Assessment of Parkinson's disease at-home using a natural interface based system

A system for the management of the automatic assessment of Parkinson's Disease (PD) at home is presented. The system is based on a non-contact and natural human computer interface made by optical RGB-Depth devices and it is suitable for motor impaired users, as are PD patients. The interface allows for both the gesture-based management of the system and the tracking of hands and body movements of the patient during the execution of standard upper and lower limb tasks as specified by the Unified Parkinson's Disease Rating Scale (UPDRS). The accurate tracking and characterization of hands and body movements allows for an automatic and objective assessment of UPDRS tasks, making feasible the monitoring of motor fluctuations at home and on daily basis, features that are important in disease management. The assessment is performed by machine learning techniques; the kinematic parameters characterizing the patient movements during task execution are input to trained classifiers to rate his/her motor performance. Results on monitoring experiments at home and on the system accuracy compared to clinical evaluations are presented and discussed.