Electromyography agreement and musculotendon parameter sensitivity analysis of predictive cycling simulations

Abstract

Abstract

Abstract Predictive musculoskeletal simulations can support rehabilitation cycling by estimating internal biomechanical variables and exploring cadence, task constraints, and neuromuscular strategies without extensive experimental testing. However, it remains unclear how well predictive cycling simulations reproduce measured muscle activation patterns and how sensitive this agreement is to assumptions about musculotendon parameters. This study quantified agreement between simulated muscle activations and experimental electromyography (EMG) during cycling and evaluated local sensitivity to key musculotendon parameters. In healthy participants, simulations captured some cadence-dependent waveform similarity and phase relationships, but EMG-simulation agreement remained limited to moderate, muscle-specific, and more consistent for tibialis anterior and knee extensors than for posterior or biarticular muscles. The two participants with incomplete spinal cord injury were analysed descriptively and showed more heterogeneous agreement. In a local one-factor-at-a-time sensitivity analysis performed in three representative participants, tendon slack length perturbations produced the largest changes in timing-based EMG and simulation agreement, whereas optimal fibre length, maximal isometric force, and pennation angle had smaller effects. These findings provide a benchmark for predictive cycling simulations and support their cautious use as hypothesis-generating tools rather than as direct surrogates for measured neuromuscular activity