Core Concepts
AI outperforms sonographers in assessing left ventricular ejection fraction.
Abstract
The study compared AI and sonographers in assessing left ventricular ejection fraction (LVEF) in echocardiographic studies. AI showed superiority over sonographers in initial evaluations, with fewer corrections needed by cardiologists. The mean absolute difference in LVEF assessments was lower for AI, indicating higher accuracy. Cardiologists had difficulty distinguishing between AI and sonographer assessments. The AI-guided workflow saved time for both sonographers and cardiologists. Study limitations included a single-center population and the need for more training examples for the AI model. The study was published in Nature.
- AI superior to sonographers in LVEF assessment
- Cardiologists had difficulty distinguishing AI from sonographer assessments
- AI-guided workflow saved time for sonographers and cardiologists
- Study limitations included single-center population and need for more training examples for AI model
- Published in Nature
Stats
More than 3500 echocardiographic studies were screened.
Proportion of studies substantially corrected after review: 16.8% in AI group, 27.2% in sonographer group.
Mean absolute difference in LVEF assessments: 2.79% for AI, 3.77% for sonographers.
Cardiologists unable to distinguish AI from sonographer assessments.
Mean absolute difference between previous and final cardiologist assessments: 6.29% for AI, 7.23% for sonographers.
Quotes
"We were surprised that the AI did better than sonographers." - Dr. David Ouyang
"We very much want clinicians to still be in charge." - Dr. David Ouyang
"AI of this sort, once trained on more than 100K videos, should generalize to most institutions." - Dr. David Ouyang