Can we trust running gait data from devices positioned at the thoracic spine?

If practitioners want to use IMUs that the athletes wear at the thoracic spine to look at metrics related to fatigue, what metrics can they have confidence in?

Although GPS units are worn at the level of the thoracic spine, it is challenging to profile fatigue from measures of distance and speed. Many factors, such as playing position and team tactics, influence a player’s distance and speed in a team sport. Additionally, athletes can maintain their distance and speed even under fatigue, likely through changes in movement strategy to produce the same outcome (e.g., speed). 

Fortunately, GPS units have inertial measurement units (IMUs) built in. These accelerometers provide an avenue to detect discrete changes in movement strategy that may precede, or manifest independently of, any potential change in distance and speed. 

Changes in movement strategy caused by fatigue can be quantified through examination of the individual accelerometer vectors that contribute to metrics like Catapult Sports’ PlayerLoad. 

For example, the contribution of the vertical vector (PlayerLoad 1D Up) to the total PlayerLoad tends to follow decreases in flight time:contraction time (FT:CT) during a countermovement jump, indicating a fatigued state. Changes in how a player accumulates their PlayerLoad also effects how coaches perceive a player’s match performance. Given the potential implications on fatigue and match performance, practitioners may find it useful to analyse the individual accelerometer vectors (e.g., PlayerLoad 1D Up) during matches or in the days post-match. 

Although PlayerLoad (and other proprietary metrics like Dynamic Stress Load from STATSports) create opportunities to monitor fatigue within running, they do not explain the mechanisms responsible for changes in movement strategy. 

One potential solution is quantifying measures related to running gait. Until recently, these were available only in labs, using systems like three dimensional motion capture or force plates; or required the athlete to wear additional IMUs across the body, e.g., foot / lower leg and lumbar spine. 

The accelerometers in GPS units can provide valid and reliable spatiotemporal gait characteristics (contact time, step frequency, etc) and estimates of peak vertical ground reaction force (vGRF_peak) and vertical stiffness (K_vert). Moreover, via acute fatiguing protocols, we can reliably detect changes within someone’s gait pattern, which has the potential to provide more granular information on running activity than distance and speed alone. 

Is there a standard protocol that lets practitioners identify these changes? Or do we not need an isolated assessment?

A 20 m section of relatively constant speed (±1 m/s, e.g. 5-6 m/s) running is enough to derive measures of running gait from thoracic mounted IMUs. We do not yet understand the impact on the derived values during running with a greater range of speed. 

Our protocol involved a 20 m build up and 20 m constant speed zone, so it is not too different from other standardised running protocols in team sport training. For example, many warm ups include a series of runs that require athletes to accelerate, hold the required speed for a certain distance (e.g., 20 m) and gradually decelerate. These could easily be used as an opportunity to assess running gait in the days after a match to assess fatigue. 

Although this approach may be beneficial for practitioners, particularly if such running sequences are already part of the training program, the major advantage of using the built-in IMUs is the potential to obtain data within matches. 

Competition provides a more chaotic environment to assess running gait compared to a controlled research setting. But more importantly, athletes in games will repeatedly run at a constant speed for about 20 m, whether that is while directly impacting the play or in an incidental circumstance (e.g., making a substitution). These provide practitioners numerous segments to analyze. 

What other metrics related to running gait can we use confidently when using thoracic mounted IMUs?

From a fatigue perspective, one metric that has particular promise for practitioners is K_vert. 

When athletes are fatigued, they may adopt a running gait pattern characterized by a lower centre of mass during midstance, increased knee flexion, and longer ground contact times. These characteristics reflect a “Groucho” running style, which has the potential to reduce running economy. 

K_vert reflects the body’s efficiency at absorbing and utilising stored elastic energy, so it would be advantageous to practitioners to quantify K_vert (and its underlying components) in a field environment. Estimates of K_vert from accelerometers in GPS units are valid and reliable compared to other placement sites for IMUs, such as the lumbar spine. Moving the IMU further up the kinetic chain does not negatively impact validity or reliability. 

Thoracic mounted IMUs also are sensitive to fatigue related changes in K_vert during running at high velocity (7-8 m/s). The changes in K_vert during high speed running could be due to the requirement to run fast placing constraints on the athletes’ ability to manipulate components of K_vert. 

By having a way to monitor fatigue within normal training or match settings, practitioners may not need additional assessment tasks that do not provide insight into changes in running itself, such as the CMJ.

Are there any areas of caution practitioners should be aware of when using thoracic mounted IMUs to detect changes in running gait?

The major word of caution is that the approaches I’ve researched and discussed here are not currently built in to the athlete monitoring systems and other commercially available products. Practitioners would need a relatively high level of expertise and dedicate the necessary time, likely collaborating across a multidisciplinary team, to derive these measures. 

Additional areas of caution include needing to consider the size of the vest (or playing jumper) to house GPS devices. The inputs to running gait measures derive from patterns in the accelerometer signal (e.g., vertical, anteroposterior and mediolateral axes). It is imperative to minimize any extraneous perturbations acting on the accelerometer sensor due to loosely fitted clothing, as these may impact the accelerometer waveform and the derived outcome measures. 

This applies more to those metrics that require the identification of toe off (i.e., the moment the foot leaves the ground), as these events are not as pronounced in the accelerometer waveform as initial contact (i.e., the moment the foot strikes the ground). 

Secondly, practitioners should reserve the analysis of gait characteristics to straight line running. Although not all running in team sport is linear, and there are often accelerations, decelerations and changes of direction, until these movements are investigated more thoroughly from thoracic mounted IMUs, practitioners should rely on straight line runs at a relatively constant speed. 

Last, our work assessed validity and reliability up to an average velocity of 7-8 m/s, which may not necessarily be the maximum capacity of many highly trained athletes. Accelerometer data from athletes running 7-8 m/s showed a high level of precision and recall of the thoracic spine site to detect gait events (i.e., foot strikes), which is promising for assessing an athlete’s gait pattern at higher speeds.

How might practitioners use this data to inform training strategies when restoring normal running gait post-injury?

One avenue that may be attractive to rehabilitation personnel is assessing lower limb loading during early phases of return to running. Following lower limb injury, athletes may alter their gait pattern, such as longer contact times or shorter step lengths. The ability to not only quantify the metric itself but also determine the meaningfulness of the change can be advantageous to rehab personnel when implementing interventions to restore or improve an athletes running gait. 

For example, practitioners could analyse values of vGRF_peak as a surrogate measure of lower limb load, which could be useful for informing an athlete’s progression through rehabilitation. Pre-injury values could be a baseline for post-injury assessments, along with other markers like high speed running exposure and strength assessments. 

Data on running gait provides insight into “how” an athlete runs, not just what they did in terms of distance and speed. By establishing reliability through statistics like the coefficient of variation (CV%), practitioners no longer need to guess whether there has been a meaningful change in a particular gait metric (e.g., contact time). Instead, they can assess the change relative to the noise (CV%), increasing the confidence with which they assess changes in an athlete’s gait pattern. 

How do these changes in running gait due to fatigue relate to CMJ derived metrics?

Our initial study assessed changes in running gait characteristics derived from thoracic mounted IMUs in conjunction with CMJ derived markers. Those latter metrics included jump height and flight time:contraction time (FT:CT), which is sensitive to fatigue. 

Jump height is not always impaired in the presence of fatigue, likely due to alterations in jump strategy to maintain height. But we showed a significant moderate decrease in jump height coincident with the significant reduction in K_vert post intervention. This result is similar to the effects on jump height of using sprint exercise as a fatiguing protocol, and likely results from an impaired ability to use the stretch-shortening cycle (SSC) in a fatigued state. Since running also has an SSC element and K_vert is an important neuromuscular component that regulates elastic energy utilisation, it is perhaps not too surprising that fatigue had this effect. 

FT:CT declined in about 69% of the participants, which is in line with other fatigue work in team sport. Although this result was not statistically significant, it suggests a change in movement strategy that coincides with changes in running strategy as measured by reductions in K_vert. 

None of this is to say that the CMJ is redundant. But until now, assessing fatigue was only possible post-activity. The ability to quantify fatigue from IMUs attached to the thoracic spine gives practitioners the potential to conduct these evaluations during performance, which may allow for more expedited interventions aimed at recovery strategies or training load modifications.