Evaluation of Parkinson's Disease at Home: Predicting Tremor from Wearable Sensors.

Abstract

The continuous monitoring of Parkinsons's disease (PD) symptoms would allow to automatically adjust medication or deep brain stimulation parameters to a patient's momentary condition. Wearable sensors have been proposed to monitor PD symptoms and have been validated in a number of lab and hospital settings. However, taking these sensors into the daily life of patients introduces a number of difficulties, most notably the absence of an observable ground truth of what the user is currently doing. In this pilot study, we investigate PD symptoms by combining wearable sensors on both wrist and the chest with a questionnaire based evaluation of PD symptoms, in the form of experience sampling method. For a tremor dominant patient, we show that experienced tremor severity can be predicted from the sensor data with correlations of up to r = 0.43. We evaluated different window lengths to calculate the features in and see better results for longer window lengths. Our results show that continuous monitoring of PD symptoms in daily life is feasible using wearable sensors.