Objective assessment of motor symptoms in Parkinson's disease using body-worn sensors

Objective: There is great need for an objective method of symptom assessment in Parkinson's disease (PD) to enable better treatment decisions and to assist evaluation of new therapies. In this work we present the use of wrist-worn accelerometers and artificial neural networks (ANN) to evaluate patients' motor symptoms during a prolonged period of unobserved home monitoring. Background: Current assessment methods; clinical rating scales and patient-completed symptom diaries, have a number of limitations. Accelerometers (sensors capable of capturing data on human movement) and analysis using ANN have shown potential as a method of motor symptom evaluation in PD1,2 1 Keijsers et al. Movement Disorders 2003.18:70-80 2 Hoff et al. Clin. Neuropharm. 2004.27(2):53-57. Methods: 34 participants with PD wore bilateral wrist-worn accelerometers for 4 hours in a research facility (phase 1) and then for 7 days in their own homes (phase 2) whilst also completing hourly symptom diaries. An ANN designed to predict a patient's motor status, was developed and trained based on accelerometer data captured during home monitoring (phase 2). Training an ANN based on home-recorded data is likely to provide more ‘real-life' data, compared to previous work which trained ANNs using data captured in a controlled, prescriptive fashion in clinical environments. ANN performance was evaluated (using a leave-one-out approach) against patient-completed home diaries during phase 2, and against clinician rating of disease state during phase 1 observations. Results: ANN-derived values of the proportion of time spent in each disease state during phase 2, showed strong, significant correlations with values derived from patient-completed symptom diaries. ANN disease state recognition during phase 1 was sub-optimal. Diary concordance was variable, with older, more cognitively impaired patients demonstrating poorer diary completion. Conclusions: *In terms of the amounts of time spent in each disease state, accelerometers and ANNs produced results comparable to those of home diaries. *The inherent subjectivity of the on-off disease state categories renders ANN training based on this system challenging. *Cognitive problems may preclude the use of patient-completed symptom diaries for some patients. Body-worn sensors have the potential to enable prolonged assessment of motor symptoms in this patient group