Home  /   Data Sources  /   Rehabilitation Dataset Directory: Search  /   Dataset Profile Center for Large Data Research & Data Sharing in Rehabilitation

Rehabilitation Dataset Directory: Dataset Profile

Dataset: Anatomical Tracings of Lesions after Stroke (ATLAS)

Basic Information
Dataset Full Name Anatomical Tracings of Lesions after Stroke
Dataset Acronym ATLAS

Release 1.2* of the Anatomical Tracings of Lesions After Stroke (ATLAS) Dataset is a collection of 304 T1-weighted MRIs (Magnetic Resonance Images) with manually segmented diverse lesions and metadata. ATLAS is designed to provide researchers with a standardized training and a testing dataset for lesion segmentation algorithms on T1-weighted MRIs. It can also be used to compare different lesion segmentation techniques.

The images were collected across 11 research groups worldwide participating in the ENIGMA Stroke Recovery Working Group consortium.  Brain lesions were identified and masks manually drawn on each individual brain MRI using MRIcron, an open-source tool for brain imaging visualization. Quality control was performed on each lesion mask by a second tracer, including categorizing lesions.  Additional metadata includes qualitative descriptions of the type of stroke, vascular territory, and intensity of white matter disease developed by an expert neuroradiologist.

*Version 1.1 was updated to v.1.2 on 11/27/2018  with corrections to images in Cohorts 1 and 2 in the previous dataset. The study also included a new version of the metadata data file. The ICPSR codebook has also been updated based on that file.

Key Terms Stroke, MRI, Images, Lesion segmentation, Automated lesion segmentation, Masks
Study Design Cross-Sectional
Data Type(s) Clinical
Sponsoring Agency/Entity NA
Health Conditions/Disability Measures
Health Condition(s) Stroke
Disability Measures NA
Measures/Outcomes of Interest
Topics Neuroimaging, Lesion segmentation, brain structure, T1-weighted images
Sample Population Convenience sample of stroke patients from 11 locations worldwide.
Sample Size/Notes 304 stroke patients
Unit of Observation Individual/Patient

Asia, Europe, North America


Global including: United States, China, Germany, Norway

Geographic Coverage Global
Geographic Specificity NA
Data Collection
Data Collection Mode Clinical data, images
Years Collected Images and data compiled in 2017
Data Collection Frequency Varies with individual patient. Some have multiple MRIs and metadata over time.
Strengths and Limitations
Strengths Large sample of MRI images. In cases with multiple lesions each non-contiguous lesion is identified and masked. Extensive meta-data developed by expert neuroradiologists are available regarding lesion properties and location(s). In some cases longitudinal data is available. Multiple tracers used to ensure data and tracing quality. All image data can be easily accessed using free open source software. Standardized MRI images are available for over three quarters of the cases (n=239).
Limitations No information regarding the lesion(s) impact on individual functioning.
Data Details
Primary Website

Data archived on ADDEP at ICPSR:


Data Access

Data archived on ADDEP at ICPSR:


Data Access Requirements Data Use Agreement, No Cost
Summary Tables/Reports


Data Components

ADDEP/ICPSR-Curated-Metadata (n=304):

  • Metadata available in multiple file types: Excel, SPSS, SAS, Stata, and R. 

MRI Data:

  • Native: Native space (subject space) MRIs (n=304)
  • Standard_MNI: Standardized MRIs (n=239)
Similar/Related Dataset(s)
Stroke focused studies:
Selected Papers
Other Papers
Publications List:

The Anatomical Tracings of Lesions After Stroke (ATLAS), 2017

 A large, open source dataset of stroke anatomical brain images and manual lesion segmentations.

ATLAS Documentation (Read Me, Codebook, PI documentation, User agreement) 


Ask Our Researchers

Have a question about disability data or datasets?
E-mail your question to our researchers at disabilitystatistics@cornell.edu

The Rehabilitation Research Cross-dataset Variable Catalog has been developed through the Center for Large Data Research & Data Sharing in Rehabilitation (CLDR). The Center for Large Data Research and Data Sharing in Rehabilitation involves a consortium of investigators from the University of Texas Medical Branch, Cornell University's Yang Tan Institute (YTI), and the University of Michigan. The CLDR is funded by NIH - National Institute of Child Health and Human Development, through the National Center for Medical Rehabilitation Research, the National Institute for Neurological Disorders and Stroke, and the National Institute of Biomedical Imaging and Bioengineering. (P2CHD065702).

Other CLDR supported resources and collaborative opportunities:

Acknowledgements: This tool was developed through the efforts of William Erickson and Arun Karpur, and web designers Jason Criss and Jeff Trondsen at Cornell University. Many thanks to graduate students Kyoung Jo Oh and Yeong Joon Yoon who developed much of the content used in this tool.

For questions or comments please contact disabilitystatistics@cornell.edu