PIGDAssess: Wearable Dual-Task Sensing for Self-Administered PIGD Assessment in Parkinson's Disease

1. Overview The PIGDAssess Dataset is the first publicly available wearable sensor dataset that provides full item-level annotations for all four subitems of the MDS-UPDRS Postural Instability and Gait Difficulty (PIGD) scale. The dataset is designed to support the development of algorithms for fully self-administered, clinician-free Parkinson’s Disease (PD) assessment in both clinical and home environments. 2. Participant Characteristics The dataset involves a diverse cohort of 35 individuals diagnosed with Parkinson’s Disease, recruited from a tertiary medical center. Demographics: 20 women, 15 men. Age: $68.2 \pm 10.5$ years (mean $\pm$ SD). Disease Duration: $5.29 \pm 4.2$ years. Clinical Distribution: The cohort spans Hoehn–Yahr (H-Y) stages 1 to 4 (mean $2.17 \pm 1.07$). Note: The dataset intentionally includes a robust proportion of Stage 1 (early-stage) and Stage 4 (severe) patients to ensure model generalizability across the entire spectrum of motor impairment. 3. Data Modalities & Equipment Data were captured using three commodity Inertial Measurement Units (IMUs) positioned as follows: Lumbar (L5): Captures trunk stability and center-of-mass dynamics. Left Foot: Captures stride-level gait metrics. Right Foot: Captures stride-level gait metrics. 4. Experimental Protocol Participants performed a standardized, brief protocol designed for self-administration. Each session includes: Motor Tasks: Sit-to-stand. Static standing. Walking tasks. Conditions: Each task was performed under two conditions to capture balance-sensitive signatures: Single-Task: Normal performance. Dual-Task: Motor task combined with a cognitive task (Serial-3 subtraction). 5. Annotations (Ground Truth) Each trial is annotated with gold-standard clinical labels provided by experienced movement-disorder clinicians: Subitem Scores: Ordinal scores (0–4) for each of the four UPDRS–PIGD items: Walking (Gait). Freezing of Gait. Postural Stability (normally requiring a physical "pull test"). Rising from Chair. Setting Labels: Data is tagged as either "Clinical Setting" or "Home Setting" to facilitate cross-environment validation. 6. Clinical Significance & Use Cases This dataset is specifically curated to address the limitations of traditional, infrequent clinic-based assessments. It enables researchers to: Benchmark ordinal regression models for PD severity estimation. Explore Domain Adaptation between clinical and at-home sensor data. Study the impact of cognitive load (Dual-Task) on postural stability and gait in a real-world context. 7. Availability In the interest of reproducible research and accelerating clinical translation, the IMU signals, processed features, and corresponding clinician-scored labels will be released to the research community.