Author Identifiers

Aaron Balloch
ORCID: 0000-0002-0305-822X

Date of Award

2020

Degree Type

Thesis

Degree Name

Doctor of Philosophy

School

School of Medical and Health Sciences

First Advisor

Professor Robert Newton

Second Advisor

Dr Nicolas Hart

Third Advisor

Dr Timo Rantalainen

Abstract

The ability to change direction rapidly and efficiently is critical to team-sport performance, including Australian football (AF), where a player’s capacity to rapidly decelerate, move laterally and re-accelerate is critical when evading opponents, tackling, or reacting to the unpredictable bounce of the ball or movement of another player. The biomechanical loading requirements of change of direction (COD) movement are angle and velocity dependant. Cumulative COD movement can impart high levels of neuromuscular and metabolic fatigue which can adversely affect the efficiency of subsequent movement efforts. Despite widespread use of microtechnology devices (the vast majority containing a global navigation satellite system receiver and inertial sensors) in elite level team-sport, a valid solution to automatically detect COD events and quantify the associated biomechanical load of these movements on-field remains absent. This project served to develop an algorithm that can automatically detect COD events, quantify the angle of the COD event and quantify the associated biomechanical load of each COD event. Study 1 and 2 were primarily focused on assessing the validity and reliability of the detection and angle quantification portions of the algorithm in both structured (Study 1) and unstructured (Study 2) movement environments, whilst Study 3 introduced a COD biomechanical load quantification technique to profile the COD demands of match play and a variety of match simulation training drills, provide comparisons between playing positions, and assess any similarities or differences with existing proprietary locomotive metrics. Whilst both COD event detection and angle quantification were highly accurate in a structured environment (Study 1), the accuracy of the angle quantification was severely reduced during unstructured, match-simulation training (Study 2). Utilising the event detection and biomechanical load quantification portions of the algorithm together, without angle quantification, the COD demands of match play were significantly lower than three different training drill types when expressed relative to time, whilst several positional differences were also present in COD demands across an entire season (Study 3).

Study one assessed the validity and reliability of a novel algorithm to automatically detect and calculate COD angle for pre-determined COD events ranging from 45° to 180° in both left and right directions. Five recreationally active males ran five consecutive predetermined COD trials each, at four different angles (45°, 90°, 135° and 180°) in each direction wearing a commercially available microtechnology unit (Optimeye S5, Catapult Innovations). Raw inertial sensor data were extracted, processed using our novel algorithm to calculate COD angle, and compared against a high-speed video (remotely piloted, position-locked drone aircraft) criterion measure. Concurrent validity was present for the following angles; (Left: 135°= 136.3 ± 2.1° and Right: 45°= 46.3 ± 1.6°; 135°= 133.4 ± 2.0°; 180°= 179.2 ± 5.9°) with a mild bias (< 5° or 6%) present for remaining angles; (Left: 45°= 43.8 ± 2.0°; 90°= 88.1 ± 2.0°; 180°= 181.8 ± 2.5° and Right: 90°= 91.9 ± 2.2°). All measurement of angles demonstrated good reliability (CV < 5%), whilst greater mean bias (3.6 ± 5.1°), weaker limits of agreement and reduced precision were evident for 180° trials when compared with all other angles (p < 0.001). These results confirm the high-level of accuracy and reliability of our novel algorithm to detect COD events and quantify COD angle during pre-determined COD trials and further advocates the use of inertial sensors to quantify sports-specific movement patterns.

Study two assessed the validity and reliability of both the original COD algorithm (Study 1) as well as an enhanced version to automatically detect COD events during Australian football match simulation training. The accuracy of detected COD angles was assessed from both absolute angle and discrete categorisation through multi-rater video observation as the criterion measure. Twenty-five elite, professional male Australian footballers’ completed a match simulation training drill on a modified playing area (140 m x 70 m) where video footage was recorded from multiple angles (rear and perpendicular to play) and a 3-minute portion of the drill was synchronised and chosen for the manual coding process to be performed by three different expert raters. Each rater was required to manually note the time-point that each player performed a COD event, whilst also required to document the direction (i.e. left or right), a precise COD angle (between 30° and 180°) and to subsequently categorise the COD angle into pre-determined angle zones (Zone 1: 30-60°, Zone 2: 61-90°, Zone 3: 91-120°, Zone 4: 121-150°, Zone 5: 151-180°). Sensitivity of the enhanced algorithm (95.1%) in correctly detecting COD events was greater than the original version of the algorithm (50.9%), however, the enhanced algorithm significantly underestimated mean COD angle (absolute) (p < 0.01, d = 1.07 – 1.13) and mean COD angle zone (discrete) (p < 0.01, d = 0.84 – 0.91) when compared against each of the expert raters, whom demonstrated excellent inter-rater reliability for both COD angle (ICC: 0.997, p < 0.001) and COD angle zone allocation (ICC: 0.993, p < 0.001). The COD event detection capacity of the enhanced algorithm remained high in unstructured, chaotic, match simulation training, whilst the COD angle detection accuracy was poor, likely due to the spontaneous nature of the training drill and individual biomechanical variation to pre-determined versus reactive COD movement. The accuracy of the COD event detection portion of the algorithm provides an opportunity to further integrate sensor signal outputs in a different manner to obtain and track mechanical loading requirements of on-field COD movement during team-sport activity.

Study three provided a method to quantify the mechanical load associated with each COD event which was subsequently used to profile the COD demands of 3 different training drills for comparison against match play data. Positional differences in COD demands were also assessed across the season, whilst our newly developed COD metrics were compared against existing proprietary metrics to determine their novelty. Forty-five elite Australian footballers’ provided player movement data via a commercially available microtechnology unit (Optimeye S5, Catapult Innovations) for both training and match play across the course of an entire AFL premiership season. Three different types of match simulation style training drills were compared to assess the effect that field size and player density has on COD frequency and load (and other proprietary movement metrics). Each of these drill types were also compared against match play to ascertain whether these drills meet the COD (and other movement) requirements of match play, where match play positional COD demands were also assessed across the course of a season. The relative COD demands (COD frequency and load relative to time) of each training drill type (Small-Sided Games (SSG), Mini Footy, Team Training) were greater than match play (p < 0.01), whilst the COD demands of each drill were proportional to field size and player density of each specific drill where SSG demonstrated the greatest demand for COD frequency and load, followed by Mini Footy, then Team Training. During match play, the relative COD requirements (CODAlg) were greater for inside midfielders when compared against mobile defenders (p = 0.001, d = 0.90, moderate) and tall forwards (p = 0.031, d = 1.04, moderate). Additionally, outside midfielders (p = 0.048, d = 0.54, small) and mobile forwards (p = 0.027, d = 1.07, moderate) accumulated significantly more COD events than mobile defenders. Inside midfielders recorded a significantly higher rate of COD load (CODLoadAlg) than mobile defenders (p < 0.001, d = 1.33, large), rucks (p = 0.026, d = 1.25, large) and mobile forwards (p = 0.047, d = 0.03, trivial). Both outside midfielders (p = 0.008, d = 0.91, moderate) and mobile forwards (p = 0.003, d = 1.39, large) recorded a significantly higher rate of CODLoadAlg than mobile defenders. CODAlg and CODLoadAlg were largely correlated (p < 0.01) with relative IMA-COD during each training drill as well as match play. Almost all relative physical output measures (except IMA-Decel) decreased during each subsequent period (quarter and half) of match play. The COD demands of three different types of match simulation training drills all exceeded the demands of match play, however, whilst at a lower rate, players are required to sustain these COD demands for a far greater duration (i.e. across an entire match where physical output markedly changes), when compared with training drills. Differences in COD demands present across positions may provide unique information that enables individualised, position-specific training prescription to more closely align with the evident differences during match play. These novel COD movement metrics may provide an alternative insight into the movement demands of team-sports and ultimately enhance load monitoring practice to optimise performance and reduce injury risk.

These three experimental studies as a collective provide a valid solution to detecting COD events and quantifying the associated mechanical load of COD movement during on-field team-sport activity; as well as provides a unique insight into the COD prevalence and mechanical load requirements of AF training and match play. The findings of this thesis may extend beyond elite AF and have the potential to influence future practice in various other team-sport environments.

Access Note

Chapters 3,4,5,6 and 7, and Appendices A, B, E and F are not available in this version of the thesis.

Available for download on Tuesday, May 31, 2022

Share

Paper Location

 
COinS