CARE-PD: A Multi-Site Anonymized Clinical Dataset for Parkinson's Disease Gait Assessment

2025-10-05

AI Summary

Parkinson’s disease (PD) lacks large-scale, multicenter, clinically annotated 3D gait datasets, hindering objective motor assessment. Method: We introduce CARE-PD—the largest publicly available PD-specific 3D mesh gait dataset to date—comprising data from eight centers and nine cohorts. Using a standardized pipeline, heterogeneous motion capture data are uniformly converted into anonymized SMPL human meshes. We propose a multimodal benchmark framework integrating 2D-to-3D keypoint lifting, full-body 3D reconstruction, supervised/unsupervised learning, and domain-adaptive transfer learning. Contribution/Results: CARE-PD enables the first large-scale, privacy-preserving, cross-center integration of clinical gait data. Experiments show that pretraining reduces mean per-joint position error (MPJPE) from 60.8 mm to 7.5 mm and improves macro-F1 for PD severity classification by 17 percentage points, significantly enhancing model generalizability and clinical interpretability.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Responsible Web: Data and user privacy-enhancing technologies for the Web
• Security and Privacy: Large-scale security measurements

Abstract

Objective gait assessment in Parkinson's Disease (PD) is limited by the absence of large, diverse, and clinically annotated motion datasets. We introduce CARE-PD, the largest publicly available archive of 3D mesh gait data for PD, and the first multi-site collection spanning 9 cohorts from 8 clinical centers. All recordings (RGB video or motion capture) are converted into anonymized SMPL meshes via a harmonized preprocessing pipeline. CARE-PD supports two key benchmarks: supervised clinical score prediction (estimating Unified Parkinson's Disease Rating Scale, UPDRS, gait scores) and unsupervised motion pretext tasks (2D-to-3D keypoint lifting and full-body 3D reconstruction). Clinical prediction is evaluated under four generalization protocols: within-dataset, cross-dataset, leave-one-dataset-out, and multi-dataset in-domain adaptation. To assess clinical relevance, we compare state-of-the-art motion encoders with a traditional gait-feature baseline, finding that encoders consistently outperform handcrafted features. Pretraining on CARE-PD reduces MPJPE (from 60.8mm to 7.5mm) and boosts PD severity macro-F1 by 17 percentage points, underscoring the value of clinically curated, diverse training data. CARE-PD and all benchmark code are released for non-commercial research at https://neurips2025.care-pd.ca/.

Problem

• Addresses limited objective gait assessment in Parkinson's Disease
• Provides large multi-site clinical dataset for PD motion analysis
• Supports supervised clinical score prediction and unsupervised motion tasks

Innovation

• Anonymized 3D mesh gait dataset from multi-site clinical centers
• Harmonized preprocessing pipeline converting recordings into SMPL meshes
• Benchmarks for clinical score prediction and unsupervised motion tasks