Skip Navigation


Artificial Intelligence Transforms COVID-19 Radiology

Artificial Intelligence Transforms COVID-19 Radiology

As part of a multidisciplinary group working to better predict the trajectory of COVID-19-positive patients once admitted to the hospital, Bharath Ambale-Venkatesh, an assistant professor of radiology and radiological science, has turned to artificial intelligence (AI) to make sense of the vast amount of patient biomarkers now available.

Ambale-Venkatesh was involved previously in the Multi-Ethnic Study of Atherosclerosis, which used similar AI techniques — in particular, machine learning and deep learning — to analyze MRI scans, electrocardiograms, blood tests, patient surveys and histories from 3,000 people to determine which of 35 different markers best predicted patient outcomes.

His collaborators on the COVID-19 work include experts from the Cardiovascular Imaging Core Lab and statisticians from the Bloomberg School of Public Health. They have access to patient data through Johns Hopkins CADRE (COVID-19 and Data Research Evaluation). The researchers’ goal is to comb through more than 1,000 markers for COVID-19 that are currently available to determine which are the most clinically useful for predicting not just disease progression, but a patient’s likelihood of survival.

“When COVID-19 first happened, there was a lot of uncertainty about who survives and who doesn’t. Age was a big marker, but everything outside of that was largely unknown,” Ambale-Venkatesh says of the ongoing AI effort.

Early on in the pandemic, Ambale- Venkatesh prioritized X-ray scans and EKGs for COVID-19 diagnosis, as the accuracy of existing tests was unreliable. As reverse transcription polymerase chain reaction testing for the disease continues to become more accurate, the role of imaging evolved to keep pace. Now, X-ray scans and EKGs are used to determine the severity rather than the presence of the disease.

The fundamentals involved in analyzing these markers has remained the same, but the field of radiology (like others in medicine) is learning new things along the way. The challenge is in adapting to that ever-changing understanding of COVID-19, according to Ambale- Venkatesh. “I have never worked on a clinical condition whose understanding evolves as fast,” he says.

In this work, Ambale-Venkatesh was one of the earliest applicants to tap into CADRE, which pulls directly from patient electronic health records across five Johns Hopkins Health System hospitals. In particular, the availability of X-rays and EKGs within 24 hours of the patient being admitted has increased doctors’ effectiveness in identifying key COVID-19 markers.

Ambale-Venkatesh foresees the lessons learned from COVID-19 to be lasting — yielding information that he and others will be studying for many years to come. As such, he has worked to assimilate his COVID-19 findings into his existing research. “I don’t think of COVID-19 so much as an interruption to what I was working on already,” he says, “but rather important work that will become a big part of imaging and clinical studies across many fields.”
back to top button