On the recognition of tremor severity in Parkinson's disease by means of inertial measurements-based ML algorithm

Assessing Parkinson's disease progression represents a crucial point to ensure suitable drug therapies, improve people's life quality and promptly intervene in case of sudden worsening of their health conditions. This is particularly useful when the progression status assessment can be achieved in real-time and quantitative analyses can be made available both to people affected by such a disease and the medical doctor. The assessment based on objective measurements requires accurate acquisition systems, trained algorithms and minimum invasiveness. To this aim, a combination of a software platform, based on an open-source Internet of Things (IoT) framework, and an IMU device mounted on a robotic arm, emulating movements and tremors typical of Parkinson's disease symptoms, have been employed in this work. To evaluate the system robustness to different measurement data quality, emulations were repeated asynchronously using two different IMU sensors, suitably placed on the end-effector of the robotic arm. The overall system allows the measurement, the processing and the classification of tremor severity and the possibility to visualize data in real-time on a cloud solution. In the emulated scenario, the overall accuracy in classifying tremor severity, into 5 levels, is higher than 97%. The developed method will be employed in the next future to perform validation and real-scenario testing.