The language of the doctoral school is English; therefore English is required, German is an advantage.

You have a completed Master's degree (or equivalent) in:

• Sport science
• Human Movement Science
• Sports engineering
• Machine Learning

Further expectations are:

• Experience in programming in Python
• Experience in working with human skeletal models
• Experience in working with marker-based motion capture systems (Vicon)
• Deep interest in and enthusiasm for scientific work.
• A critical and analytical mind.
• Strong communication, data presentation and visualisation skills.
• Ability to work, both independently and collaboratively.
• An asset: experience in writing scientific publications.

Sports injuries, prevalent among both elite and amateur athletes, impose significant societal and healthcare burdens. In Austria alone, 47,000 football-related injuries occur annually, with an incidence rate of 36 injuries per 1,000 hours, costing €373 million in healthcare costs. Also, non-contact injuries, such as ACL ruptures, often lead to prolonged rehabilitation (6–13 months) and low return-to-sport rates (43% in competitive athletes), increasing dropout risks and long-term physical/social consequences. These numbers highlight the need for effective prevention programs to minimize injury occurrence, especially for non-contact injuries. Currently efficient injury prevention is hindered by insufficient field-based biomechanical tools and validated thresholds for abnormal movement patterns exacerbated by lab-field discrepancies (Di Paolo et al., 2023). Most of the lab studies on non-contact injuries (e.g., ACL/hamstring) lack ecological validity and statistical power. Various research groups tried to close this gap using, such as Zago et al.’s (2024) 3D skeletal models spanning 600 ms around injuries. However, these studies focus narrowly on incident moments, overlooking causative movement patterns. Furthermore, these approaches are labour-intensive, thereby restricting sample sizes, and depend on low-frequency broadcast footage. In addition, these approaches do not provide joint load approximations which are crucial for skeletal modelling. A potential solution to these issues lies in the use of advanced computer vision techniques, such as deep learning-based pose estimation, which captures data from specific body landmarks at a sampling rate of 50 Hz (Jiang et al., 2025). These techniques have been developed in recent years and are already utilized by organizations like FIFA during official matches (e.g., for offside detection), providing opportunities for large-scale analysis of movement patterns that precede non-contact injuries. Therefore, the aim of this project is to create musculoskeletal models of non-contact injury incidents based on large scale analysis of official football matches. It is thereby hypothesized that: Deep learning-based pose estimation will reveal consistent pre-injury biomechanical patterns (e.g., excessive lateral trunk flexion, reduced knee flexion during deceleration) in athletes who sustain non-contact injuries. These patterns will differ significantly from uninjured athletes, providing actionable targets for prevention programs.

This project seeks to redefine injury prevention by unifying lab-validated biomechanics with real-world football data. The candidate will combine deep learning and markerless motion capture to predict joint loads—validated against gold-standard lab systems—and apply these models to  pose estimation data from elite matches football. By isolating biomechanical precursors (e.g., asymmetrical loading) within the 10-second pre-injury window and comparing injury sequences to the same player’s non-injury movements, we seek to eliminate individual variability. The biomechanical analysis will further be supported by workload metrics (fatigue, exertion) to build a real-time injury risk model.

• S. Di Paolo, E. Nijmeijer, L. Bragonzoni, E. Dingshoff, A. Gokeler, A. Benjaminse, Comparing lab and field agility kinematics in young talented female football players: implications for ACL injury prevention, Eur. J. Sport Sci. 23 (5) (2023) 859–868, https://doi.org/10.1080/17461391.2022.2064771.
• Jiang, T. et al. (2025). WorldPose: A World Cup Dataset for Global 3D Human Pose Estimation. In: Leonardis, A., Ricci, E., Roth, S., Russakovsky, O., Sattler, T., Varol, G. (eds) Computer Vision – ECCV 2024. ECCV 2024. Lecture Notes in Computer Science, vol 15077. Springer, Cham. https://doi.org/10.1007/978-3-031-72655-2_20.
• Zago M, Esposito F, Stillavato S, Zaffagnini S, Frigo CA, Della Villa F. 3-Dimensional Biomechanics of Noncontact Anterior Cruciate Ligament Injuries in Male Professional Soccer Players. Am J Sports Med. 2024 Jun;52(7):1794-1803. https://doi.org/10.1177/03635465241248071. Epub 2024 May 14. PMID: 38742580.