Simulation of the stride-to-stride variability of patients with Parkinson’s disease and prediction of the impact of a robotic assistance

Parkinson's disease is the second widespread neurodegenerative disease. Gait disorders are one of the most critical symptom, as it impacts the mobility, and thus the autonomy of the patient. Specifically, the stride-to-stride variability increases and the long range autocorrelations exhibited in healthy gait pattern are reduced. Consequently, gait is unstable and the risk of falling increases. Robot-assisted gait training has shown interesting effects on the gait of these patients. Through the development of a gait model, we predict the potential benefits of an active pelvic orthosis, not yet tested on Parkinson's disease patients. The model, based on optimal control, brings a novel insight on the origins of the long range autocorrelations in healthy gait. These would be due to a poor control on the stride duration and on the stride length, with still a tight control on the velocity. Furthermore, the model adapted to Parkinson's disease suggests that the decline in persistence observed in this population is due to a compensatory strategy set to reduce the variability. Finally, the orthosis is embedded in the model as a low-pass filter on the stride signals. With a suitable level of assistance, the results of the simulations indicate that patients recover a healthy gait pattern when using the orthosis. The results also highlight the importance of the adaptation of the patient to the orthosis. In conclusion, the predicted results give encouraging perspectives for the rehabilitation of patients with Parkinson's disease.