Gait assessment in Parkinson's disease using waist-mounted smartphone

Parkinson's disease (PD) is the second neurodegenerative disorder, expressed by motor and non-motor symptoms. One of the primary motor symptoms is gait impairment, that occurs with episodic (freezing of gait and festination) and continuous disturbances, characterized by reduction of step length and gait velocity and increase in stride-to-stride variability, that with postural instability intensifies fall risk. Currently, to assess motor condition and gait impairment progress, subjective and qualitative clinical evaluations are used, which may have negative effects on diagnosis, follow-up and treatment. Therefore, gait analysis systems (laboratory-based or laboratory-free) have been developed, providing objective and additional information to neurologists. The aim of this study was to carry out the assessment of PD walking, by the construction of a systems capable to detect steps, identify initial (heel-strike) and final (toe-off) contacts of gait cycle and discriminate them between left and right, in order to prove the feasibility of a simple and low cost system for home monitoring of gait impairment in PD subjects. Data acquisition was executed with a waist-mounted smartphone, which includes several inertial sensors. A total number of 75 participants took part in the study, divided into three gropus, consisting of neurological healthy people and two composed of PD subjects, respectively. Data processing of acceleration signals was executed offline: wavelet transform has been applied to vertical and anteroposterior components, and autocorrelation function of vertical acceleration was computed. Spatio-temporal parameters and symmetry indices have been extracted and then correlated with subject age, disease duration, UPDRS items and H&Y scores. Significance tests between different groups have been also performed. Results were promising, with 95.3% of total steps detected and over 85% of the parameters values in physiological ranges. The high correlation coefficients and parameters trends are well in line with results found in literature. The alghorithm showed good sensitivity to gait variability, expressed by step and stride variability and symmetry indices. Given the promising results, togheter with ADL-like data aquisition, the proposed alghorithm could be used for remote monitoring of PD patients clinical (e.g. disease progression) and therapeutic (e.g. on/off state) condition, also in free-living environments.