Artificial intelligence (AI) has once again been proven to be an effective tool in the fight against COVID-19. A new study out of the University of Central Florida demonstrated how AI can be almost just as accurate as a physician in diagnosing the virus present in the lungs, as well as how it can be used to improve testing.
The study was published in Nature Communications.
The team of researchers developed an AI algorithm that could be trained to identify COVID-19 pneumonia in computed tomography (CT) scans, and it demonstrated an accuracy rate up to 90 percent. It was also able to correctly identify positive cases and negative cases, 84 percent and 93 percent of the time, respectively.
CT scans have been proven to be more effective when it comes to COVID-19 diagnosis and progression compared to transcription-polymerase chain reaction (RT-PCR) tests. These tests are often used, but they have high false negative rates and usually take longer to be processed.
One of the biggest reasons CT scans are utilized to diagnose COVID-19 is that they can detect the virus even in individuals who show no symptoms. It doesn’t stop there, however, as they can also detect it in individuals with early symptoms, those who are in the worst stage of the disease, as well as those who have made it out and no longer have symptoms.
With all of its benefits, CT scans also have their downfalls, which is why they are sometimes not recommended for COVID-19 identification. This has to do with the similarities between influenza-associated pneumonia and COVID-19.
The New Algorithm
Taking all of this into account, the team of researchers at UCF developed a new algorithm that is able to accurately spot COVID-19. Not only that, but it can also tell the difference between COVID-19 and influenza, which is extremely useful for physicians.
Ulas Bagci is an assistant professor in UCF’s Department of Computer Science and co-author of the study.
“We demonstrated that a deep learning-based AI approach can serve as a standardized and objective tool to assist healthcare systems as well as patients,” Bagci says. “It can be used as a complementary test tool in very specific limited populations, and it can be used rapidly and at large scale in the unfortunate event of a recurrent outbreak.”
In the study, the team of researchers trained a computer algorithm to detect COVID-19 in lung CT scans, with there being a total of 1,280 patients from China, Japan and Italy observed. The next step was to test the algorithm on 1,337 patients suffering from various lung diseases including COVID-19, cancer and pneumonia not caused by COVID-19.
The results from the computer were then compared to diagnoses from physicians, and the researchers found that the algorithm was extremely effective at accurately detecting COVID-19 pneumonia in lungs, as well as distinguishing between COVID-19 and other diseases.
“We showed that robust AI models can achieve up to 90 percent accuracy in independent test populations, maintain high specificity in non-COVID-19 related pneumonia, and demonstrate generalizability to unseen patient populations and centers,” Bagci says.
The study also involved co-authors Baris Turkbey, who is an associate research physician at the NIH’s National Cancer Institute Molecular Imaging Branch, and Bradford J. Wood, who is the director of NIH’s Center for Interventional Oncology and chief of interventional radiology at NIH’s Clinical Center.
The new development out of UCF is one of the more recent examples of how AI can be leveraged during a pandemic. The technology has been implemented in various different areas relating to COVID-19 including tracking, testing, prevention, diagnosis, research and vaccine development.