Reviewer: Dr. Bengisu Keskin Meriç

Name of the Study: Safety and Efficacy Study of artificial intelligence (AI) LVEF (EchoNet-RCT)

Published Congress: ESC 2022

Background:

Although there has been great progress in the application of artificial intelligence (AI) to the field of cardiology, there are no randomized controlled studies yet.

Objective:

The aim of this study is to determine whether an integrated artificial intelligence program can save time and improve the accuracy of evaluation of echocardiograms.

Methods:

Researchers; designed a blinded, randomized controlled trial to compare left ventricular ejection fraction in the initial evaluation of echocardiograms with sonographer versus AI-guided measures. Echocardiograms were randomized 1:1 to either the sonographer's pre-interpretation finding or the AI pre-interpretation finding, based on the final decision made by the cardiologist. In the study, it was evaluated how much the preliminary interpretation was edited and changed by the cardiologists from the first interpretation. Primary outcomes was considered; the proportion of studies in which the LVEF was modified by more than 5% in the final report and the mean change in LVEF.

Results:

3495 patients were randomized into the study. Cardiologists could not distinguish between the initial assessments of AI and sonographers [ (Correct 32.3%), (Unsure 43.4%), (Incorrect 24.2) bangs dazzling index: 0.088 ]. Significant change between both reports; 292 (16.8%) in the AI arm and 478 (27.2%) in the sonography design arm (difference -10.4%, 95%CI -13.2% to -7.7, p<0.001). The mean absolute difference was 2.79±5.53 in the AI arm and 3.77±5.22 in the sonographer arm (difference -0.97, CI 95% -1.31 to -0.61, p<0.001). The safety endpoint was the difference between the final cardiologist report and a historical cardiologist report. 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). Results from subgroups based on patient characteristics, imaging study characteristics and cardiologist prediction were consistent with the overall study results.

Conclusion:

In adult patients undergoing echocardiographic assessment of cardiac function, the preliminary assessment of LVEF by AI was superior to that of the sonographer and more consistent with the previous assessment of the cardiologist.

Interpretations:

The incorporation of AI into clinical use could potentially lead to earlier detection of clinical deterioration or response to treatment, providing more precise and consistent assessments.