Buttfield, Alec (2016) The Development and Application of a Novel Method of Analysing Within-step Accelerations Collected During Australian Rules Football Games. PhD thesis, Victoria University.

Abstract

Resolving intra-stride accelerations from training and game data routinely collected by athlete tracking devices is rarely attempted, even though these data can provide important insights into the physical condition of athletes. This thesis proposes a new method of extracting stride accelerations from athlete tracking data via a novel analysis tool, describes methods of analysing the results generated by the analysis tool and reports and the influence of instances of missed or modified training and game activity on those results. Accelerometer and GPS Data from twenty-two professional Australian Rules Footballers were examined from competitive games during an Australian Football League season. These data were processed with a novel analysis tool developed specifically for the purpose of identifying instances of high speed running in a straight line during games, extracting step waveforms in three axes from those sections and determining the variability of those waveforms via a within-section and between-section co-efficient of multiple determination (CMD) over the course of the game. The steps taken in the development of the analysis tool are described in the thesis. Numerous approaches to identifying matched sections of high speed running in a straight line were investigated, with the method resulting in the highest number of waveforms while still being mindful of theoretical considerations adopted. Similarly, numerous statistical approaches to identifying step waveform variability were investigated and the methods demonstrating the highest repeatability within the context of the number of waveforms available for analysis were adopted, and methods with a high possibility of providing limited value in an applied setting eliminated. Results exported from the analysis tool were analysed in a number of contexts. Season averages from raw CMD scores were calculated on steps taken on the left and right foot, and the magnitude of the difference between those scores within each subject was estimated through determining the 99% confidence interval for the mean raw CMD on each side and identifying where those confidence intervals for the left and right foot did not overlap. There was one subject whose 99% confidence intervals did not overlap in any analysis condition (within-section and between-section CMD across x, y and z axes), one subject where the 99% confidence intervals did not overlap in four of the six analysis conditions, ten subjects where there was an overlap in between one and three of the analysis conditions, and ten subjects where there were no analysis conditions in which there was an overlap. Raw co-efficient of multiple correlation scores were converted to z-scores within side and axis for each subject, and confidence intervals for z-scores collated by axis (combining steps from all subjects on right and left side) were determined via an empirical bootstrapping procedure. When combined with data on ii instances of missed or modified training in the week preceding or following a game, some significant results were identified. Instances of missed or modified training were divided into five categories; “load”, “groin”, “leg soft tissue”, “leg structural” and “other”. A lower within-section z-score (indicating more step waveform variability) was found when a training was modified due to “load” (p=0.02) and higher between-section z-scores (which indicates less step waveform variability) were found in the week preceding a training modification due to “leg structural” injuries encompassing injuries to a leg not encompassed by soft tissue injuries, such as an ankle ligament sprain (p=0.02). Subjects with no difference between sides in average within-section z-axis raw CMD scores or average between-section x-axis raw CMD scores were unlikely to require training modifications due to “load” (correctly predicted in 82% of cases) and “groin” (correctly predicted in 92% of cases) respectively. These procedures and results can immediately be integrated into athlete monitoring systems, though investigations into combining these procedures with more established parameters may enhance their ability to predict adverse events. In addition, results supported previous research into the association between movement variability and pathology, and further research into the mechanism behind the changes in step waveform variability utilising the procedures outlined in this study will aid in the development and testing of our theoretical hypothesis.

Search Google Scholar

Repository staff login