AI improves cardiologist evaluation of heart function – EchoNet-RCT: Blinded, Randomized Controlled Trial of Sonographer vs. Artificial Intelligence Assessment of Cardiac Function

Presented at the ESC congress 2022 by: David Ouyang, MD- Los Angeles, CA, USA

Introduction and methods

Echocardiography has many advantages; it is portable, involves no radiation and is inexpensive. However, the challenge with this technique is the variability in interpretation.

EchoNet-RCT is an investigator initiated study and the first blinded randomized trial of artificial intelligence (AI) in cardiology. A total of 3,495 transthoracic echocardiograms performed on adults for any clinical indication were included in the study. The echocardiograms were randomly allocated in a 1:1 ratio to initial assessment by an AI algorithm [1] or to initial assessment by the sonographer. The initial assessment was subsequently reviewed by blinded cardiologists who provided a final report of LVEF.

The primary endpoint was the proportion of reports with a >5% change in LVEF between the initial assessment (by AI or sonographer) and the final assessment by the cardiologist.

Main results

• Cardiologists were asked for each echocardiogram to guess whether the initial assessment was made by AI or a sonographer. Cardiologists could not distinguish between AI or sonographer initial assessments (32.3% correct, 43.4% unsure and 24.2% incorrect).
• The proportion of reports with a >5% change in LVEF in initial vs. final assessment was 16.8% in the AI group and 27.2% in the sonographer group (difference -10.5%, 95% CI -13.2% to -7.7%, p<0.001 for both noninferiority and superiority). The mean absolute difference was 2.79% in the AI group and 3.77% in the sonographer group in terms of absolute LVEF (difference -0.97%, 95% CI -1.31% to -0.61%, P<0.001 for superiority).
• The key secondary safety endpoint was the difference between the final cardiologist assessment and a historical cardiologist assessment. A >5% change in LVEF in final vs. historical assessment was found in 50.1% of cases in the AI group and in 54.5% in the sonographer group (difference -4.5%, 95% CI -7.8% to -1.2%, P=0.008 for superiority). The mean absolute difference was 6.29% in the AI group and 7.23% in the sonographer group (difference -0.96%, 95% CI -1.34% to -0.54%, P<0.001 for superiority).

Conclusion

This randomized study showed that initial echocardiogram assessment of LVEF by AI was noninferior and superior to initial sonographer assessment.

-Our reporting is based on the information provided at the ESC Congress-

References