Phase-specific neuromechanical signatures of 40-m sprint performance: an integrated two-dimensional kinematic and surface electromyographic profiling study

Background Sprint performance reflects coordinated mechanical and neuromuscular behavior across distinct sprint phases. This study identified phase-specific neuromechanical signatures of 40-m sprint performance and tested whether integrated kinematic and electromyographic information added predictive value beyond direct spatiotemporal determinants of running velocity. Methods In this cross-sectional study, 160 young adults classified as sprint-trained athletes or physically active non-sprinters completed three maximal 40-m sprints. Split times were recorded every 10 m. Four synchronized high-speed cameras (240 frames·s -1) and wireless surface electromyography of six lower-limb muscles captured sagittal-plane kinematics, muscle excitation, coactivation, and excitation–movement coupling across four sprint phases. Velocity-adjusted mixed-effects models and one-dimensional statistical parametric mapping assessed discrete and waveform differences. Reliability-qualified features were analyzed using elastic-net models with repeated nested five-fold cross-validation. Mean 40-m velocity was the primary outcome. Step length and step frequency formed the direct spatiotemporal benchmark; ground-contact and flight times were added in an extended benchmark. Sprint-training-status classification was secondary and exploratory. Results Sprint-trained athletes completed 40 m faster than non-sprinters (5.20 ± 0.31 vs 5.95 ± 0.40 s). Their profile showed longer normalized steps, shorter contact times, phase-specific trunk and lower-limb configurations, greater early-phase gluteus maximus and vastus lateralis excitation, and greater later-phase biceps femoris preactivation and medial gastrocnemius excitation. In the complete synchronized sample ( n = 152 ), the integrated non-spatiotemporal model achieved R 2 = 0.74 and RMSE = 0.34 m·s -1 , versus R 2 = 0.60 /RMSE = 0.43 for kinematics and R 2 = 0.55 /RMSE = 0.45 for sEMG. Adding neuromechanical features to the direct benchmark improved R 2 from 0.72 to 0.86 and reduced RMSE from 0.36 to 0.25 m·s -1 (paired reduction 0.11, 95% CI 0.07–0.15). Improvement persisted for the extended benchmark ( R 2 0.76→0.87; RMSE 0.33→0.24). Transition and high-speed features accounted for 82% of coefficient magnitude, not variance explained. Exploratory classification achieved AUC = 0.94 and balanced accuracy = 0.89. Conclusion Forty-metre sprint performance was characterized by a phase-specific neuromechanical signature that provided incremental out-of-fold predictive information beyond direct and closely related spatiotemporal velocity metrics. External validation is required before broader generalization.