Gait biomechanics in the era of data science

Abstract

Data science has transformed fields such as computer vision and economics. The ability of modern data science methods to extract insights from large, complex, heterogeneous, and noisy datasets is beginning to provide a powerful complement to the traditional approaches of experimental motion capture and biomechanical modeling. The purpose of this article is to provide a perspective on how data science methods can be incorporated into our field to advance our understanding of gait biomechanics and improve treatment planning procedures. We provide examples of how data science approaches have been applied to biomechanical data. We then discuss the challenges that remain for effectively using data science approaches in clinical gait analysis and gait biomechanics research, including the need for new tools, better infrastructure and incentives for sharing data, and education across the disciplines of biomechanics and data science. By addressing these challenges, we can revolutionize treatment planning and biomechanics research by capitalizing on the wealth of knowledge gained by gait researchers over the past decades and the vast, but often siloed, data that are collected in clinical and research laboratories around the world.

A hierarchical cluster analysis to determine whether injured runners exhibit similar kinematic gait patterns

“…A sample of 291 injured, and healthy runners were queried from an existing database of running kinematics . Participants were recruited through standard advertisements (eg, posters, facebook posts) approved by the Ethics Board.…”

“…A sample of 291 injured, and healthy runners were queried from an existing database of running kinematics. 19,20 Participants were recruited through standard advertisements (eg, posters, facebook posts) approved by the Ethics Board. In this study, runners were identified for inclusion provided they: (a) had 3D gait analysis performed by the same experienced clinician using the same motion capture system, (b) were known to be injury free for the 6 months preceding data collection (25 runners) or were suffering some form of lower body injury at the time of data collection (266 runners) and (c) contained complete kinematic data with no missing values.…”

Meta-Analysis of Integrated Therapeutic Methods in Noninvasive Lower Back Pain Therapy (LBP): The Role of Interdisciplinary Functional Diagnostics

“…Given the relationship between functional disorders and nonspecific back pain, the authors believe in the importance of diagnostic procedures to evaluate the quality of movement and in the need to conduct scientific and research works in this area. Particularly, in the era of time-space research technologies, procedures should be focused on the implementation of measurable, objective, noninvasive, and economic rehabilitation methods [145][146][147].…”

The use of real-time feedback to improve kinematic marker placement consistency among novice examiners

“…While comprising the database with healthy adults is ideal in the case of the current study involving a single healthy runner free of lower limb pathology, care must be taken when applying the placement feedback to individuals with surgical changes, or developmental disorders affecting posture or anatomical alignment. However, one of the strengths of the feedback tool is that the reference database can be easily adjusted and expanded to include more individuals that possess similar morphological characteristics to the target individual (Ferber, Osis, Hicks, & Delp, 2016;Osis et al, 2016). In a similar fashion, the study's outcomes must be interpreted within the context of the JCS used, which may vary between research groups.…”