Researchers say their AI algorithm can analyze clinical data and images of a patient's heart and calculate the probability of cardiac arrest and other concerns over several years.
Researchers at Cedars-Sinai have developed an AI algorithm aimed at predicting heart attacks before they happen.
A team led by Piotr Slomka, PhD, the hospital's director of innovation in imaging and a research scientist at the Smidt Heart Institute's Division of Artificial Intelligence in Medicine, created a tool that collects and sifts through climical data and images of a patient's heart to identify cardiac concerns and determine the likelihood of a heart attack, requirement for an invasive cardiovascular intervention (such as a stent or bypass surgery) or even death over several years.
“This general patient data, together with heart imaging, is what the deep-learning platform uses to make cardiac health predictions,” Slomka said in a press release. “Doctors and patients can use these graphs to track how risk changes over time and to identify individual risk factors. They can also interactively modify certain risk factors to see how it impacts a patient’s particular risk.”
“AI algorithms of this nature could enable physicians to communicate more personalized information regarding potential timing of imminent heart disease events, allowing patients to engage more meaningfully in the shared decision-making process,” added Sumeet Chugh, MD, director of the Center for Cardiac Arrest Prevention in the Smidt Heart Institute and director of the Division of Artificial Intelligence in Medicine and the Pauline and Harold Price Chair in Cardiac Electrophysiology Research. “Even more importantly, this tool has the potential to lend data-led, appropriate urgency to heart disease prevention efforts by both patients and providers.”
The results of the project were recently published in NPJ Digital Medicine. It's being touted as the first study to "evaluate prediction at multiple time points of multiple events in a large multi-site registry of cardiovascular imaging data that also explicitly takes advantage of time-to-event data during model training."
"The model relies on the combined predictive potential of the clinical features, stress test data, and direct image analysis, similarly to the way clinicians try to integrate all available information to provide the most accurate study interpretation," the research team wrote in its study. "Moreover, this approach also leverages time-to-event data to provide more robust risk estimation over time, which could potentially be applied to a broad range of AI tasks."
"In addition to informing the physician about the rationale behind model predictions, the visualization of factors contributing to increased risk of adverse events might serve as a powerful tool in shared decision-making after the exam, utilizing all available information," the team concluded. "When discussed with the patient, a special focus might be given to modifiable risk factors such as high BMI, hypertension, diabetes, and dyslipidemia, leading to optimal, goal-directed medical therapy of these risk factors. That could be a starting point for a discussion on how these factors can be targeted through lifestyle modifications and medications. Such an approach could be an important step towards patient empowerment and could improve adherence to physicians’ recommendations."
Eric Wicklund is the associate content manager and senior editor for Innovation at HealthLeaders.
KEY TAKEAWAYS
Researchers at Cedars-Sinai have developed an AI tool that can both predict the chance of a cardiac event and plot the likelihood of adverse events over time.
The tool analyzes clinical data, such as age, weight, gender, heart rate, and blood pressure, alongside images of the heart that show blood flow to the heart muscle and expansions and contractions.
Researchers say these tools could help care providers develop more personalized care plans for patients and improve patient engagement.