Smartwatch-Based Screening to Identify Candidates for Deep Brain Stimulation in Parkinson's Disease

Identifying deep brain stimulation (DBS) candidates, particularly those without access to an advanced specialty center, presents ongoing challenges. This study evaluates the feasibility of a smartwatch system for identifying DBS candidates in Parkinson's disease (PD).

We recruited adults diagnosed with PD and motor complications. Participants wore a consumer smartwatch for at least 4 days per month for 8 months. The smartwatch continuously recorded motion data from its internal motion sensors. Previously validated algorithms used motion data to measure tremor, slowness, and dyskinesia. We compared various metrics in participants who were and were not recommended for DBS and developed an artificial intelligence (AI) model to predict DBS candidacy using features extracted from the motion data.

Twenty-three participants were included in the data analysis. Sixteen participants were considered DBS candidates; among them, ten initiated DBS during the study, and six did not due to age or personal preference. Seven participants were not DBS candidates. Bad time (presence of tremor, slowness, and/or dyskinesia as measured by the smartwatch) occurred more often in DBS candidates (3.46 ± 2.23 vs. 1.24 ± 1.23 h/day; p < 0.001) compared to those who were not. Additionally, the system captured a significant reduction in bad time (3.46 ± 2.23 vs. 2.25 ± 2.76 h/day; p < 0.001) after receiving DBS. The AI model achieved an area under the receiver operating characteristic curve of 0.96 for identifying DBS candidates.

The results suggest that sensors in commercial smartwatches and AI can help identify DBS candidates and detect improvements resulting from the therapy. This type of remote patient monitoring could expand access to patients who might not otherwise have considered DBS.