Measuring the effect of rhythmic auditory stimuli on parkinsonian gait in challenging settings

Rhythmic auditory stimuli (RAS) improve the disabling motor symptom of Parkinson's disease patients. In the large majority of studies, the effect of RAS has been assessed during common activities such as walking and turning. However, how RAS modulates parkinsonian behaviors in more challenging settings of daily living and whether a machine learning algorithm could classify them remains unclear. Eleven patients with idiopathic PD (age 72±7 years) were asked to walk under four conditions: straight walking, walking over an irregular surface, walking within a narrow pathway, and walking along a curving path (eight-shaped), with and without external stimulation. RAS pace was set at 110% of the normal cadence and spatio-temporal gait parameters were measured through two inertial measurement units placed on feet. k-Nearest Neighbor (k-NN) algorithm, with and without principal component analysis (PCA) as feature selector, was used for the classification of walking conditions. Cadence, gait speed, and gait time improved during RAS walking, regardless of challenging walking conditions. On the contrary, stride length increased only in straight walking, while gait speed showed improvement also in walking over an irregular surface and walking within narrow pathway conditions. k-NN algorithm reported higher accuracy (72.4%) in the classification of eight- shaped curving path both considering the overall feature set and a reduced one. These results open to the possibility of measuring RAS-induced effects on PD mobility in an ecological scenario and improving their administration based on the actual motor activity.