Predictive Simulation of Hemiparetic Gait Based on Neural Impairments

Hemiparetic gait is the most common walking impairment after a stroke that causes abnormal movement patterns. However, the neuromuscular impairments that may lead to this type of gait are not fully understood. Predictive neuromechanical simulations allow for exploring how neural and biomechanical impairments affect gait performance based on dynamic optimization method. To simulate a hemiparetic gait, we used a predictive simulation software (SCONE) with a neuromuscular model of human, walking in the sagittal plane. The model is used to implement different neural impairments such as decreased unilateral muscle excitations and spastic hyperreflexia in quadriceps and ankle plantar flexors. Neural impairments were modeled in three cases of pure neural weakness, pure spastic hyperreflexia, and a combination of both, to investigate their contribution to hemiparetic gait. For validation, all simulation results were compared with experimental gait data of hemiparetic stroke patients. Our results exhibited better agreement with experimental data when a combination of neural impairments was used to simulate hemiparetic gait. In this case, the walking speed was slower than in other cases (0.47 m/s), and spatiotemporal parameters matched well with experimental data. The cross-correlation was highest in paretic knee angle $(\mathrm{R}=0.99)$ and strong in paretic and non-paretic hip angle $(\mathrm{R}\gt 0.92)$. In terms of joint kinetics, hip and ankle moments showed strong cross-correlation in the paretic side $(\mathrm{R}\gt 0.71)$ and moderate in knee moment $(\mathrm{R}=0.66)$. Future works could utilize our predictive simulations to design and develop assistive devices and control strategies of lower-limb exoskeletons to improve gait performance in post-stroke patients.