Study at a glance
Summary
Disclaimer: The following summary details independent academic research. Theia3D is intended for research, sports performance and educational use only. It is not cleared as a medical device and is not intended to be used for the diagnosis, treatment, mitigation or prevention of any disease, injury or medical condition.
Editor's note: Theia3D was used as part of the biomechanics laboratory data collection in this protocol. The specific software version was not specified in the published paper.
Study Overview
Design
A prospective, observational cohort study following 120 competitive adult runners over a full training season (~10 months), conducted according to SPIRIT clinical trial protocol standards.
Participants
Runners aged 18+, training at least 20 km/week, with at least one competition during the study period and no injury in the prior three months.
Baseline Assessment
Before the season, assessment includes subjective questionnaires covering demographics, injury history, training history, sports participation, and validated psychological screens for depression (CESD-R) and anxiety (STAI); a musculoskeletal and sports medicine clinical exam plus a blood panel assessing bone health markers (vitamin D, parathyroid hormone, osteocalcin, bone-specific alkaline phosphatase, and bone resorption markers); and a two-part biomechanical assessment, a reference run on an outdoor track measured by wireless IMUs, followed by a laboratory session combining a 45-minute treadmill run and overground trials.
Continuous Monitoring During the Season
Monitoring includes a training diary, weekly injury screening via the Oslo Sports Trauma Research Center Questionnaire, and ongoing IMU-based tracking of running mechanics during normal training and competition. This lets the team observe how biomechanics shift in the lead-up to an injury, not just at one point in time.
What the Study Will Investigate
The protocol is built around five planned outcomes:
- Injury incidence and severity per 1,000 exposure hours across training and competition
- Bone health and nutritional status, specifically the prevalence of vitamin D deficiency and other skeletal alterations, and their relationship to bone stress injuries
- Individual running-pattern identification, including stride length, cadence, ground contact time, tibial shock, and tibial acceleration, measured continuously via IMU throughout the season
- Other contributing risk factors, including training periodization, subjective and objective stress, gender, running experience, and surface
- Biomechanical changes around the time of injury, comparing a runner's mechanics before and after they get hurt rather than only comparing injured versus uninjured runners
All of this feeds into the DGCN model, which will use global variance-based sensitivity analysis to rank which input variables matter most to injury risk. In effect, it produces a data-driven priority list of risk factors rather than a hypothesis-driven one.
What This Means for Research Teams
The core contribution here isn't a finding yet. It's a methodological template. For researchers designing overuse-injury studies, the protocol demonstrates a few things worth borrowing regardless of sport: combining biomechanics, bone and metabolic health, and load monitoring in the same cohort rather than studying them in isolation; treating risk factors as dynamic and interacting rather than static; and using a probabilistic ML model that reports its own uncertainty, which matters a great deal when a prediction is going to inform an athlete's training decisions.
The Role of Theia3D in Movement Analysis
The laboratory biomechanics assessment is where markerless capture does the heavy lifting. Each participant's 45-minute treadmill run and 20 overground trials are recorded with 14 synchronized video cameras and processed through Theia3D, then evaluated further in Visual3D alongside synchronized plantar pressure and ground-reaction-force data. Across 120 runners tracked over a full season, with repeat sessions for anyone who gets injured, that adds up to a substantial volume of biomechanics-lab visits. Markerless capture removes the marker-placement time and inter-session marker-placement variability that would otherwise make a protocol at this scale considerably harder to run consistently.
Study Design Considerations and What's Next
As a protocol, the paper is explicit about anticipated constraints rather than results-stage limitations. Recruitment (target 120 runners) was still ongoing at the time of publication, the study is single-cohort and single-country (Germany), and the DGCN model itself is a novel application not yet validated against simpler ML approaches for this specific task. The authors note that results will help determine whether this combined biomechanical, skeletal, and loading approach to injury prediction, along with the DGCN model itself, generalizes to other sports.
Citation: Rahlf, A.L., Hoenig, T., Stürznickel, J., Cremans, K., Fohrmann, D., Sanchez-Alvarado, A., Rolvien, T., & Hollander, K. (2022). A machine learning approach to identify risk factors for running-related injuries: study protocol for a prospective longitudinal cohort trial. BMC Sports Science, Medicine and Rehabilitation, 14, 75. Trial registration: DRKS00026904 (German Clinical Trial Register), registered October 18, 2021
Link: Read the full study protocol
Designing a Longitudinal Study?
This protocol is a good illustration of what markerless capture makes possible at scale: repeated, marker-free biomechanics sessions across a full training season, integrated with load and health data most labs already collect. If you're planning a running or overuse-injury cohort study and want to build markerless motion capture into the protocol from day one, talk to our team.