Characterising individual running gait: A field based, scalable approach
Date of Award
Doctor of Philosophy
School of Medical and Health Sciences
Athlete physical preparation and rehabilitation in professional field-based sports are both extraordinarily complex tasks and ones in which hundreds of millions of dollars are invested across the world annually. The need for accurate and insightful information describing individuals is paramount to achieving successful outcomes in both preparation and rehabilitation.
With specific respect to running in team-based sports, the proliferation of geospatial tracking over the last two decades has produced a surplus of rudimentary displacement information which is often misinterpreted as describing how athletes run. Fundamentally displacement data e.g., distance covered in speed bands, average speed etc. lacks the detail required to describe a deep understanding of how an individual runs overground. This is specifically true when addressing the question of variability between limbs. Practitioners such as physiotherapists and strength & conditioning coaches are left only with outcome measures (distance/time) as an assessment of locomotor efficiency and effectiveness. At best, some practitioners seek to infer knowledge of running via the use of surrogate measures in activities such as vertical jump or its derivatives as an indicator of proficiency in or preparedness for running.
Historically the study of human gait in running has been restricted to the laboratory. All advances in our understanding of this area have come by way of instrumented treadmills and force plates, with the advent of multiple sequenced force plates providing unique insights in recent years. While these studies have provided invaluable foundational knowledge, the insights cannot specifically be implemented in the field due to the logistical and financial limitations of processing large groups of athletes through a laboratory setting.
Technical advances in wearable inertial motion sensors, such as accelerometers, offer a viable means by which to assess gait in an applied field setting at large scale. To date no specific research has been conducted to determine a scientifically valid methodology for utilising this technology in a highly applied and translatable manner.
The aim of this doctoral research is to describe a methodology (RUN SIGNATURE) for determining the characteristics of individual running gait via widely accessible equipment (commercial IMU’s) that can be scaled to assess large numbers of athletes. The development of this work serves to inform practitioners in the rehabilitation and performance fields about changes over time in running characteristics that may relate to output, injury, and associated rehabilitation. Further, this research seeks to scientifically substantiate RUN SIGNATURE via examination against four core elements of clinimetric evaluation; validity, reliability, discrimination, and capacity to describe change. In total the thesis is comprised of four studies and one extensive case study.
Study 1 is comprised of three independent investigations of the validity of RUN SIGNATURE. Synchronised data collected from both lumbar mounted accelerometers (housed in commercial IMU’s) and force plates in normal overground running is evaluated in the first two studies. First, criterion validity of the accuracy of determining ground contact time was established. Second, construct validity evaluation was conducted on the continuity of waveforms between the two signals to confirm the degree of agreement. In the third study assessing validity, a large historic data set was analysed across a range of speeds to determine if the accelerometer signal responds to changes in speed in a manner expected based on published laboratory-based research. The findings from these studies confirm the proposed methodologies as having both criterion and construct validity.
Reliability of the methodology was investigated in Study 2. This study was designed to evaluate both within and between day variation. An unexpected evolution within this study was the determination of reliability for speed windows. This finding radically improves the usability of the methodology as it specifies the speed range in which a practitioner can confidently assess the reported results.
One of the most unique aspects of this research was Study 3, describing the capacity to discriminate between subjects using the developed methodology. Using previously published research as prior established knowledge, the entire available data set (all collected in field-based running) was evaluated using a multiple cluster algorithmic approach (independent cluster techniques using centroid and neural networks) to determine if it was possible to “classify” the athletes against the model described by the prior established knowledge. Both clustering techniques resulted in very high level of agreement, and collectively conformed very closely to the prior established knowledge. Further, evaluation of each athlete via the models developed in the cluster analysis illustrated very clearly the degree of inter-athlete variability within the 50 strong cohort examined. These findings provide a unique framework to classify athletes based on their natural running mechanics. This framework contributes significantly to the fields of strength & conditioning and rehabilitation by providing baseline information on individual running mechanics not previously available. This insight has the potential to alter the training intervention design for many athletes toward a model more suited to their intrinsic needs as distinct from arbitrary training program design.
Data collected over 12-months of an AFL pre-season training and competition was utilised in Study 4. A linear mixed effects model approach was implemented, to examine the capacity of the proposed methodology to evaluate changes over time. Results were categoric in describing several inter-related variables and their evolution over the course of an AFL season in a manner supported by previous research. The results of this study further strengthen the practical application of the RUN SIGNATURE method.
The penultimate chapter is a thorough case study of a single athlete over a 20-month period. This athlete was profiled early in his AFL tenure, presenting with a significant injury history which was identifiable using RUN SIGNATURE. Over the ensuing 20-month period the athlete suffered two unique injuries which upon close and considered examination, using data developed using RUN SIGNATURE, can be demonstrated to be related. This finding was in complete contradiction to the medical conclusions delivered at the time. Further, a program design based on the findings of RUN SIGNATURE was implemented successfully, with the results again captured using RUN SIGNATURE and described in detail.
The general conclusion provided from the collective studies of this thesis support the use of RUN SIGNATURE as a scientifically plausible tool for the provision of unique gait characteristic data at scale in over-ground running. Further, and arguably more importantly, it has been demonstrated in detail the magnitude of variability within a generally assumed homogenous population. The level of insight provided by this work is unique as it offers the practitioner information that is simply not practically available by any other means in a team-based setting, in addition to a philosophical framework that demonstrably has an impact in changing training and rehabilitative intervention design.
Access to this thesis is not available.
Weber, J. (2022). Characterising individual running gait: A field based, scalable approach. https://ro.ecu.edu.au/theses/2533