Initial application of AI visual object tracking technology to assist in evaluating high-altitude military training-related injuries

Qin Gang; Tang Lin; Wang Zhao; Qi Xiao; Li Linfeng; Liu Yong; Guo Shaolan

doi:10.3969/j.issn.1009-4237.2025.07.003

Abstract：Objective The new AI visual object tracking technology was applied to assessing functional movement screening(FMS)scores in military training at high altitude,in order to identify its early diagnostic value for military training-related injuries.Methods A total of 136 soldiers from a grass-roots army in Tibet,China were selected as the research subjects,including 113 males and 23 females aged 18-25 years,with an average age of 20.3 years.The FMS scores,assessed both manually and by AI technology,were obtained to evaluate soldier activities be-fore and 1 month after entering the plateau(altitude:3,000 m).The manual FMS scoring was carried out with one-to-one guidance from certified professionals,while the AI evaluation was achieved by building a visual object tracking and motion capture 3D video acquisition system and scoring according to the identification of key feature points.Training-related injury data were collected as final results.The receiver operating characteristic(ROC)curve was drawn and the area under the curve(AUC)was calculated to evaluate the accuracy of the two evaluation methods of FMS scores in predicting training-related injuries.Results One month after entering the plateau,15 soldiers(11.0%,95%CI:6.7%-17.4%)were diagnosed as having military training-related injuries.The FMS showed that:(1)for each soldier,the total score was statistically reduced after entering the plateau,by both manual and AI tests(both P<0.05);compared with manual evaluation,the AI test showed a lower total score before and after entering the plateau.(2)For the 7 activity items of FMS,all showed significant differences after entering the plateau(all P<0.05)except for linear lunge(P>0.05),by both manual and AI tests.(3)AI assessment was more accurate in pre-dicting training-related injuries,with the AUC(95%CI)being 0.880(0.813-0.929),much higher than the 0.733(0.650-0.805)for manual test(P<0.05).The cutoff value of the total FMS score was 12.7 and 15.5 for AI and manual tests respectively.Conclusion AI visual object tracking technology can reduce the interference of human factors and objectively evaluate motor function through quantitative evaluation indicators,which can identify the inju-ries related to military training in plateau early.