we often do not model muscle behaviour at that biological level. Apart from the fact that this would be prohibitively complex, it’s not necessary for most scenarios we are trying to study. Instead we utilise phenomenological models. That is, we observe the physical behaviour of muscles and build mathematical models to represent that observed behaviour.

One important phenomenon is the activation-contraction behaviour.  The higher a muscle is activated, i.e. the bigger the electrochemical signal, the more force it generates. However, this is not all. Two other very important properties we observe in muscles are the force-length and force-velocity properties.

  • Force-length property: Every muscle has an optimal fibre length, at which the muscle generates maximal force. As the muscle fibres shorten during contraction, they generate less force. Similarly, if muscles are stretched beyond their optimal fibre length, they also generate less force. The resulting force-length behaviour can be represented by a bell-shaped curve (see figure).
  • Force-velocity property: When muscle fibres are not contracting, they also generate the maximum force possible at the current fibre length. However, when the muscle contracts, the speed of contraction affects the muscle’s ability to generate force. The greater the speed at which muscle fibres contract, the less force they generate. The resulting force-velocity behaviour can be represented by curve which gradually falls to zero (see figure).

Together these are known as the force-length-velocity property of muscles.

Force-LengthForce-VelocityDuring gait, an important group of muscles involved in moving the body are the ankle plantarflexors, of which the “calf muscles” – i.e. the soleus and gastrocnemius – are the most important. They activate during in the stance phase during both walking and running, and provide the push-off which propels the body forward. In 2005, Kotaro Sasaki and Rick Neptune reported in an electromyography-based modelling study that near the transition speed, the forces generated by the ankle plantarflexor muscles decreased with increasing walking speed despite the higher activation levels of those muscles. Yet while running at speeds close to the transition speed, those same muscles generated increasing forces with increasing speed and activation. They concluded that the force-length-velocity property was a determining factor in the walk-run transition.

This finding has, in the last few months, been reinforced by a new modelling study undertaken by Edith Arnold and colleagues at Stanford University, with respect to the soleus muscle. They simulated a wide variety of walking and running speeds. This study also examined an important part of the calf muscles which we have not yet considered – the Achilles tendon.

The ideas proposed by both studies is that as walking speeds approach the transition speed of about 2.0 m/s, muscle activation begins to reach its limit, yet the ankle plantarflexor muscles are contracting and then relaxing faster and faster. In this way, the force-velocity relationship comes into play – increasing fibre velocities result in reduced force production despite increased activation levels. The switch to running enables the leg to turn over at much lower speeds, enables the calf muscles to generate more force.


Image courtesy of IAAF (www.iaaf.net)

The large range of motion of the ankle during running also causes very large changes in the total length of the calf muscle and the Achilles tendon combined, i.e the “muscle-tendon unit length”. While this large length change is detrimental to force generation (via the force-length property), it enables the highly compliant Achilles tendon to store and return large amounts of energy as a passive force, like an elastic band. This lessens the mechanical demands on the ankle plantarflexor muscles. Ultimately, the result is a sudden boost in the muscle force generating capabilities of the calf muscles while running at speeds around the transition speed compared to walking.

And so I will leave you to ponder this new and highly plausible theory of the walk-run transition – but keep in mind the story isn’t over. There is still much work to be done, and possibly many other theories to explore. Until then, when faced with the dilemma of whether to walk or run, running might be the better option biomechanically, but you will probably end up with a sweaty shirt.

Tagged , , , , , ,

Murray Man 2013


Road Trip and Registration

As I loaded up the car on Friday 8 November for the long 9-hour drive north-west to the lovely Riverland region of South Australia, I was filled with some trepidation. I was heading up to Barmera for the Murray Man Triathlon, part of the Australian Long Course Triathlon Championships, and my experience with this race last year –  a 40-deg-C epic with strong desert-fired northerly winds that took me an excruciating 6h 20min to complete – made me very nervous. Nevertheless, the forecast was for low-20s and I was hoping it would stay that way. After spending most of the trip up in typically Victorian freezing cold and driving rain, crossing the border into South Australia brought a warm cloudless evening. I was staying in  Berri, about 10km east of the the race hub in the town of Barmera.

The next morning was an absolute pearler and I headed out with my mate Brett, who’d also driven up from Melbourne, for a spin on the bike course along the eastern shore of Lake Bonney. Sure enough, as soon as we left the outskirts of town, that wind hit us like a ton of bricks, although there was one difference from last year – it was a cool southerly. Not that it made any difference.

As I did last year, I lunched at the wonderful Banrock Station winery and wetland centre in Kingston-on-Murray. The wine isn’t spectacular but the food is great and they do maintain a restored section of the Murray River using a portion of the profits from wine sale, although the quiet riparian scene was unceremoniously interrupted by a bunch of very loud middle-aged women up from Adelaide.

1463104_10152035395491535_540036416_nThen it was straight back to Barmera to register and check in my bike. I don’t have a TT bike. As a poor grad student I can’t afford it for the moment. But I used my trusty SRAM 60/80 wheels – they’re quite heavy and not lightning fast, but then again neither am I. In fact I can rarely ride fast enough to get any aero advantage from them. The ceramic bearings make the ride a hell of a lot easier though. I also chose to use my ITU mini-aerobars instead of the full-length ones. I have not been able to find a good TT position on my road bike, and so using the full bars is actually a detriment. The mini-bars meant I could keep my road set-up at the expense of aero advantage. I really should see the boys at Cycle Works for a proper TT fit.

Murray Man is a small race – about 300 people max – in a small town so there’s none of the irritating overdone over-hyped commercialisation that anyone who’s done an Ironman-branded race is familiar with. Just smiling faces, lots of “g’days” and a great community vibe. There I ran into my Nunawading Triathlon clubmate  Michael, and his partner Kym, where we caught up for some friendly banter and discussed race strategy. Michael did this race last year too, and he was back leaner and fitter, and chomping at the bit to beat his previous time. Finally, I headed back to Berri, picked up some fruit at a roadside stall and had a carb-loaded dinner at the local Chinese restaurant, with a menu and decor stuck in the 80s, then packed my race gear and hit the sack to catch some z’s.

Race Day

I slept surprisingly well. Woke up on time at 5am, ate some toast, and prepped my drinks.  I have recently discovered Hammer Nutrition’s Perpetuem. It is like some kind of magic rocket fuel with carbs, a full spectrum of electrolytes and a spot of protein. It is my primary bike nutrition and I made up one 4-hr bottle – as I am a woefully slow cyclist, and I mean slooooow, 3 hours in that wind would be what I would expect to do + 1 hour in case I’m really granny-gearing it. My second bottle is water with Nuun electrolyte tabs. I don’t drink plain water as I am a very salty sweater and need to replenish electrolytes more than most people, otherwise I cramp in glorious fashion, usually beginning with my hip adductors. I sip on the Perpetuem, but drink a bottle of Nuun-ified water per hour or so, replacing the water at the bottle exchanges and adding Nuun tabs from my bento or back pocket. For the run, I have gels and Endurolytes – full-spectrum electrolyte pills also made by Hammer. I don’t use them on the bike because I prefer the reassuring flavour of salty water hence the Nuun, but Endurolytes are effective and very convenient to carry on the run.

Having loaded up the car, I began the drive – really a procession of athletes – from Berri to Barmera. It was still dark. On the drive I consumed my regular pre-race “breakfast” – vanilla-flavoured Up’n’Go, the highly-processed sugar-laden liquid breakfast made by Sanitarium. Purists will admonish me, but I like it because it has very little fibre (despite being advertised as high-fibre – fibre is the endurance athlete’s worst enemy, trust me), has plenty of quick carbs and empties relatively fast so it doesn’t slosh around on the swim. No bananas this year!


Photo by Kym Hentschel

Transition was the usual buzz. Setting up in the morning is no longer the nerve-wracking task it used to be, although I almost forgot to pump up my tyres before I left. I always have my bike shoes clipped into my pedals but no rubber bands to keep them horizontal. One day I would like to give it shot, but I think I need to improve my riding first or risk looking like a try-hard. For the first time ever, I decided to try socks for the bike and run. For the run, I had two shoe options – both of them Inov-8 – neither of which I had ever used in a triathlon. As the run course loops around town and is 95% on bitumen, I chose to use my Road-X 233s (the alternative would have been Terrafly).

As the sun rose, it was clear that it would be another wonderful cloudless day. Put on the wetsuit, chugged down a whole bottle of Nuun and headed to the water with the throngs. Will only about 300 people, it was a mass start. As I jumped into the water I ran into both Brett and Michael who were also warming up and wished them well. The water in Lake Bonney is cool and turbid but clean, unlike that awful muddy gunk they have at Shepparton. Despite the wind, the water was relatively flat. Of course, I undertook that great triathlon tradition of peeing in my wetsuit.

The horn went off and the 2km swim was underway. As always I started out the back. It was a strange swim – I was completely relaxed and cruising in my own space. I don’t think I fought a single battle out there. Weird. The new swim course, designed to eliminate sun glare, did its job. Finished the swim mid-field and fully expected to cramp in my hip adductors as often happens when I begin the run into transition but there was none – proper hydration saved me. Transition was uneventful, although I forgot to dry my feet, which made pulling on my socks somewhat difficult.

903463_10152035398481535_253278091_oThe 80km bike leg would decide my race. Four 20km laps out and back along the eastern shore of Lake Bonney with rolling gentle hills. It’s a wonderfully unique landscape though. The water on one side, and the acre upon acre of red soil and low saltbush scrub on the other reminds you that you’re on the edge of the desert. Of course the scrub means there’s no protection from that crazy prevailing wind which exists here. I began well, riding north out of town along the Lake, the big tail wind made it all so cruisy – and I actually overtook people, very rare in my world. The Perpetuem/Nuun combination worked really well. The return leg was tough with the headwind but I was mentally well-prepared for it after last year’s experience.

There is one important rule I am constantly reminded of for races of this distance but have never followed: you must pee at least once on the bike. It’s an important indicator of hydration. And to my glee, for the first time ever, that point came during lap 3. Of course, I’d already passed the toilets at the town-end of the course so had to try the ones are the far end of the course. And of course they were fully occupied. After waiting a few minutes, I gave up so back to town it was – it’s desperately hard work riding into a headwind with an exploding bladder. That was the only real adventure on the bike. The main downside of being a slow rider is that you’re out there for longer. Which means the wind gets stronger, the sun gets hotter and you expend more energy. Otherwise it was relatively enjoyable. Saw both Brett and Michael woosh by several times. They were really flying.


Ruined 19th century hotel at the turnaround point

Eighty kilometers, 4 bottles of Nuun, 1 bottle of Perpetuem later, I rolled into transition. My routine is always to leave my shoes clipped in and run in my socks. Put on the road shoes, grabbed my visor, gels and bag of Endurolytes and jogged out with a large group of athletes who’d also just come in. I felt great. A bit tired, but great It was so strange. Never felt this way. One of the race marshals recognised me from last year and called out “Prazzle Dazzle, looking fresh!” and soon I found myself well in front of the group I started with. Woot!

The 20km run was 4 laps of 5km around town, each with a short section through the scrub. Not particularly exciting, but it’s great having the locals out supporting you, and providing the occasional spray from their garden hoses. My plan was to run at an even 5min/km. Unfortunately, during the ride leg, my watch inexplicably went beserk and rebooted itself, which meant I’d lost all my race data and couldn’t reconnect to GPS without stopping my ride. So I had no idea of pace and had to approximate by time elapsed alone. In the end I was slightly slower than 5min/km for the first lap. I didn’t see Brett at all. Turns out he’d finished well before any of us. What a freak.

As I started the second lap, I heard a voice behind me say “You’re a hard man to catch!”. It was Michael on his third lap and looking comfortable. We had a chat and it turned out he was aiming for 5min/km as well. So we ran together. It was great having the company. Kym was out there too running back and forth taking photos and shouting encouragement. Once we began my third lap (Michael’s fourth and final) he was feeling great so he upped his speed and disappeared into the distance, but I managed to keep my pace consistent. In the middle section of my last lap, fatigue caught up with me and I slowed down considerably, but recovered at the end. In hindsight, it was more mental than anything and I should have kept my speed up. Oh well. Next time.

The final 100m was the best. A small but vocal crowd cheering you on as you cross the line. Another Murray Man done. And done well. I’d taken a whole hour off last year’s time. Bloody marvellous.


Although I finished where I expected – well down the back end as always – I had some significant wins this year. Which made this the best long course race I’ve done in years. And Murray Man is a wonderful race full stop, for all reasons – the course, the people, the region, etc.

For the first time ever, my hydration and nutrition went to plan. No cramps –  not one. And never felt hungry or light-headed. Peed wonderful glorious clear pee. Got off the bike and ran with fresh-ish legs. Perpetuem, Endurolytes, Nuun and gels did their jobs. Long course triathlon is as much about nutrition as it is about training, and when you get it right for the first time after years of failure, the feeling is indescribable.


Photo by Kym Hentschel

However, I do want to end my reliance on gels in the run leg – they don’t always go down well, taste awful and if anything, it’s one less thing I need to buy. Given how good Perpetuem is on the bike and after watching Michael’s consistent and uninterrupted run with his Fuel Belt, I have now procured one and am testing how the Perpetuem sits in my belly while running. Unfortunately, I can’t get away from using Endurolytes, but they’re easy to carry.

My other great win was on the bike. I am a terrible cyclist, mostly because I don’t do enough cycling. Although my split time wasn’t great compared to most of the competitors, it was a huge improvement on my previous long course results. I spent much my training period specifically working on bike fitness – not so much to go faster, but to ride easier and to get off the bike fresher. Saturdays were spent at Yarra Boulevard in Burnley with long brick sessions, Sunday tri club rides were done at tempo where possible. But what helped most was the wind trainer. Like many I used to abhor it, lasting no more than 20min, even if I had a programme. Then I invested in Troy Jacobson’s Spinervals. I got the Endurance Builder 5-Pack. And they are bloody marvellous – intense, focussed, really hard. I really look forward to wind trainer sessions now. I don’t have a TV in my garage, so I ripped the DVDs to MP3 and listen to them on my phone. It’s like having Coach Troy standing next to you yelling instructions in your ear.

The unexpected reboot of my watch is still a mystery but I’m not surprised. I use a Timex Ironman Run Trainer 2.0. I have to admit, it’s a pretty frustrating watch to use. It has all the great features and programmability of Timex Ironman watches, but the GPS connectivity is poor and very temperamental. Internet reviews report the same issues as I have experienced. If it continues, I may have to abandon Timex and move to Garmin or Suunto.


Photo by Kym Hentschel

My Road-X 233s did their job well but they are not made for triathlon. They are a more minimal shoe and I love them for road running, but after several hours on the bike, with the associated fatigue, they felt quite harsh underfoot. Inov-8 don’t make a triathlon-specific shoe, and I would love for them to do so –  their road and trail shoes are really that good. A month later I did an impromptu long-course race and used my Terraflys on that occasion. Although they have the same midsole as the Road-X 233, they are more forgiving and feel wonderful even after sitting on the bike for hours, and I think will be my shoes of choice going forward. I know I should consider brands which make triathlon-specific running shoes (Newton, etc), but I love Inov-8 so much, I can’t give them up!

And finally there’s nothing like taking a massive chunk off your previous time – a WHOLE HOUR in my case. Admittedly, it was half the temperature this year – but I’m pretty sure that wasn’t the only reason. Hard work, careful planning and attention to detail really paid off.

Plans for 2014

As I expect to join my wife who works in the South Pacific for the first quarter of the year, I won’t be doing much triathlon in this period. Instead I’ll be building up my trail-running and open-water swimming credentials. I have signed up for my first ultramarathon – the North Face 100 in the Blue Mountains in May, which I am very excited about. I’ll be pulling out the bike again once we’re settled back in Melbourne, hopefully doing one more Murray Man in November. I will work hard on my cycling and hopefully I will have a TT bike by then. If possible, I’d love to try and crack the elusive sub-90 minute half marathon, maybe the GCCC Half Mara in Geelong or RunMelbourne. After many years of long course racing, I’m finally hoping to do my first full iron-distance race in 2015.

What are your goals for 2014?

Tagged , , , , , , , ,

What Next, Static Optimisation?

It’s been some time since I have posted, due to an intense period of PhD-related work, which has had me thinking about some of the optimisation-based modelling techniques used in computational biomechanics.


Static optimisation is part of an inverse approach for calculating muscles forces in musculoskeletal models that has been around the computational biomechanics world for many decades now. It is used to distribute net joint torques across muscles which span the joint using some pre-defined criteria. However, as musculoskeletal modelling is increasingly used to study the biomechanics associated with demanding tasks like sprinting or disorders and pathologies such as ostoarthritis and cerebral palsy, its limitations are becoming clearer and more difficult to ignore.

From the point of view of musculoskeletal biomechanics, human movement arises from neural excitation signals, which causes muscles to contract in a manner akin to a step-response in muscles (called activation). Contracting muscles generate linear forces. As muscle wrap around joints, the distance from the joint at any time is the muscle’s lever arm, and so the force generated by a contracting muscle creates a torque about that joint. Torques induce joints to accelerate, which in turn accelerates the whole body, creating movement and interaction with the external environment.

Currently, there are typically 2 ways to model these muscles forces during human movement – a forward approach, and an inverse approach. In either case, the challenge is selecting the desired muscle forces at any time instant to generate the necessary joint torques. This is difficult because there are typically many muscles which span a joint. This leads to an over-determined problem, and requires mathematical techniques, such as optimisation theory, to solve.

The forward approach simulates movement in a model as it would typically in the human being, beginning from neural excitations and culminating in spatial movement of the body. For example, Frank Anderson and Marcus Pandy first simulated human walking using a forward approach called dynamic optimisation. Unfortunately, while the forward approach seems most natural, it is also computationally complex and challenging, hence rarely used outside of a research context.

GaitlabThe inverse approach aims to reverse engineer the muscle forces, activations and excitations by working backwards from a known pattern of movement and measured kinetics. When used in a clinical or research context, movement data is first measured in humans or animals in a gait laboratory using motion capture technology and highly sensitive force plates in the ground. From this spatial and kinetic data, the time history of joint kinematics (joint angles, velocities and accelerations) can be calculated. This can then be used to compute the net joint torques using the principles of rigid body dynamics. Finally, using the process of static optimisation, the net joint torques can be distributed across the muscles spanning the respective joints.

As there are more muscles than torques, this is also an over-determined problem. Torques are distributed across muscles in a manner which minimises the total activation of muscles, a simple surrogate measure of metabolic cost. However, this problem is much easier to solve than the forward approach since the step-by-step time history of joint torques is known beforehand. Thus muscle forces can be solved for quasi-statically, i.e. at each time step without any need to consider forward dynamics. In this way the process is “static”.

Unfortunately, the assumption of minimum metabolic cost as the criteria for directing the static optimisation process is questionable in anything other than level walking in healthy humans. Consider sprinting. The point of sprinting is to get from A to B as fast as possible, so it’s reasonable to assume that the recruitment of muscles is governed by anything other than metabolic optimality. Consider also cerebral palsy (CP). In spastic CP, the most common form of the disorder, lesions in the brain and spine can cause impairment of the neural excitation pathway, resulting in spasticity, tetanic muscle contractures and spasms. This is responsible for the unique and varied gait patterns observed in people with CP. Here too the assumption of metabolic optimality can be questioned.

Furthermore, static optimisation minimising activations cannot account for antagonist muscle co-contraction that commonly occurs during the performance of certain tasks. For example, at the knee, the action of the quadriceps in extending the knee is almost always accompanied by some level of contraction by the hamstrings. This is of course suboptimal from a metabolic point of view. However co-contraction is necessary to stabilise joints, improve postural control and absorb impacts. Knee-joint osteoarthritis (OA) is a case in point. People with OA typically have very lax knees which bow outwards during gait. Hence OA patients often increase the co-contraction of the quadriceps and hamstrings compared to healthy people. This presses the tibia against the femur, stabilising the knee against the high loading rates experiences during walking, although this may come at a cost of increased pain.

StatOptHowever static optimisation, in its current form, cannot wholly account for these elevated levels of co-contraction in muscles since it can only distribute the net joint torques across the muscles. For example, a 10 Nm extension torque at the knee can be generated by a 12 Nm extension torque due to quadriceps and 2 Nm flexion torque from the hamstrings, or a 22Nm extension torque and a 12 Nm flexion torque. Static optimisation will select the the first pair since that would minimise activations.

These issues arise from the selection of a suitable cost function to govern the distribution of muscle forces in the static optimisation process. On the surface, it only remains to find better, more task- and/or patient-specific cost functions. While it is undoubtedly a topic of much academic research and debate, suitable alternative criteria have yet to be developed. While potentially acceptable in a research context, where  metabolic optimality is just a limitation of the work, this may not be so in a clinical context.

One aim of developing user-friendly musculoskeletal modelling packages such as OpenSim is to enable complex analyses to be undertaken in a clinical setting by people who are not necessarily biomechanists or engineers. For example, surgeons and physiotherapists could calculate muscle forces generated in young CP patients to determine the best form of intervention, such as muscle re-assignment surgery, and then to measure quantitatively the outcomes after the intervention has been implemented. If the information provided by modelling is incorrect, it could lead to potentially disastrous consequences for the patient.

Modelling has the potential to revolutionise the clinical assessment and treatment of musculoskeletal conditions, but much needs to be undertaken to improve its capabilities. So what can be done? Is it time for the static optimisation to be superseded by different techniques? Or can it be improved to fit the needs of its potential clinical applications?

Tagged , , , , ,