Yes, you can download our open access images at data.midrc.org. In addition to the de-identified images, there are clinical data, annotations and reports available for many images.  Data collection is ongoing, so check back often for updates.

MIDRC is a collaborative initiative led by the medical imaging community, funded by the National Institute of Biomedical Imaging and Bioengineering (NIBIB) and hosted at the University of Chicago. It is co-led by the American College of Radiology® (ACR®), the Radiological Society of North America (RSNA), and the American Association of Physicists in Medicine (AAPM). MIDRC aims to accelerate innovation in medical imaging by providing researchers with a large-scale, de-identified and curated dataset of medical images and associated clinical data. This resource supports the development of effective and useful AI algorithms to improve disease characterization and guide clinical interventions.

MIDRC was created to provide a large, high-quality dataset of medical images to support AI research. Developing AI for medical imaging is often limited by small datasets, inconsistent image quality, and a lack of reliable reference data. MIDRC helps overcome these challenges by offering a standardized collection of curated images and clinical data.

MIDRC was established in 2020 as a collaboration between the American Association of Physicists in Medicine (AAPM), the American College of Radiology (ACR), and the Radiological Society of North America (RSNA). Initially funded by the National Institute of Biomedical Imaging and Bioengineering (NIBIB), MIDRC is hosted at the University of Chicago and benefits from contributions by over 20 research institutions. The data commons is built on the Gen3 platform.

Today, MIDRC receives funding from the Advanced Research Projects Agency for Health (ARPA-H) serving as the exclusive medical imaging performer for Biomedical Data Fabric (BDF) program, advancing AI-driven medical imaging solutions. MIDRC was also selected to participate in the National Artificial Intelligence Research Resource (NAIRR) program, partnering with other federal and private organizations  to build a shared research infrastructure that will strengthen access to critical resources necessary to power responsible AI discovery and innovation.

MIDRC brings together experts in medical imaging to create a comprehensive resource that integrates imaging data with clinical metadata. Unlike other datasets that primarily focus on general patient records, MIDRC prioritizes high-quality imaging data in the standardized DICOM format, ensuring researchers have access to rich, structured data that enhances AI model development.

What sets MIDRC apart is its ability to link imaging data with expansive clinical datasets (e.g., N3C, BioDataCatalyst) and aggregate disparate datasets from affiliated data enclaves (IDC, ACRDart, Stanford AIMI, TCIA) through the Biomedical Imaging Hub (BIH). This centralized approach allows researchers to efficiently index, display, and retrieve imaging and clinical data through a common portal, facilitating large-scale medical AI research.

Additionally, MIDRC uniquely supports a sequestered data commons designed for benchmarking and regulatory use, providing a controlled environment for validation testing and AI model evaluation. The platform also includes a robust set of AI research tools, making it a leading resource for advancing medical imaging innovations.

MIDRC provides image files along with any associated structured data, including patient, imaging study, and imaging series DICOM tags. Where available, MIDRC also provides supplemental patient demographic and clinical data, including conditions, medications, observations, procedures and measurements.  In addition to image files, MIDRC also provides radiology reports and annotation files, as well as textual annotations and other derived/calculated labels.

MIDRC uses the Digital Imaging and Communications in Medicine (DICOM) standard for medical images. DICOM is the international standard for transmitting and storing medical images and related information, providing rich information on acquisition and imaging protocol data as well as the images.  The corresponding image annotations/labels, reports, and clinical supplements may be delivered in several formats including DICOM SR, DICOM SEG, JSON, NIfTI, TXT, XLSX, and CSV.

Medical images follow standardized formats like DICOM, but other health data—such as electronic health records (EHRs), lab results, and patient demographics—can vary widely in structure and terminology. MIDRC is now accepting de-identified radiology reports with labels assigned by the data donor as well as by information extraction software it has developed. MIDRC collaborates with federally funded initiatives to align imaging data with broader medical datasets, ensuring consistency and compatibility across different research projects. For example, MIDRC supports linkages to other non-imaging clinical enclaves to provide a comprehensive record of patient metadata including N3C, All of Us, and BioData Catalyst.

MIDRC began by collecting chest X-ray and chest CT examinations. It has since branched out to include other modalities (e.g., MR, PT, RF, NM, US, XA, and MG) and body parts, including head, abdomen, pelvis, long bones, spine, breast, heart, and esophagus. In many instances, there are longitudinal image records to track changes over time to support research on long-term effects.

MIDRC is working to link imaging data with treatment details and patient outcomes, including long-term follow-ups when possible. MIDRC is enriched by direct secure linkages to repositories that contain extensive clinical metadata that correspond to imaging data contained in the MIDRC enclave.  This includes repositories such as N3C, All of Us, and BioData Catalyst.

The Biomedical Imaging Hub (BIH) is a federated data platform developed by the Medical Imaging and Data Resource Center (MIDRC) under the Biomedical Data Fabric (BDF) program. It enables researchers to search, discover, and analyze biomedical imaging data across multiple repositories through a unified interface. https://imaging-hub.data-commons.org/

Key Features of BIH:

·       Federated Data Access: BIH integrates imaging datasets from various sources, allowing researchers to query and retrieve data without navigating multiple platforms.​

·       Advanced Search and Discovery: The platform offers robust search functionalities, enabling users to locate specific imaging data based on criteria such as modality, anatomy, or clinical attributes.​

·       Analytical Tools: BIH provides tools for data analysis, facilitating tasks like image processing, annotation, and AI model development directly within the platform.​

By consolidating diverse imaging datasets and offering integrated analysis tools, BIH streamlines the research process, promoting collaboration and accelerating advancements in medical imaging and related fields.

The MIDRC Data Commons provides direct access to MIDRC’s centralized image stores, while the MIDRC Biomedical Imaging Hub provides access to imaging structured data distributed across many repositories, including the MIDRC data commons. The structured data in MIDRC BIH includes links to platforms, DOIs, and other persistent digital identifiers (PIDs) like data GUIDs, which can be used to access image files and other data directly from host platforms, but no images can be downloaded directly from the MIDRC BIH. Thus, MIDRC BIH is a data and repository discovery platform that provides cohort building tools over structured data aggregated from various repositories, while the MIDRC Data Commons provides direct, open access to many unique image collections as well as tools to search and select subsets of those imaging data.

No, data access is free. Access is subject to our data use agreements (one for non-commercial use and one for commercial use, see https://www.midrc.org/midrc-data-use-agreement). Available datasets can be found on data.midrc.org.

For technical support, contact MIDRC at midrc-support@gen3.org.

Users must register and agree to the data use policy. All research using MIDRC data must acknowledge MIDRC in publications (see https://www.midrc.org/midrc-acknowledgements).

Yes! MIDRC welcomes contributions from medical centers and hospitals. Visit the MIDRC data intake page for details (https://www.midrc.org/donate).

All submitted data must be de-identified before upload using approved tools. MIDRC reviews and processes all data to ensure no protected health information (PHI) remains. Secure transmission methods are available for data that may contain PHI, which will remain embargoed until fully de-identified.

Most data is openly accessible, but part is reserved for our sequestered dataset for validation testing.

Yes, MIDRC data are available worldwide.

All research using MIDRC data must acknowledge MIDRC in publications. See https://www.midrc.org/midrc-acknowledgements) for how to properly cite MIDRC.

Attend our monthly seminars: https://us06web.zoom.us/webinar/register/WN_FxteE7VTRqOTtMTzojzeGw#/registration

Sign up for the MIDRC newsletter:

https://www.midrc.org/register-to-receive-newsletter

Follow us on LinkedIn for updates:

https://www.linkedin.com/in/midrc/

Subscribe to our YouTube channel for seminar recordings, town halls, and more: https://www.youtube.com/@MIDRC_

MIDRC does not provide funding, but researchers are encouraged to seek independent grants. We can provide letters of support for funding applications.

Visit our contact page for general inquiries. For technical support, email midrc-support@gen3.org.