Abstract
Background: Wearable sensors are mainly used in Parkinson's disease (PD) to assess motor symptoms and to aid clinicians in patient management. Inertial measurement units that simultaneously register accelerometric and gyroscope signals have been one of the most studied and practicable methods. The heterogeneity of described methods and clinical settings studied can discourage wearable device use and highlight the need for standardization. This study compares previously proposed accelerometry and gyroscope signal features for tremor assessment measured at the wrist. Methods: An inertial measurement unit registered accelerometry and gyroscope signals at the wrist from 18 PD patients treated with deep brain stimulation (DBS). Measurements were made in DBS on and off states. Signal features for both accelerometry and gyroscope were calculated-mean linear acceleration, mean angular velocity, root mean square, maximal amplitude and power of the 3-7 Hz frequency band. The outcome features were log-transformed and correlated to the Movement Disorder Society Unified Parkinson's Disease Rating Scale (MDS-UPDRS) item 3.17 using linear regression. Intraclass correlation coefficient (ICC) values were calculated for the signal features. Results: A total of 108 tremor episodes were investigated. All signal features exhibited a strong correlation with the MDS-UPDRS tremor amplitude scale. Tremor ratings showed a stronger correlation with accelerometry (r = 0.964-0.970) than with gyroscope-derived features (r = 0.942-0.956). The best-performing feature was the mean linear acceleration (r = 0.970, R(2) = 0.940), which also showed high reliability (ICC = 0.921). Conclusions: Different accelerometry and gyroscope signal features are viable in characterizing rest tremor at the wrist. Simpler accelerometry signal features can be preferred in conducting the MDS-UPDRS item 3.17 examination in PD patients with DBS using a wrist-worn inertial measurement unit. Future research to expand the validity and usefulness of wearable technologies in PD is warranted.