Smart Watch Sensors for Tremor Assessment in Parkinson’s Disease—Algorithm Development and Measurement Properties Analysis

Highlights What are the main findings? A smartwatch-based algorithm was developed to assess upper limb tremor in Parkinson's Disease (PD) using spectral and spatiotemporal features. The algorithm showed moderate to strong agreement with a commercial IMU and was capable of distinguishing PD patients from healthy individuals. What are the implications of the main findings? Smartwatches can be used as low-cost and accessible tools for remote clinical assessment of PD motor symptoms. This wearable approach may support the transition of tremor evaluation from controlled lab environments to free-living settings.Highlights What are the main findings? A smartwatch-based algorithm was developed to assess upper limb tremor in Parkinson's Disease (PD) using spectral and spatiotemporal features. The algorithm showed moderate to strong agreement with a commercial IMU and was capable of distinguishing PD patients from healthy individuals. What are the implications of the main findings? Smartwatches can be used as low-cost and accessible tools for remote clinical assessment of PD motor symptoms. This wearable approach may support the transition of tremor evaluation from controlled lab environments to free-living settings.Abstract Parkinson's disease (PD) is a neurodegenerative disorder commonly marked by upper limb tremors that interfere with daily activities. Wearable devices, such as smartwatches, represent a promising solution for continuous and objective monitoring in PD. This study aimed to develop and validate a tremor-detection algorithm using smartwatch sensors. Data were collected from 21 individuals with PD and 27 healthy controls using both a commercial inertial measurement unit (G-Sensor, BTS Bioengineering, Italy) and a smartwatch (Apple Watch Series 3). Participants performed standardized arm movements while sensor signals were synchronized and processed to extract relevant features. Statistical analyses assessed discriminant and concurrent validity, reliability, and accuracy. The algorithm demonstrated moderate to strong correlations between smartwatch and commercial IMU data, effectively distinguishing individuals with PD from healthy controls showing associations with clinical measures, such as the MDS-UPDRS III. Reliability analysis demonstrated agreement between repeated measurements, although a proportional bias was noted. Power spectral density (PSD) analysis of accelerometer and gyroscope data along the x-axis successfully detected the presence of tremors. These findings support the use of smartwatches as a tool for detecting tremors in PD. However, further studies involving larger and more clinically impaired samples are needed to confirm the robustness and generalizability of these results.