Smartphone-based continuous mobility monitoring of Parkinsons disease patients reveals impacts of ambulatory bout length on gait features
Wei‐Yi Cheng, Florian Lipsmeier, Alf Scotland, Andrew P. Creagh, Timothy Kilchenmann, Liping Jin, Jens Schjodt‐Eriksen, Detlef Wolf, Yan-Ping Zhang-Schaerer, Ignacio Fernandez Garcia, Juliane Siebourg‐Polster, Jay Soto, Lynne M. Verselis, Meret Martin Facklam, Frank Boess, Martin Koller, Michael J Grundman, Andreas Urs Monsch, Ron Postuma, Anirvan Ghosh · 2017
Smartphone-based remote monitoring is a potential solution for providing long-term, objective assessment of gait and mobility in patients with Parkinsons disease (PD). In the Multiple Ascending Dose study of PRX002/RG7935, forty-four mild to moderate PD patients from cohorts 4 to 6 were included in a smartphone-based assessment for up to 24 weeks, while in a separate control study, thirty-five age-and gender-matched healthy individuals performed the same assessment up to 6 weeks. In total, over 30,000 hours of sensor data from subjects' daily activities were collected. A convolutional recurrent neural network was used for human activity recognition and extracted gait-related activities, followed by a mobility analysis on extracted mobility features during ambulatory bouts and turns. The analysis revealed that PD patients showed significantly lower mobility in terms of average ambulatory bout length - length of time of one continuous ambulatory segment, average per-step power, turn speed, and number of turns per ambulatory minute. In addition, bout-length stratified analysis shows the between-group difference of multiple features is associated with bout lengths. These study results support the potential use of smartphones for long-term mobility monitoring in future clinical practice, and also shed lights on previously inaccessible relationships between bout length and gait features under free-living condition.