Using Wearable Sensors and Motion Parameters for Recognizing Progressive Supranuclear Palsy Phenotypes

Progressive supranuclear palsy (PSP) is a neurodegenerative disease characterized by severe motor complications. The aim of the study was to distinguish between PSP phenotypes by means of spatiotemporal, postural sway, anticipatory postural adjustment (APA) and turning parameters. Thirty-two subjects (25 PSP with Richardson's syndrome and 5 PSP with other variants) performed gait analysis while wearing wearable sensors and following a three-trial protocol: a walking test, a sway test and a turning test. Univariate statistical analysis was used to compare the two groups. Results highlighted the role of APAs in PSP phenotypes recognition. Then, a machine learning analysis was carried out by using quantitative parameters as input to Random Forest (RF), Naïve Bayes (NB) and k-Nearest Neighbor algorithms. The Synthetic Minority Oversampling TEchnique was implemented to balance the minority class. RF reached the highest accuracy (90.0 %)and Area Under the Curve Receiver Operating Characteristic (96.0 %). The best sensitivity value (96.0 %) was achieved by NB. Despite needing further investigation, predictive models show encouraging results in distinguishing PSP phenotypes.