An on-field evaluation of race walking gait variability during a Conconi test through a video analysis with a new markerless pose estimation system: a pilot study

Purpose: The literature presents numerous accurate laboratory analysis, but the “everyday field” conditions have not been sufficiently assess. Indeed there are no studies carried out on approved tracks with a sufficient number of steps. In addition, the analysis were made through the use of expensive instruments, equipments that are very often not transportable outside the laboratories or through the use of wearable tools that could affect the gait kinematics. Therefore, it is essential to find an economical and applicable method on-field environment that allows an analysis that is as close as possible to the reality of training and competition. The aim of this study is to assess the race walking gait variability in athletes with different ages, through a video analysis with a markerless pose estimation based on transfer learning with deep neural networks named DeepLabCut. Methods: 4 athletes (2 males) of different categories perform a Conconi test in an official athletics field. The last 20m on the 400m track were recorded by two Go Pro Hero 5 video cameras. Then, the videos have been analyzed by the software of DeepLabCut. We evaluated the parameter of length, speed and time of step. Moreover, we calculated contact time to investigate eventually suspension phases. Results: During the progression of the test, as the speed increased, length and speed of step augmented. Furthermore, body parts tracked with DeepLabCut has achieved excellent results, by reaching human accuracy. Conclusions: In agreement with previous literature with the increasing of speed correspond a higher length and velocity of the step. In addition, DeepLabCut turned out to be a very useful and versatile tool for the study of gait kinematics. The next step will foresee a 3D kinematic analysis, through the use of both video cameras. The comparisons of this system with a consolidate gold-standard will be necessary to understand the strengths and weaknesses this new emerging technology.