ECSS Paris 2023: OP-BM09
INTRODUCTION: The squat in powerlifting is executed under standardised conditions, yet elite athletes exhibit marked inter- and intra-individual variability in movement patterns, challenging group-based interpretations of squat mechanics [1]. Musculoskeletal (MSK) modelling allows for detailed estimation of muscle forces (MF) [2]; and although certain general patterns of MF adaptation with increasing intensity have been described, substantial variability in individual force production strategies persists across athletes [3]. Consequently, investigating individual MF contributions rather than group means is essential for athlete-specific technique and training optimisation. METHODS: 3D motion capture and ground reaction forces were recorded from 29 elite powerlifters (age: 26.1±5.4 years; 1-repetition-maximum (1RM): 2.4±0.4xbody mass (BM)) performing squats at 70 to 90% 1RM. Personalised MSK-models based on the “Catelli”-model [4] were used to estimate MF in OpenSim [2]. Peak forces were normalised to BM and used to calculate MF ratios between hip and knee extensor muscles. Individual MF ratios and changes in MF ratios (i.e., slopes) were derived in order to visualise inter-individual strategies with increasing intensity during the concentric phase and the sticking region. This region is defined as the interval between the initial maximal upwards velocity of the barbell and its first local minimum [5] (i.e. the most challenging part of the squat). RESULTS: The knee-to-hip extensor MF ratio decreased significantly from 70% to 90% 1RM (1.25 (IQR 1.16-1.49) to 1.18 (IQR 1.11-1.30); p<0.001). This shift was driven primarily by greater contributions of the single-joint hip extensors (p<0.001). The average change in MF ratio with increasing intensity was negative and significantly different from zero (-0.004±0.005; range: −0.012 to 0.007; p<0.001), indicating a systematic shift toward greater hip-extensor involvement at higher loads. Individual MF relationship changes showed marked heterogeneity (ICC=0.017): 44.8% of the athletes with a hip-dominant pattern at 70% 1RM were becoming even more hip-dominant as intensity increased. 41.4% of athletes shifted from knee- to hip-dominant strategies, whereas knee dominance or reductions in hip dominance were rare (6.9% each). CONCLUSION: Individual muscle force strategies varied markedly across athletes. Overall, squats became increasingly hip-extensor-dominant with rising intensity, mainly due to greater involvement of the single-joint hip extensors. Knee extensor dominant strategies were rare. These findings highlight the need for individualised technique and training adjustments. Considering individual muscle-force contribution strategies may help optimise squat performance. Literature: 1) Kristiansen et al., Hum Mov Sci, 2019 2) Delp et al., IEEE Trans Biomed Eng, 2007 3) Pürzel et al., Scand J Med Sci Sports, 2025 4) Catelli et al., Comput Methods Biomech Biomed Engin, 2019 5) Van den Tillaar et al., J Hum Kinet, 2014
Read CV Alexander PürzelECSS Paris 2023: OP-BM09
INTRODUCTION: Muscle force simulation is fundamental in sports science and biomechanics, enabling the estimation of internal muscle loads that cannot be measured directly in vivo. These simulations support training optimization, performance enhancement, injury risk reduction, and technique analysis by linking movement kinematics to neuromuscular demands. Modeling muscle fatigue is also essential for understanding performance decline, workload accumulation, and recovery. However, conventional musculoskeletal and fatigue models are computationally intensive, limiting their use in real-time applications. Previous work demonstrated the feasibility of real-time muscle-level biofeedback for load management and injury prevention [1], but was limited in range of motion and did not account for muscle wrapping or fatigue. This work aims to integrate efficient musculotendon kinematics and muscle fatigue modeling into a unified framework for future real-time biofeedback in sport. METHODS: An extended Kalman filter (EKF) framework was used for whole-body motion capture, kinematic reconstruction, and inverse dynamics computation [1]. For this preliminary study, the framework was implemented offline to assess feasibility and accuracy. Reconstructed kinematics were used as inputs to a neural network trained on 1,000 scaled musculoskeletal models posed across 500 joint coordinates, yielding 500,000 input vectors, to estimate muscle moment arms, muscle–tendon lengths, and contraction velocities [2]. This surrogate model accounts for muscle wrapping. A four-compartment muscle fatigue model was integrated to represent metabolic and non-metabolic fatigue effects and optimize muscle load-sharing [3]. The framework was evaluated using three cycles of three sets of ten squat repetitions with 90 s rest. After each cycle, two maximal voluntary contractions (MVCs) of the quadriceps were performed. Kinematics were recorded using 18 OptiTrack cameras (100 Hz), ground reaction forces using two AMTI force plates (1000 Hz), and MVC forces using a strain gauge (100 Hz). RESULTS: The unified framework was successfully implemented with high computational efficiency. Neural network–derived musculotendon kinematics required 2.8 ms per frame, while the integrated fatigue model required 4.7 ms per frame, both implemented in Python. These processing times enable real-time compatibility of the unified framework at 100 Hz when implemented in C++. Estimated muscle fatigue showed good agreement with MVC force decay (<15%). Subject-specific calibration using an isokinetic dynamometer is expected to further improve estimation accuracy. CONCLUSION: Integrating efficient musculotendon kinematics with muscle fatigue modeling advances the state of the art and enables intelligent training systems capable of delivering real-time, muscle-level feedback to optimize performance, manage fatigue, and reduce injury risk. [1] Lugrís 2024 [2] Cornish 2024 [3] Michaud 2024
Read CV Florian MichaudECSS Paris 2023: OP-BM09