RobustPDx: Remote Parkinson's Assessment Data with Device Metadata

Two de-identified datasets collected via a remote, web-based platform where participants completed gamified motor and cognitive tasks on personal computers. The data includes raw task-level records like mouse trajectories and key presses, plus derived performance features, designed to support research on remote digital health assessment and AI robustness. The dataset was created by Zerin Nasrin Tumpa and last updated in May 2026.

Use Cases

• Train Parkinson's disease classification models based on derived motor and cognitive performance features.
• Evaluate AI model robustness across device types based on included device type metadata.
• Analyze the impact of handedness on digital motor task performance based on dominant hand data.
• Study remote digital health assessment methodologies using raw web-based interaction data like mouse trajectories and reaction times.

Strengths

• Includes both raw session-level data and a feature-extracted dataset for direct machine learning use.
• Contains metadata for evaluating AI robustness across real-world conditions like device type and handedness.
• Data is de-identified and shared under a CC0-1.0 public domain license.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred from separate dictionary files after download.
• The dataset is small in size (614.6 KB), indicating a limited scope or sample size.

Provenance

Source

figshare

Collection Method

Collected through a remote, web-based Parkinson's disease assessment platform with gamified tasks.

Freshness

Last updated 2026-05-27 06:22:54; freshness should be verified.