Artificial Intelligence and Surface Electromyography for Assessing Quadriceps Neuromuscular Fatigue and Knee Injury Risk in Professional Basketball Players: A Pilot Study

Background: Neuromuscular fatigue reduces a muscle’s capacity to generate force and is a key determinant of performance and injury risk. In basketball, the quadriceps femoris (QF) is central to jumping, acceleration, and changes of direction, so objective detection of QF fatigue is clinically valuable. Objective: To develop and validate ma-chine-learning models for detecting QF neuromuscular fatigue in professional basket-ball players from surface electromyography (sEMG), and to relate fatigue-induced changes to knee biomechanics. Methods: In this pilot study, eight male athletesIn a real cohort, eight male athletes (n = 8; basketball; 18–40 years) performed standardized isometric and isotonic 30-repetition knee-extension fatigue protocols (isometric con-traction at 60% of maximal voluntary contraction [MVC]; isotonic 30-repetition set against a fixed submaximal load of 60% of MVC), measured on both legs (16 limbs). sEMG from vastus lateralis (VL) and vastus medialis (VM) was recorded per SENIAM guidelines at baseline, during, and post-fatigue. Time-domain (RMS), frequen-cy-domain (MDF, MNF), and non-linear features (approximate entropy, sample en-tropy, recurrence quantification analysis), plus the VL/VM ratio, were extracted; Ran-dom Forest, Support Vector Machine, and CNN–LSTM classifiers were trained and evaluated with participant-level nested cross-validationwere trained (k-fold cross-validation) and knee loading examined with finite-element (FEM) analysis. Re-sults: Fatigue reduced MDF and MNF and increased RMS in both muscles (p < 0.001), increased signal determinism, and shifted the VL/VM ratio (0.97 → 1.14, p < 0.01). The CNN–LSTM model reached the highest within-subject accuracy (0.93, AUC 0.97); however, under participant-level (leave-one-subject-out) nested cross-validation this did not generalize across athletes, with the classical models reaching only modest per-formance (accuracy ~0.69, AUC ~0.73)achieved the best classification (accuracy 0.93, AUC 0.97), exceeding classical approaches; FEM indicated increased localized carti-lage stress under fatigue. Conclusions: A multimodal framework combining sEMG, biomechanics, and artificial intelligence detected the canonical fatigue signature (ris-ing RMS with falling MNF) in 32/57 trials and, under participant-level (leave-one-subject-out) nested cross-validation the classifiers generalized only mod-estly (accuracy ~0.69, AUC ~0.73), well below the within-subject estimate; these pre-liminary findings require confirmation in a larger, adequately powered cohort.enables accurate, objective assessment of neuromuscular fatigue, with translational potential in sports medicine and injury prevention. In this pilot cohort (eight athletes, 16 limbs, 57 sEMG trials)In the real cohort (eight athletes, 16 limbs, 57 sEMG trials), RMS in-creased in 50 and MNF decreased in 37 trials, with the fatigue signature present in 32/57 (56%).