First RCT evidence for use of AI in daily practice

Initial assessment of left ventricular ejection fraction (LVEF) by artificial intelligence (AI) was superior to initial sonographer assessment. After blinded review of initial LVEF assessment, cardiologists were less likely to substantially change their final report with initial AI assessment than sonographer assessment. Furthermore, AI-guided assessment took less time for cardiologists to overread, and was more consistent in test-retest modelling.

Tremendous progress in applying AI to cardiology has recently been made. In that context, Dr David Ouyang (Cedars-Sinai Smidt Heart Institute, CA, USA) reported the findings of the EchoNet-RCT (NCT05140642) [1].

The researchers measured how often cardiologists changed the initial assessment by AI compared with how often they changed the initial assessment by sonographer. Researchers randomly assigned transthoracic echocardiograms performed on adults for any clinical indication to either AI initial assessment or sonographer initial assessment, after which blinded cardiologists reviewed the assessment and provided a final report of LVEF. The AI element made use of a deep learning algorithm called EchoNet-Dynamic, which was trained on echocardiogram videos to assess cardiac function and was previously shown to assess LVEF, using information across multiple cardiac cycles to minimise error and produce consistent results [2]. The primary endpoint was the frequency of a >5% change in LVEF between the initial assessment (AI or sonographer) and the final cardiologist report. The trial was designed to test for non-inferiority, with a secondary objective of testing for superiority.

Transthoracic echocardiograms (n=3,495) were performed on adults for any clinical indication, and then randomised for LVEF assessment to either AI or an experienced sonographer (results summarised in Table). The proportion of assessments that was substantially changed was 16.8% in the AI group and 27.2% in the sonographer group (difference -10.4%, 95% CI -13.2% to -7.7%; P<0.001 for non-inferiority, P<0.001 for superiority). The safety endpoint was the difference between the final cardiologist report and a historical cardiologist report. The mean absolute difference was 6.3% in the AI group and 7.2% in the sonographer group (difference -0.96%; 95% CI -1.34% to -0.54%; P<0.001 for superiority).

Table: EchoNet trial results [1]

In conclusion, Dr Ouyang said: “We were initially quite conservative. This was built as a non-inferiority study, but it actually met that endpoint as well as the endpoint of superiority, so we were pleasantly surprised to see that it was able to work so well. This shows that AI in certain cases is ready for primetime.”

1. Ouyang D, et al. EchoNet-RCT – Safety and Efficacy Study of AI LVEF. Hot Line Session 3, ESC Congress 2022, Barcelona, Spain, 26–29 August.
2. Ouyang D, et al. Nature. 2020;580(7802):252–256.

 

Copyright ©2022 Medicom Medical Publishers

Posted on 27 October, 202227 October, 2022