Smartphone application for classification of motor impairment severity in Parkinson's disease

Advanced hardware components embedded in modern smartphones have the potential to serve as widely available medical diagnostic devices, particularly when used in conjunction with custom software and tested algorithms. The goal of the present pilot study was to develop a smartphone application that could quantify the severity of Parkinson's disease (PD) motor symptoms, and in particular, bradykinesia. We developed an iPhone application that collected kinematic data from a small cohort of PD patients during guided movement tasks and extracted quantitative features using signal processing techniques. These features were used in a classification model trained to differentiate between overall motor impairment of greater and lesser severity using standard clinical scores provided by a trained neurologist. Using a support vector machine classifier, a classification accuracy of 0.945 was achieved under 6-fold cross validation, and several features were shown to be highly discriminatory between more severe and less severe motor impairment by area under the receiver operating characteristic curve (AUC > 0.85). Accurate classification for discriminating between more severe and less severe bradykinesia was not achieved with these methods. We discuss future directions of this work and suggest that this platform is a first step toward development of a smartphone application that has the potential to provide clinicians with a method for monitoring patients between clinical appointments.