Parkinson's Disease Assessment Data from 200 Processed Facial Movement Recordings

200 CSV files contain numerical facial movement data for Parkinson's disease assessment, processed from raw video footage using OpenFace 2.0. The dataset, created by Jaratsaeng, Sitthatka, was last updated on April 22, 2026. Each file contains 714 columns of frame numbers, timestamps, confidence scores, and facial landmark coordinates.

Use Cases

• Train classifiers to differentiate between healthy controls and Parkinson's patients based on numerical facial landmark coordinates.
• Analyze facial movement dynamics over five recording sequences per participant for disease progression modeling.
• Develop feature extraction pipelines for clinical video analysis using the OpenFace 2.0 output format described.

Strengths

• 200 CSV files provide a substantial collection of processed recordings.
• Data is privacy-preserving, as original videos and identifiable images have been converted to mathematical estimates.
• Clear file naming convention distinguishes between 20 healthy control ('hal') and Parkinson's disease ('Par') participants across five recording sequences.

Limitations

• Column-level documentation is absent; field semantics for all 714 columns must be inferred after download.
• Row count per file is unknown, which may limit suitability assessment for certain modeling tasks.
• The dataset's geographic and demographic scope is not specified, which may indicate inherent sampling bias.

Provenance

Source

Harvard Dataverse

Collection Method

Raw video footage was processed and feature-extracted using OpenFace 2.0.

Time Range

null

Freshness

Last updated 2026-04 22 15:37:54; freshness should be verified.

Geography

null