Detection of Freezing of Gait in People with Parkinson{'}s Disease using Smartphones

Freezing of Gait (FOG) is one of the most troublesome motor symptoms associated with Parkinson's disease (PD), characterised by brief episodes of inability to step. It involves increased risk of falls and reduced quality of life, and correlates with motor fluctuations and progression of the disease. Hence, the knowledge of FOG event frequency, duration, daily distribution and response to drug therapy is fundamental for a reliable patient's assessment. In this study, we propose a FOG detection algorithm that takes as input inertial data from a single waistmounted smartphone, and provides information about presence and duration of FOG episodes. Data acquisition was carried on 38 PD patients and 21 elderly subjects executing a standard 6-minute walking test. More than 3.5 hours of acceleration data have been collected. A combination of Support Vector Machine and k-Nearest Neighbour classifiers has been designed. Sensitivity of 95.4%, specificity of 98.8%, precision of 92.8% and accuracy of 98.3% in the 10-fold cross validation, and a detection rate of 84% in Leave-one-Subject-Out validation were obrained. These results, along with a good time resolution in the FOG duration identification and very efficient processing times, make the algorithm a promising tool for reliable FOG assessment during activities of daily living.