Artificial Intelligence for Rapid Exclusion of COVID-19 Infection

Artificial intelligence (AI) can be a way to accurately determine whether a person is not infected with COVID-19. An international retrospective study shows that infection with SARS-CoV-2, the virus that causes COVID-19, causes subtle electrical changes in the heart. An AI-enhanced EKG can detect these changes and potentially be used as a fast, reliable COVID-19 screening test to rule out COVID-19 infection.

In the test, the AI-enhanced EKG was able to detect a COVID-19 infection with a positive predictive value – people infected – of 37% and a negative predictive value – people not infected – of 91%. When additional normal controls were added to reflect a 5% prevalence of COVID-19 — similar to a real population — the negative predictive value increased to 99.2%. The findings are published in Mayo Clinic Proceedings.

COVID-19 has an incubation period of 10 to 14 days, which is long compared to other common viruses. Many people show no symptoms of infection and can unknowingly put others at risk. Also, the turnaround time and clinical resources required for current testing methods are significant, and access can be an issue.

“If this is prospectively validated using smartphone electrodes, it will make it even easier to diagnose COVID infection, highlighting what can be done with international collaborations,” said Paul Friedman, MD, president of the Department of Cardiovascular Medicine at Mayo Clinic in Rochester. dr. Friedman is the study’s senior author.

The realization of a global health crisis brought stakeholders around the world together to develop a tool that could address the need to rule out the presence of acute COVID-19 infection quickly, non-invasively and cost-effectively. The study, which included data from racially diverse populations, was conducted by a global volunteer consortium spanning four continents and 14 countries.

“The lessons of this global working group showed what is achievable, and the need forced members in industry and academia to collaborate in solving the complex questions of collecting and transferring data from multiple centers with their own EKG. systems, electronic health records, and variable access to their own data,” says Suraj Kapa, ​​MD, a cardiac electrophysiologist at Mayo Clinic. support validation of new algorithms.”

The researchers selected patients with EKG data around the time their COVID-19 diagnosis was confirmed by a genetic test for the SARS-Co-V-2 virus. These data were compared with comparable EKG data from patients not infected with COVID-19.

Researchers used more than 26,000 of the EKGs to train the AI ​​and nearly 4,000 others to validate the measurements. Finally, the AI ​​was tested on 7,870 EKGs that had not been used before. In each of these sets, the prevalence of COVID-19 was approximately 33%.

To accurately represent a real population, more than 50,000 additional normal EKGs were then added to achieve a 5% prevalence of COVID-19. This increased the negative predictive value of the AI ​​from 91% to 99.2%.

Zachi Attia, Ph.D., a Mayo Clinic engineer in the Department of Cardiovascular Medicine, explains that prevalence is a variable when calculating positive and negative predictive values. Notably, as the prevalence decreases, the negative predictive value increases. dr. Attia is co-first author of the study with Dr. Kapa.

“Accuracy is one of the biggest hurdles in determining the value of a test for COVID-19,” says Dr. Attia. “We need to know not only the sensitivity and specificity of the test, but also the prevalence of the disease. Adding the additional control ECG data was critical in demonstrating how variable disease prevalence – as we have encountered in regions with widely different disease rates at different stages of the pandemic – would affect how the test would perform.”

“This study shows the presence of a biological signal in the EKG that corresponds to a COVID-19 infection, but many sick patients were involved. While it is a hopeful sign, we need to prospectively test it in asymptomatic people using smartphone-based electrodes to confirm its practical use in the fight against the pandemic,” notes Dr. Friedman up. “Studies are now underway to answer that question.”

Reference: June 15, 2021, Mayo Clinic Proceedings.

This study was designed and conceived by Mayo Clinic researchers, and the work was made possible in part by a philanthropic donation from the Lerer Family Charitable Foundation Inc. and through the voluntary support of participating physicians and hospitals around the world who contributed to an effort. to fight the COVID-19 pandemic. Technical support was provided by GE Healthcare, Philips and Epiphany Healthcare for the transfer of ECG data.