Luke David, Masood Zaryan, Bondi Daniel, Zhang Chaokai, Kenny Rebecca, Clansey Adam, van Donkelaar Paul, Rauscher Alexander, Ji Songbai, Wu Lyndia
Department of Mechanical Engineering, University of British Columbia, Vancouver, BC, Canada.
School of Biomedical Engineering, University of British Columbia, Vancouver, BC, Canada.
Ann Biomed Eng. 2025 May 21. doi: 10.1007/s10439-025-03747-6.
Accurate quantification of head acceleration event (HAE) exposure is critical for investigating brain injury risk in contact sports athletes. However, missing HAEs may be unavoidable in real-world data collection. This study introduces missing data imputation methods to estimate complete video- and sensor-based HAE exposure.
We captured and verified university men's ice hockey HAEs using video and instrumented mouthguards (iMGs) in one varsity season (n = 27, n = 31). A statistical mapping technique was first introduced to impute missing video-based HAEs during away games with limited camera angles. We then applied multiple imputation to impute missing iMG-based HAEs using captured data, including the complete video-based HAE exposure. This enabled estimation of complete exposure data at a per-athlete level over all games of the season.
Among 591 athlete-games, 45% did not have any recorded iMG data. We find that data imputation increased the median values of per-athlete-season video- and iMG-based HAE counts by 10% and 69%, respectively. Consequently, common head kinematics- and brain deformation-based cumulative exposure metrics also increased substantially (median per-athlete-season cumulative peak linear acceleration by 95%, peak angular acceleration by 109%, and corpus callosum strain by 69%).
This study highlights the potential underestimation of exposure metrics due to missing HAEs and fills a critical gap in sports HAE exposure research. Future studies should incorporate missing data imputation methods for more accurate estimation of HAE exposure in investigating acute and long-term brain trauma risks.
准确量化头部加速事件（HAE）暴露对于调查接触性运动运动员的脑损伤风险至关重要。然而，在实际数据收集过程中，HAE缺失可能不可避免。本研究引入缺失数据插补方法来估计基于视频和传感器的完整HAE暴露情况。
我们在一个大学男子冰球校际赛季中（n = 27，n = 31），使用视频和仪器化护齿（iMGs）捕捉并验证了HAE。首先引入一种统计映射技术，用于在客场比赛中摄像机角度有限时插补基于视频的缺失HAE。然后，我们应用多重插补，利用捕获的数据（包括基于视频的完整HAE暴露情况）来插补基于iMG的缺失HAE。这使得能够估计整个赛季所有比赛中每个运动员的完整暴露数据。
在591场运动员比赛中，45%没有任何记录的iMG数据。我们发现，数据插补分别使每个运动员赛季基于视频和iMG的HAE计数中位数增加了10%和69%。因此，基于常见头部运动学和脑变形的累积暴露指标也大幅增加（每个运动员赛季累积峰值线性加速度中位数增加95%，峰值角加速度增加109%，胼胝体应变增加69%）。
本研究强调了由于HAE缺失可能导致暴露指标被低估，并填补了运动HAE暴露研究中的一个关键空白。未来的研究应纳入缺失数据插补方法，以便在调查急性和长期脑创伤风险时更准确地估计HAE暴露情况。
