Understand walking
Study patterns in pace, steps, balance, and posture that may help describe changes in gait.
Parkinson’s movement research
Every step
tells a story.
CARE-PD studies how people move to build better tools for understanding Parkinson’s disease.
Image created for illustration; it does not show a study participant.
Walking can change with Parkinson’s.
We work to understand those changes.
Parkinson’s can change the way a person walks, and those changes can differ from person to person and day to day. CARE-PD studies how video and machine learning could help researchers measure those changes more consistently, alongside clinical assessment.
We bring together movement data from different people and settings so new methods can be tested more widely. The work also looks beyond walking, including facial movement.
Looking ahead: CARE-PD’s long-term aim extends beyond walking to other Parkinson’s symptoms and daily challenges that machine learning may help researchers study.
Our research
From a single walk to shared datasets, we connect computer vision with clinical questions.
Learn from video
Turn recordings into movement information that researchers can study while protecting identity.
Measure change
Explore whether movement measures reflect differences across symptoms and treatment conditions.
Build shared foundations
Create datasets and benchmarks that help teams test ideas across people and clinical sites.
These are research directions. CARE-PD tools are being studied and are not a clinical diagnosis or treatment service.
Datasets you can explore.
Open work and shared resources from the wider CARE-PD research community.
CARE-PD
A shared view of Parkinsonian gait across clinical centers.
CARE-PD brings walking recordings together as anonymized 3D body movement data. It gives researchers a way to study gait and test whether models hold up across different sites.
9cohorts
8clinical centers
362participants
8,477walking segments
Figures from the CARE-PD project page and dataset card.
Dataset & model
PD-GaM & GAITGen
PD-GaM brings together 3D walking data across a range of gait severity. GAITGen explores how generated movement can help researchers study less common walking patterns.
Explore the projectResearch dataset
Parkinson’s Pose Estimation Dataset
Movement trajectories and clinical ratings from video based Parkinson’s assessments, with a notebook showing how to use the data.
View dataset notebookCommunity & events
MoCha brings computer vision, movement science, and clinical research into one conversation. Its CARE-PD challenge asks a practical question: can a model trained on one set of sites work well on new ones?
ECCV 2026 workshop
Human Motion Challenges in Real-World and Clinical Settings
A forum for research on human movement, including the CARE-PD benchmark and challenge on Parkinsonian gait.
Workshop detailsPublications
Selected papers on gait, movement, and video based assessment in Parkinson’s disease and related conditions.
- 2026
SynthGait-19K: A Physically Grounded Synthetic Video Dataset for Gait Parameter Estimation
S. Mehraban, X. L. Lin, V. Adeli, M. Mirmehdi, A. Dadashzadeh, C. Hansen, A. Iaboni & B. Taati · arXiv · Project page ↗
- 2026
GAITGen: Disentangled Motion-Pathology Impaired Gait Generative Model — Bringing Motion Generation to the Clinical Domain
V. Adeli, S. Mehraban, M. Mirmehdi, A. Whone, B. Filtjens, A. Dadashzadeh, A. Fasano, A. Iaboni & B. Taati · IEEE/CVF WACV
- 2025
CARE-PD: A Multi-Site Anonymized Clinical Dataset for Parkinson’s Disease Gait Assessment
V. Adeli et al. · NeurIPS, Datasets and Benchmarks Track
- 2024
Benchmarking Skeleton-based Motion Encoder Models for Clinical Applications: Estimating Parkinson’s Disease Severity in Walking Sequences
V. Adeli, S. Mehraban, I. Ballester, Y. Zarghami, A. Sabo, A. Iaboni & B. Taati · IEEE FG
- 2023
Evaluating the Ability of a Predictive Vision-Based Machine Learning Model to Measure Changes in Gait in Response to Medication and DBS Within Individuals with Parkinson’s Disease
A. Sabo, A. Iaboni, B. Taati, A. Fasano & C. Gorodetsky · BioMedical Engineering OnLine
- 2023
Pose2Gait: Extracting Gait Features from Monocular Video of Individuals with Dementia
C. Malin-Mayor, V. Adeli, A. Sabo, S. Noritsyn, C. Gorodetsky, A. Fasano, A. Iaboni & B. Taati · Predictive Intelligence in Medicine (PRIME 2023) · PDF on arXiv ↗
- 2022
Concurrent Validity of Zeno Instrumented Walkway and Video-Based Gait Features in Adults with Parkinson’s Disease
A. Sabo, C. Gorodetsky, A. Fasano, A. Iaboni & B. Taati · IEEE Journal of Translational Engineering in Health and Medicine
- 2022
Estimating Parkinsonism Severity in Natural Gait Videos of Older Adults with Dementia
A. Sabo, S. Mehdizadeh, A. Iaboni & B. Taati · IEEE Journal of Biomedical and Health Informatics
- 2021
Prediction of Parkinsonian Gait in Older Adults with Dementia Using Joint Trajectories and Gait Features from 2D Video
A. Sabo, S. Mehdizadeh, A. Iaboni & B. Taati · IEEE EMBC
- 2020
Assessment of Parkinsonian Gait in Older Adults with Dementia via Human Pose Tracking in Video Data
A. Sabo, S. Mehdizadeh, K.-D. Ng, A. Iaboni & B. Taati · Journal of NeuroEngineering and Rehabilitation
- 2020
Estimation of Orofacial Kinematics in Parkinson’s Disease: Comparison of 2D and 3D Markerless Systems for Motion Tracking
D. L. Guarin, A. Dempster, A. Bandini, Y. Yunusova & B. Taati · IEEE FG
- 2018
Vision-Based Assessment of Parkinsonism and Levodopa-Induced Dyskinesia with Pose Estimation
M. H. Li, T. A. Mestre, S. H. Fox & B. Taati · Journal of NeuroEngineering and Rehabilitation
- 2018
Automated Assessment of Levodopa-Induced Dyskinesia: Evaluating the Responsiveness of Video-Based Features
M. H. Li, T. A. Mestre, S. H. Fox & B. Taati · Parkinsonism & Related Disorders