DSI Collaboration Projects

ACR Connect is expanding to give sites the flexibility to keep data local, in addition to centrally, when participating in a research registry, while keeping this underlying change transparent to the researcher. Distributed support will allow more sites and facilities to participate in research registries, expanding the fidelity of the datasets and richness of research that can be performed. We are working with Rhino Health's solutions to make this pilot a reality. 

AI-LAB is expanding the models available for evaluation to also support operational and cloud-based models. Sites can now use AI-LAB to send their prepared datasets to cloud based operational models and understand how the model would work on their clinical data. This is done in a safe, secure, and free method, allowing organizations to understand how a potential solution would work for them. The first commercial vendor to participate in this new functionality is Whiterabbit.ai with their WRDensity product. 

ACR has been working in collaboration with RSNA, AAPM (American Association of Physicists in Medicine), and the University of Chicago on the MIDRC (Medical Imaging and Data Resource Center) project, funded by the National Institute of Biomedical Imaging and Bioengineering (NIBIB). MIDRC is a large safe-harbor repository of COVID-19 imaging data curated by having multiple scalable intake systems which allows data to be gathered from diverse sources. Over time, MIDRC's goals are to re-use its infrastructure for many different medical imaging scenarios across healthcare. To learn more about MIDRC please visit: https://www.midrc.org/ 

ACR has been working with College of American Pathologists, CAP, to produce discrete data mappings for care scenarios in radiology-pathology concordance. These discrete data mappings will also quicken and higher fidelity concordance workflows as it allows users to compare direct, equivalent details for concordance instead of only being able to compare report text to report text. The first two scenarios produced are BI-RADS and PI-RADS. The ACR has also produced a reference implementation workflow for concordance using this discrete data mapping, which is available under open-source licensing. 

DART is gaining the functionality to allow researchers to annotate datasets, even when the data is not held centrally in DART to create standardized annotations. This will help research as it allows annotations to be centrally and with a small cadre of annotators, without requiring all data be centrally hosted. We are working with Rhino Health's solutions to make this pilot a reality.