With the help of artificial intelligence (AI), an inexpensive test found in many doctors' offices may soon be used to screen for hidden heart disease.
Structural heart disease, including valve disease, congenital heart disease, and other issues that impair heart function, affects millions of people worldwide. Yet in the absence of a routine, affordable screening test, many structural heart problems go undetected until significant function has been lost.
"We have colonoscopies, we have mammograms, but we have no equivalents for most forms of heart disease," says Pierre Elias, assistant professor of medicine and biomedical informatics at Columbia University Vagelos College of Physicians and Surgeons and medical director for artificial intelligence at NewYork-Presbyterian.
Elias and researchers at Columbia University and NewYork-Presbyterian developed an AI-powered screening tool, EchoNext, that analyzes ordinary electrocardiogram (ECG) data to identify patients who should have an ultrasound (echocardiogram), a non-invasive test that is used to diagnose structural heart problems.
In the absence of a routine, affordable screening test, many structural heart problems go undetected until significant function has been lost.
In a study published in Nature, EchoNext accurately identified structural heart disease from ECG readings more often than cardiologists, including those who used AI to help interpret the data.
"EchoNext basically uses the cheaper test to figure out who needs the more expensive ultrasound," says Elias, who led the study. "It detects diseases cardiologists can't from an ECG. We think that ECG plus AI has the potential to create an entirely new screening paradigm."
The (Echo)Next step in cardiovascular screening
The ECG is the most used cardiac test in health care. The test, which measures electrical activity in the heart, is typically used to detect abnormal heart rhythms, blocked coronary arteries, and prior heart attack. ECGs are inexpensive, non-invasive, and often administered to patients who are being treated for conditions unrelated to structural heart disease.
While ECGs have their uses, they also have limitations. "We were all taught in medical school that you can't detect structural heart disease from an electrocardiogram," Elias says.
Echocardiograms, which use ultrasound to obtain images of the heart, can be used to definitively diagnose valve disease, cardiomyopathy, pulmonary hypertension, and other structural heart problems that require medication or surgical treatment.
EchoNext was designed to analyze ordinary ECG data to determine when follow-up with cardiac ultrasound is warranted. The deep learning model was trained on more than 1.2 million ECG-echocardiogram pairs from 230,000 patients. In a validation study across four hospital systems, including several NewYork-Presbyterian campuses, the screening tool demonstrated high accuracy in identifying structural heart problems, including heart failure due to cardiomyopathy, valve disease, pulmonary hypertension, and severe thickening of the heart.
In a head-to-head comparison with 13 cardiologists on 3,200 ECGs, EchoNext accurately identified 77% of structural heart problems. In contrast, cardiologists making a diagnosis with the ECG data had an accuracy of 64%.
Finding undiagnosed structural heart problems
To see how well the tool worked in the real world, the research team ran EchoNext in nearly 85,000 patients undergoing ECG who had not previously had an echocardiogram. The AI tool identified more than 7,500 individuals-9%-as high-risk for having undiagnosed structural heart disease. The researchers then followed the patients over the course of a year to see how many were diagnosed with structural heart disease. (The patients' physicians were not aware of the EchoNext deployment so they were not influenced by its predictions). Among the individuals deemed high-risk by EchoNext, 55% went on to have their first echocardiogram. Of those, nearly three-quarters were diagnosed with structural heart disease-twice the rate of positivity when compared to all people having their first echocardiogram without the benefit of AI.
"You can't treat the patient you don't know about."
At the same positivity rate, if all the patients identified by EchoNext as high-risk had had an echocardiogram, about 2,000 additional patients may have been diagnosed with a potentially serious structural heart problem.
"You can't treat the patient you don't know about," Elias says. "Using our technology, we may be able to turn the estimated 400 million ECGs that will be performed worldwide this year into 400 million chances to screen for structural heart disease and potentially deliver life-saving treatment at the most opportune time," Elias says.
Next steps
Elias and his team released a deidentified dataset to help other health systems improve screening for heart disease. The researchers have also launched a clinical trial to test EchoNext across eight emergency departments.
References
Additional information
The study, "Detecting structural heart disease from electrocardiograms using AI," was published July 16 in Nature.
Authors (all from Columbia University except where noted): Timothy J. Poterucha, Linyuan Jing (NewYork-Presbyterian), Ramon Pimental Ricart, Michael Adjei-Mosi, Joshua Finer (NewYork-Presbyterian), Dustin Hartzel (NewYork-Presbyterian), Christopher Kelsey (NewYork-Presbyterian), Aaron Long, Daniel Rocha (NewYork-Presbyterian), Jeffrey A. Ruhl (NewYork-Presbyterian), David van Maanen (NewYork-Presbyterian), Marc A. Probst, Brock Daniels (Weill Cornell Medicine), Shalmali D. Joshi, Olivier Tastet (Montreal Heart Institute), Denis Corbin (Montreal Heart Institute), Robert Avram (Montreal Heart Institute), Joshua P. Barrios (UCSF), Geoffrey H. Tison (UCSF), I-Min Chiu (Cedars Sinai and Kaohsiung Chang Gung Memorial Hospital, Taiwan), David Ouyang (Cedars Sinai), Alexander Volodarskiy (NewYork-Presbyterian-Queens), Michelle Castillo, Francisco A. Roedan Oliver, Paloma P. Malta, Siqin Ye, Gregg F. Rosner, Jose M. Dizon, Shah R. Ali, Qi Liu, Corey K. Bradley, Prashant Vaishnava, Carol A. Waksmonski, Ersilia M. DeFilippis, Vratika Agarwal, Mark Lebehn, Polydoros N. Kampaktsis, Sofia Shames, Ashley N. Beecy (Weill Cornell Medicine), Deepa Kumaraiah, Shunichi Homma, Allan Schwartz, Rebecca T. Hahn, Martin Leon (Columbia and Cardiovascular Research Foundation), Andrew J. Einstein, Mathew S. Maurer, Heidi Hartman, Christopher M. Haggerty (Columbia University and NewYork-Presbyterian), and Pierre Elias.
The study was funded by the National Institutes of Health (R01HL149680), Amyloidosis Foundation, American Heart Association, Patient-Centered Outcomes Research Institute, Fonds de la Recherche en Santé du Quebec, Montreal Heart Institute Research Center, Montreal Heart Institute Foundation, Des Groseillers-Bernard Research Chair, and New York Academy of Medicine.
Columbia University has submitted a patent application on the EchoNext ECG algorithm. Additional disclosures may be found in the paper.
