The NIH is funding a project that aims to utilize medical imaging and AI to answer unknown questions about COVID-19 and aid diagnosis and treatment worldwide.
NIH to leverage AI and medical imaging for open-source COVID-19 database
August 14, 2020
by John R. Fischer
, Senior Reporter
The National Institutes of Health (NIH) has established the Medical Imaging and Data Resource Center (MIDRC), an initiative seeking to leverage the power of AI and medical imaging for early detection of COVID-19.
The plan is to create an open-source database in which COVID-19 related medical images can be collected, annotated, stored and shared to answer questions researchers have worldwide about the disease. The NIH’s National Institute of Biomedical Imaging and Bioengineering (NIBIB) will fund the project, while the American College of Radiology, the Radiological Society of North America and the American Association of Physicists in Medicine will co-lead its development, which includes the creation and implementation of machine-learning algorithms.
"Imaging can be quite helpful in determining the presence of COVID-19 pneumonia, both before and during the presence of symptoms," Dr. Etta Pisano, ACR chief research officer, told HCB News. "It also helps determine the presence of disease progression."
Medical images of the lungs and heart can be used to evaluate disease severity, predict treatment response and improve patient outcomes. But questions remain about what further roles imaging can play in the fight against COVID-19.
The initiative is expected to provide the international imaging community with a secure, technological network for the development and ethical application of AI to make the best medical decisions for COVID-19 patients, and could stand to benefit other conditions too.
ACR, RSNA and AAPM will collect and annotate thousands of images, and will recruit engineers, physicians and scientists to collect and organize the data. They plan to soon upload more than 10,000 COVID-19 thoracic radiographs and CT images, including from the ACR COVID-19 Imaging Research Registry and the RSNA International COVID-19 Open Radiology Database (RICORD). The information conveyed by the images will be used to train the algorithms so they can answer clinical and logistical questions to help researchers worldwide diagnose and personalize treatments for COVID-19 patients.
The development and implementation of the algorithms is expected to help physicians quickly and accurately assess the signs of COVID-19 and correlate them with other clinical symptoms and tests. The MIDRC will consist of five infrastructure development projects and oversee twelve research projects, including approximately 20 university labs, as part of efforts to find solutions for the COVID-19 pandemic. Its services will later be expanded to provide imaging data and AI technologies to help assess and diagnose other diseases.
"The MIDRC is seeking sites to contribute clinical and imaging data to its COVID-19 efforts," said Pisano. "Interested sites can contribute data through the ACR’s COVID-19 Imaging Research Registry (CIRR), which feeds into the MIDRC. There is a short form to get started available on acr.org."
The MIDRC is funded under the National Institutes of Health’s special emergency COVID-19 process.