Portal de Periódicos da CAPES

Rapid Exclusion of COVID Infection With the Artificial Intelligence Electrocardiogram

2021; Elsevier BV; Volume: 96; Issue: 8 Linguagem: Inglês

10.1016/j.mayocp.2021.05.027

1942-5546

Zachi I. Attia, Suraj Kapa, Jennifer Dugan, Naveen L. Pereira, Peter A. Noseworthy, Francisco López-Jiménez, Jessica Cruz, Rickey E. Carter, Daniel C. DeSimone, John Signorino, John Halamka, Nikhita R. Chennaiah Gari, Raja Sekhar Madathala, Pyotr G. Platonov, Fahad Gul, Stefan Janssens, Sanjiv M. Narayan, Gaurav A. Upadhyay, Francis J. Alenghat, Marc K. Lahiri, Karl Dujardin, Melody Hermel, Paari Dominic, Karam Turk-Adawi, Nidal Asaad, Anneli Svensson, Francisco Fernández‐Avilés, Darryl D. Esakof, Jozef Bartúnek, Amit Noheria, Arun Raghav Mahankali Sridhar, Gaetano Antonio Lanza, Kevin P. Cohoon, Deepak Padmanabhan, Jose Gutierrez, Gianfranco Sinagra, Marco Merlo, Domenico Zagari, Brenda D. Rodriguez Escenaro, Dev Pahlajani, Goran Loncar, Vladan Vukomanovic, Henrik Kjærulf Jensen, Michael E. Farkouh, T F Luescher, Carolyn Lam Su Ping, Nicholas S. Peters, Paul A. Friedman,

SARS-CoV-2 and COVID-19 Research

ObjectiveTo rapidly exclude severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection using artificial intelligence applied to the electrocardiogram (ECG).MethodsA global, volunteer consortium from 4 continents identified patients with ECGs obtained around the time of polymerase chain reaction–confirmed COVID-19 diagnosis and age- and sex-matched controls from the same sites. Clinical characteristics, polymerase chain reaction results, and raw electrocardiographic data were collected. A convolutional neural network was trained using 26,153 ECGs (33.2% COVID positive), validated with 3826 ECGs (33.3% positive), and tested on 7870 ECGs not included in other sets (32.7% positive). Performance under different prevalence values was tested by adding control ECGs from a single high-volume site.ResultsThe area under the curve for detection of acute COVID-19 infection in the test group was 0.767 (95% CI, 0.756 to 0.778; sensitivity, 98%; specificity, 10%; positive predictive value, 37%; negative predictive value, 91%). To more accurately reflect a real-world population, 50,905 normal controls were added to adjust the COVID prevalence to approximately 5% (2657/58,555), resulting in an area under the curve of 0.780 (95% CI, 0.771 to 0.790) with a specificity of 12.1% and a negative predictive value of 99.2%.ConclusionInfection with SARS-CoV-2 results in electrocardiographic changes that permit the artificial intelligence–enhanced ECG to be used as a rapid screening test with a high negative predictive value (99.2%). This may permit the development of electrocardiography-based tools to rapidly screen individuals for pandemic control.