Optimising Instrumented Mouthguard Data Analysis: Video Synchronisation Using a Cross-correlation Approach

Abstract

Purpose: Head acceleration events (HAEs) are a growing concern in contact sports, prompting two rugby governing bodies to mandate instrumented mouthguards (iMGs). This has resulted in an influx of data imposing financial and time constraints. This study presents two computational methods that leverage a dataset of video-coded match events: cross-correlation synchronisation aligns iMG data to a video recording, by providing playback timestamps for each HAE, enabling analysts to locate them in video footage; and post-synchronisation event matching identifies the coded match event (e.g. tackles and ball carries) from a video analysis dataset for each HAE, this process is important for calculating the probability of match events resulting in HAEs. Given the professional context of iMGs in rugby, utilising commercial sources of coded match event datasets may expedite iMG analysis.

Methods: Accuracy and validity of the methods were assessed via video verification during 60 rugby matches. The accuracy of cross-correlation synchronisation was determined by calculating synchronisation error, whilst the validity of post-synchronisation event matching was evaluated using diagnostic accuracy measures (e.g. positive predictive value [PPV] and sensitivity).

Results: Cross-correlation synchronisation yielded mean synchronisation errors of 0.61-0.71 s, with all matches synchronised within 3 s' error. Post-synchronisation event matching achieved PPVs of 0.90-0.95 and sensitivity of 0.99-1.00 for identifying correct match events for SAEs.

Conclusion: Both methods achieved high accuracy and validity with the data sources used in this study. Implementation depends on the availability of a dataset of video-coded match events; however, integrating commercially available video-coded datasets offers the potential to expedite iMG analysis, improve feedback timeliness, and augment research analysis.

Keywords: Brain injury; Head acceleration; Instrumented mouthguards; Video analysis.

Fig. 1

Manual video analysis approaches for two tasks involved in the analysis of iMG data with computational alternatives cross-correlation synchronisation and post-synchronisation event matching. These manual video analysis approaches have been used in the previous research [–10], whereas the accuracy and validity of the computational approaches are the subject of this study. Note that the synchronisation point is the UTC value at the start of a video recording

Fig. 2

A bespoke MATLAB GUI used in conjunction with video recordings during manual synchronisation and labelling processes

Fig. 3

The predicted synchronisation point using the cross-correlation synchronisation method. Coded match events were extracted from coded match events. SAE—Sensor acceleration event, UTC—Universal Coordinated Time

Fig. 4

Synchronisation error for each match following cross-correlation synchronisation in rugby union (blue) and rugby league (yellow).

Fig. 5

Confusion matrices with a heatmap showing the proportion of SAEs classified by the triggering event using the post-synchronisation event matching method (rows) and manual video analysis (columns) for rugby union (a) and rugby league (b). Classifications used for validity are shown by overlaid circles and squares. Triggering event definitions are provided in Table 1 and classification definitions are shown in Table 2. The incorrect event column included SAEs that had the same match event type (i.e. coded tackles, carries, and rucks), but were linked to a different video-coded event by each method.