Smartphone Motor Tasks for iRBD, PD, and Control Differentiation

522 participants performed seven smartphone-based motor tasks to evaluate voice, balance, gait, finger tapping, reaction time, rest tremor, and postural tremor. The data was collected in clinic and at home to distinguish individuals with idiopathic REM sleep behavior disorder from controls and Parkinson's disease patients. Results showed mean sensitivity and specificity values ranging from 84.6% to 91.9% for pairwise group discrimination.

Use Cases

• Train a classifier using features from voice, postural tremor, and rest tremor tasks to discriminate between iRBD, PD, and control groups.
• Analyze the salience of the seven motor tasks (e.g., gait, finger tapping, reaction time) in distinguishing prodromal PD from established PD.
• Build a model to identify which specific motor tasks are most discriminatory for the pairwise comparisons of Controls vs. iRBD and iRBD vs. PD.
• Validate digital biomarkers for early PD by correlating smartphone-recorded balance and gait data with parallel in-clinic assessments.

Strengths

• Includes data from 522 participants across three distinct clinical groups: 334 PD, 104 iRBD, and 84 controls.
• Data collection occurred in both controlled clinic and non-controlled home environments over several days, capturing real-world variability.
• The study achieved high discriminatory performance, with reported sensitivity and specificity values between 84.6% and 91.9% for group comparisons.

Limitations

• The dataset size of 522 participants may be limited for training complex, generalizable machine learning models without careful validation.
• Data was collected using a customized smartphone application, which may introduce device-specific sensor variances not accounted for.
• The temporal coverage and specific time range of data collection are not provided, limiting longitudinal analysis.

Provenance

Source

Dryad, authored by Siddharth Arora.

Collection Method

Data gathered via a customized smartphone application where participants performed seven motor tasks in clinic and at home.

Freshness

Last updated in June 2019; the data may not reflect the latest smartphone sensor technologies or clinical diagnostic criteria.