Author(s): ALQAHTANI, S., SPEARS, I.R., JEFFRIES, O., COUSSENS, A., Institution: NEWCASTLE UNIVERSITY, Country: UNITED KINGDOM, Abstract-ID: 921

INTRODUCTION:
When horse riders become fatigued, coordination and postural control can decline, increasing fall and injury risk. Countermovement jumps (CMJs) are used to assess neuromuscular fatigue, but single outcomes (e.g., jump height) can miss fatigue-related changes in movement strategy. We developed a statistical parametric mapping (SPM)-guided machine-learning approach to classify fatigue from CMJ waveforms and identify the minimal inertial measurement unit (IMU) configuration.
METHODS:
Twenty experienced riders performed three CMJs before and after a 3-min high-intensity simulated horse-riding bout. Thirteen body-worn IMUs and dual force plates recorded CMJ waveforms. SPM was used to locate fatigue-sensitive time regions, and extracted features were entered into an L1-regularised logistic regression classifier. Performance was evaluated using leave-one-subject-out cross-validation, and optimal 1-4 IMU sets were selected using backward elimination.
RESULTS:
Blood lactate increased from 1.7 to 6.6 mmol/L, while jump height, peak force and peak power decreased by 11-18% on average. The best single IMU (left shank) achieved AUC 0.83 (accuracy 0.75). Adding sensors improved performance: left shank + right humerus AUC 0.96 (accuracy 0.88); pelvis + left shank + right humerus AUC 0.98 (accuracy 0.93); pelvis + left thigh + left shank + right humerus AUC 0.99 (accuracy 0.97).
CONCLUSION:
Waveform-based classification detected fatigue after a brief riding-specific bout using a small number of wearable sensors. A four-IMU configuration (pelvis, left thigh, left shank, right humerus) provided near-maximal performance, supporting practical field monitoring with minimal instrumentation. References: Pataky TC (2010) J Biomech 43:1976-1982; Gathercole R et al. (2015) Int J Sports Physiol Perform 10:84-92.