Improving the accuracy of CT scans for COVID-19 diagnosis

June 4, 2020

BIDS Consulting Data Scientist and LBL Computational Research Scientist Dani Ushizima is working with a team at Berkeley Lab to evaluate and improve the accuracy of CT scans to diagnose COVID-19.   

The American medical community has been skeptical of using CT scans to diagnose COVID-19, because the abnormalities on the lungs of COVID-19 patients often look very similar to other infections and upper respiratory illnesses, including influenza, H1N1, other SARS viruses, and MERS.

3d lung imagingSo Ushizima's team — which includes medical researchers at UCSF, San Francisco's Veterans Affairs Medical Center, and Thomas Jefferson University — is exploring the feasibility of using image recognition algorithms and a data analysis pipeline to evaluate CT scans and chest X-rays, and to accurately distinguish COVID-19-specific abnormalities.

“Before the pandemic, I was already working with researchers at the University of California, San Francisco, to develop image recognition algorithms for early detection of cancer tumors in the body. We’ve developed algorithms that can search CT scans of different materials. Although the problems aren’t exactly the same, I believe I can leverage these efforts to help classify COVID-19 specific lesions in CT scans and chest X-rays,” said Ushizima. 

Since last March, the team has developed a method to narrow a scan's search area to focus only on the lungs, which enables faster analysis and detection of the lung lesions. As part of this process, the team is also developing a central database to collect and consolidate high-quality, anonymized, and publicly available chest X-ray data. The data will eventually be stored at computing facilities that are part of the COVID-19 HPC Consortium, which enables researchers around the country to openly access data through an online portal and use it to test the accuracy of their image recognition algorithms.

"Never has the world been so urgently in need and waiting for more diagnostic testing," said Ushizima. “Data is where the collaboration with medical schools is so vital; in addition to their medical knowledge, they’ve been a vital resource for helping to guide us toward obtaining IRBs, or International Review Board approvals.”


Can CT Scans Be Used to Quickly and Accurately Diagnose COVID-19? Berkeley Lab researchers are working to improve the accuracy of CT scan diagnosis
June 1, 2020  |   Linda Vu   |   Berkeley Lab Computing Sciences News

Featured Fellows

Daniela Ushizima

Computational Research Division, CAMERA, LBNL
Research Affiliate