Forum Numerica - Estelle Delouche - Integrating AI and Biomechanics for Sports Injury Prevention

Injury prevention remains a major challenge in professional sports, particularly in high-impact disciplines such as rugby, where cruciate ligament injuries significantly affect both performance and long-term athlete health. Traditional biomechanical analysis methods fall short in capturing the complexity of neuromuscular dynamics, especially in real-world, high-variability contexts.
I present an ongoing research project led by Talan R&D in collaboration with Stade Français Paris and ENSAM, which leverages AI to detect early signals of injury risk. By combining electromyographic (EMG) time-series data with markerless 3D motion capture, we develop multimodal models capable of extracting subtle neuromuscular patterns and predicting high-risk movements in athletes. Our approach integrates machine learning classifiers and deep learning architectures to identify injury precursors, while addressing challenges such as signal noise, anatomical variability, and limited dataset size.
Beyond technical innovation, this project raises fundamental questions about generalizability, data ethics, and the future of AI-assisted decision-making in sports medicine. The long-term goal is to create scalable tools not only for elite athletes but also for broader public health and rehabilitation applications across multiple sports.

Estelle Delouche received her Ph.D. in 2023 from Université de Grenoble Alpes, France, after an initial academic background in geophysics at University of Nice - Côte d’Azur, where she specialized in seismology and large-scale data processing. She developed software and machine learning methods to monitor and predict fluctuations in groundwater systems using multimodal geophysical data, including seismic noise, satellite, and meteorological data. She also contributed to the international RISE project aimed at real-time earthquake risk reduction.

Since 2024, she has held a position as a research and development engineer at Talan's R&D center, conducting applied research projects on algorithmic fairness in generative AI and the integration of EMG signal analysis with computer vision to optimize injury prevention and performance monitoring for professional athletes. In addition to her work in geophysics and applied machine learning, she has authored recent publications on AI ethics, focusing on bias detection and mitigation in generative models. Her research lies at the intersection of science, technology, and society, and she is particularly interested in the responsible development of AI systems. She is also actively involved in giving courses in artificial intelligence and data science at EPITA and within MBA programs.