Artificial intelligence outperforms humans in cardiac ultrasound examinations. It’s the leading test for overall heart health and the most rigorous test ever conducted on the subject, discovered Wednesday.
While previous studies have shown the potential power of AI models to read medical scans, the authors of a new US study say this is the first blinded, randomized clinical trial on heart health. said.
“AI holds great promise,” but rigorous evaluation remains important, the study’s lead author David Ouyang told AFP.
This successful trial “really strengthens the argument that we are now prime-time ready,” added a cardiologist at Cedars-Sinai Medical Center in Los Angeles.
An ultrasound of the heart, also known as an echocardiogram, is usually performed on the patient by a sonographer who performs an initial evaluation of the scan before giving it to a cardiologist.
A new study published in Nature compares AI models with sonographers to see who makes the most accurate initial assessments.
Both evaluated ultrasound for what is called the left ventricular ejection fraction, which measures the heart’s ability to pump blood into the body in the space between heartbeats.
Tests are the primary way to measure how well your heart is functioning. It is used to determine if a patient has had a heart attack or is eligible for serious treatment such as the implantation of a defibrillator.
In this study, approximately 3,500 echocardiograms were randomly split between sonographers and AI models.
Their assessments were evaluated by cardiologists, who did not know which were from humans and which were from AI models.
– ‘Exciting’ –
The study found that cardiologists made significant changes in more than 27% of sonographer assessments and about 17% of AI model assessments.
“AI was faster, more accurate, and indistinguishable from a cardiologist,” said Ouyang.
There is a “huge shortage” of sonographers in the U.S. and around the world, and this will save valuable time, he added.
An AI model called EchoNet-Dynamic is trained on approximately 145,000 echocardiograms and uses so-called deep learning to process large amounts of data.
The researchers are now applying for approval of the method by the Federal Drug Administration and hope to get similar approval in the European Union and elsewhere, Ouyang said.
Patricia Pellica, a Mayo Clinic cardiologist who was not involved in the study, told AFP that the study was “exciting” and that the integration of AI tools would improve efficiency and standardization. .
Florian Zolles, a French cardiologist, said the study was well done, but that the technology isn’t as useful in France, where cardiologists do initial evaluations of heart ultrasounds.
