Movement disorders associated with Parkinson’s Disease include tremor, slowness of movement, lack of movement, and involuntary movement. During the clinicial assessment of Parkinson’s disease, patients typically self-report their daily clinical states, which includes the amount of time they experienced dyskinesia (i.e., involuntary twisting or writhing movements). The clinician then uses the self-reported information to adjust treatments in the form of medication or deep brain stimulation. Because the accuracy of the self-report is often very low, the treatment modification may not be optimal. The overall objective of this study is to develop computational algorithms that automatically identify periods of dyskinesia in patients of Parkinson’s disease from body-worn accelerometer data during activities of daily living (ADL). Specifically, unlike previous studies which used supervised learning algorithms (i.e., knowledge of prior events is used to “train” the algorithm to identify future events), our goal is to classify the periods of dyskinesia solely by identifying key features from the accelerometer data. Our desired long-term outcome is to provide clinicians a timeline showing the presence of dyskinesia over an extended time period without the clinician having to train the computational algorithm by examining video for each patient.

This content is only available via PDF.
You do not currently have access to this content.