Tools, Agreements and Instructions for Sites Contributing Data to the RSNA International COVID Open Radiology Database (RICORD)
Early in the pandemic crisis, RSNA distributed a survey regarding the interest of radiology sites to share COVID-related imaging data to support research and education. The strong positive reaction--more than 200 respondents from around the globe--speaks to the determination of the radiology community to contribute its resources and expertise to addressing the global pandemic. Additional interested sites are invited to complete the data sharing survey. RSNA has developed partnerships with key organizations to launch the RSNA International COVID-19 Open Radiology Database (RICORD).
We have put in place a pathway for radiology sites anywhere in the world to contribute data to RICORD. Data gathered at contributing sites according to the defined data protocols can be de-identified and transferred to RSNA’s data staging platform (provided by MD.ai), organized into a consolidated database, checked, annotated and made available to the research community.
This resource page is intended to provide radiology organizations the agreements, tools and instructions needed to enable them to submit data in accord with best practices, including respect for patient security and privacy concerns. Please review and follow the steps below in order to participate in this important initiative.
Review the Data Inclusion Criteria and determine whether your institution is able to contribute the required and optional data requested. If so, begin planning to identify and extract the relevant data. Engage your site's technical staff in discussions regarding how to extract, de-identify and transfer data.
Review the De-identification Protocol and determine your institution's ability to meet its requirements. If you decide to use the free DICOM Anonymizer tool provided by RSNA, have your technical staff download the tool and carefully review the instructions document.
Review the Data Submission Agreement and have an authorized signatory for your organization review, sign and return the document to informatics@rsna.org. RSNA staff will return the fully counter-signed agreement.
When the Data Submission Agreement is signed, you will receive an email message from RSNA staff with unique https:// security codes and a DICOM address to use in transmitting data to the RICORD data staging site.
Extract, organize and de-identify your data and transmit it to the RICORD data staging site following the data transfer instructions provided below.
We plan to start with a core set of data that will rapidly be made available for research and to expand the volume and types of data in the RICORD repository over time.
Patients >= 18 years old with chest imaging (CT and/or X-ray) for suspected COVID-19 infection (for example ICD-10-CM code U07.1 among discharge diagnosis).
o Positive infection should be confirmed with one or more of the following:
RT-PCR test
Positive IgM antibody test
Clinical diagnosis using hospital specific criteria.
o Positive infection encompasses:
Patients who are diagnosed prior to hospitalization, but still symptomatic during the hospitalization
Patients with a positive test or diagnosis during hospital admission
Patients who are symptomatic during the hospitalization and have a confirmed test available only after hospital discharge
Exclude:
o Patients < 18 years of age
o Patients who did not receive chest imaging (CT and/or X-ray)
Imaging examinations:
Chest CT examination (axial series only, preferred slice thickness 2.5-5.0 mm; any protocol, body kernel preferred) of COVID-19 patients
and/or
Chest x-ray examination (any protocol) of COVID-19 patients
Data for each imaging examination (please review section 3. Data De-Identification, below):
Unique Patient ID (following de-identification)
Unique Imaging Exam ID
Age
Sex
COVID-19 positive confirmation testing method (PCR, serological, clinical)
COVID testing method and other supporting clinical information should be provided in spreadsheet format that correlates with the subject's imaging study information. A sample spreadsheet is provided here.
Imaging examination type:
Chest CT exams (axial series only, preferred slice thickness 2.5-5.0 mm; any protocol, body kernel preferred) of COVID-19 NEGATIVE patients
and/or
Chest x-ray exams (any protocol) of COVID-19 NEGATIVE patients
Data for each imaging examination:
Patient ID (following de-identification)
Unique Imaging Exam ID
Age
Sex
COVID-19 NEGATIVE confirmation testing method (PCR, serological)
Alternative Diagnosis (if any)
Optional clinical data elements
Optional clinical data elements for each patient are requested if available. A table of relevant data is provided here: COVID-Relevant Clinical Data Elements
Please complete and return a signed copy of the data submission agreement found here: Data Submission Agreement
The agreement must be signed by an authorized signatory for your organization. It should include a detailed description of the submission data, including the imaging locations; number and modality of images; number of subjects; data elements and number of cases for supporting clinical data; and associated analyses, etc. When completed, send the signed agreement to informatics@rsna.org. RSNA will quickly return the fully executed agreement.
The terms of the agreement are identical with those of the National Cancer Institute's Cancer Imaging Archive (TCIA), which will be the platform initially used to share data with the research community.
Contributing sites are responsible for de-identifying imaging examinations and associated clinical information for inclusion in this public research and educational database (non-commercial use). Please follow your institutional policies and procedures and adhere to locally applicable regulations and best practices for doing so.
RSNA has worked with experienced DICOM experts to develop a data de-identification protocol for RICORD based on recognized standards and best practices. The protocol is available for download here: Data De-Identification Protocol
Please review the protocol and ensure that your de-identification procedures are able meet this protocol. Note that the protocol requires that a unique (pseudonymous) identifier be created for each subject.
RSNA has also developed a free open-source DICOM de-identification software tool called Anonymizer that implements the Data De-Identification Protocol. It can be downloaded from github here:
RSNA has contracted with MD.ai to provide the data staging platform and annotation tools for RICORD. When your institution has signed the Data Submission Agreement, you will receive an email message containing unique security information and a DICOM address for your institution to use in transfering data to the MD.ai staging platform. Please follow the instructions here to access your personalized secure link to transfer data: https://covid.dev/rsna-ricord-data-transfer-instructions
Please send any questions or concerns you have about the process to informatics@rsna.org. We have also set up a discussion list for communication among participating sites and the RICORD task force. You can view and request to join the list here.