Recent significant strides in AI’s potential to revolutionize cardiovascular care mark a promising era for both clinicians and patients.
As the leading cause of mortality globally, heart disease necessitates precise and early diagnosis to enhance treatment outcomes. AI’s ability to process vast amounts of data swiftly and accurately presents opportunities for advancements in prediction and diagnosis— both in speed and reliability. Could AI become an essential tool for cardiologists in managing patients’ heart health?
Faster heart condition diagnosis
The speed at which heart conditions are diagnosed can be improved through the use of AI, according to recent developments at Cedars-Sinai and the Smidt Heart Institute. Researchers have developed a new AI model that combines image analysis with natural language processing to enhance how cardiologists interpret heart scans, potentially speeding up diagnosis and treatment planning.
The tool, called EchoCLIP, is an AI algorithm designed to analyze ultrasound videos of the heart called echocardiograms (ECGs).
“
THE AI MODEL WAS MORE ACCURATE IN PREDICTING WHO WOULD SUFFER FROM SUDDEN CARDIAC ARREST THAN THE TRADITIONAL ECG RISK SCORE METHOD.
”
—Dr. Ouyang and Dr. Chugh researchers, study published in Communications Medicine
What makes EchoCLIP special? It analyzes ECGs using a training database of more than 1 million videos. That’s around 10 times more than previous AI models. With this immense data pool, EchoCLIP can mimic doctor-level evaluations, tracking critical changes in heart health over time.
“To our knowledge, this is the largest model trained on echocardiography images,” coauthor David Ouyang, MD, a faculty member in the Department of Cardiology in the Smidt Heart Institute and in the Division of Artificial Intelligence in Medicine said in a news release.
Published in 2024 in Nature Medicine, the study behind EchoCLIP shows it can identify devices like pacemakers and provide preliminary readings of heart images, making it a valuable tool for cardiologists. Unlike older models trained on fewer examples, EchoCLIP’s extensive data training allows for more accurate heart analysis.
Predicting the deadliest heart disease
Sudden cardiac arrest claims more than 436,000 lives annually in the U.S., according to the American Heart Association. The condition disrupts the heart’s electrical activity, causing a sudden stop. It’s fatal in 90% of cases and affects those with and without known heart issues. Two recent studies at the Smidt Heart Institute showcase AI’s potential to predict sudden cardiac arrest—a medical emergency that to date has been hard to reliably forecast.
Sumeet Chugh, MD, director of the Division of Artificial Intelligence in Medicine at Cedars-Sinai, says that preventing sudden cardiac arrest is crucial but challenging, and AI could be a game-changer. AI might help doctors identify patients at higher risk for this condition.
In a study featured in Communications Medicine, researchers, including Dr. Ouyang and Dr. Chugh, taught a deep learning algorithm to, in short, analyze patterns in ECGs that display the heart’s electrical activity. The finding: The AI model was more accurate in predicting who would suffer from sudden cardiac arrest than the traditional ECG risk score method.
A second study, published in January 2024 in Circulation: Arrhythmia and Electrophysiology, involves an AI model that distinguishes between two causes of sudden cardiac arrest—pulseless electrical activity and ventricular fibrillation. Pulseless electrical activity can’t be treated with a defibrillator, often leading to death, while ventricular fibrillation can be treated with a defibrillator shock. Identifying which patients are prone to each type through AI could save lives.
This AI model was trained to review ECG patterns and patient traits, then identify risk factors for each type of cardiac arrest. Researchers learned that those with pulseless electrical activity were often older, overweight, had anemia, or experienced shortness of breath. Meanwhile, ventricular fibrillation victims tended to be younger, have coronary artery disease, or experience chest pain.
Being able to identify patterns in the body that the human eye or standard medical tests are unable to detect means getting that much closer to preventing dangerous events like sudden cardiac arrest.
Revolutionizing pediatric heart imaging
AI also holds promise for pediatric heart imaging. The method of choice for assessing heart function in children with complex congenital heart diseases (CHD) is through MRIs. However, hearts in children with CHD differ from those in healthy children and adults, and analyzing their MRI data is highly challenging, time-consuming and prone to variability in interpretation by physicians.
A UC Irvine team developed an AI method for analyzing MRI scans in children with CHD—and it matched the accuracy of physician analysis. Unlike other automated systems, UC Irvine’s AI doesn’t rely on pre-set assumptions, and instead learns directly from MRI images. The team trained their algorithm on data from 64 children at Children’s Hospital of Los Angeles, finding no significant difference in accuracy compared to manual methods.
By generating new segmented MRI data artificially, the AI overcomes the challenge of limited datasets. This advancement suggests a future where machines handle cardiac MRI analyses, easing the load on pediatric cardiologists.
Interpreting medical reports
AI isn’t just for images—it can read medical reports, too. Teaching AI models how to read written notes in physician reports could make it easier for doctors and health systems to track and manage patients with chronic heart conditions. One such condition, aortic stenosis, was the basis of a study by Kaiser Permanente.
Published in the Cardiovascular Digital Health Journal, the research showed that a computer programmed to recognize certain abbreviations, words and phrases could scan nearly a million medical records and echocardiograms, pinpointing almost 54,000 patients with aortic stenosis. This task would take doctors years; the AI model accomplished it in mere minutes.
The research team taught a computer to recognize descriptions of aortic stenosis, then had it examine patient records for those who had an ECG over a 10-year span. This process, known as natural language processing, enabled the computer to analyze nearly 928,000 reports and identify the patients with aortic stenosis.
In a Kaiser Permanente article, the study’s senior author, Alan Go, MD explained, “We were able to train the computer to do what a physician or trained abstractor would do, but on a large scale and without ever getting tired or making mistakes. Importantly, we were also able to train the system to look at the information and measurements from the exams to tell us not only whether a patient had aortic stenosis, but the severity of their condition.”
The evolution of artificial intelligence is reshaping the landscape of cardiology, offering transformative tools that promise to speed up and improve patient care. These advancements not only provide cardiologists with new ways to solve complex challenges but also bring us closer to a future where AI-driven solutions are integral to healthcare.
This supplement is printed and distributed by the Los Angeles Times Media Group. It did not involve the editorial or reporting staffs of the Los Angeles Times.
