AI Algorithm Identifies High-Risk Heart Patients

Hypertrophic cardiomyopathy (HCM) is a complex condition characterized by the thickening of the heart muscle, which impairs the heart's ability to pump blood effectively. This forces the heart to work harder to circulate blood throughout the body and can also cause disruptions in heart rhythm, sometimes leading to fatal outcomes. HCM is typically inherited and results from defects in genes that regulate heart growth. The condition affects approximately one in 200 people globally and is one of the leading causes of heart transplantation. Unfortunately, many individuals are unaware they have HCM until symptoms appear, often when the disease has already progressed. In response, a team of researchers has developed an artificial intelligence (AI) algorithm to more accurately and quickly identify patients with HCM, marking them as high risk for closer monitoring during medical appointments.

The AI algorithm, known as Viz HCM, had previously received approval from the U.S. Food and Drug Administration (FDA) for detecting HCM through electrocardiograms (ECGs). However, in a new study led by researchers at Mount Sinai (New York, NY, USA), the algorithm’s capabilities were enhanced by assigning specific numeric probabilities to its findings. While the algorithm once provided broad categories such as "suspected HCM" or "high risk of HCM," the refined model now offers more precise information, such as a percentage likelihood—such as a 60% chance—of a patient having HCM. This advancement allows patients who have not been previously diagnosed with HCM to better understand their individual risk, leading to more personalized evaluations and timely treatment. Such early intervention could help prevent severe complications like sudden cardiac death, particularly in younger patients.

In the study, published in the journal NEJM AI, the team analyzed nearly 71,000 ECG readings taken from patients between March 2023 and January 2024. Of these, the Viz HCM algorithm flagged 1,522 cases as showing potential signs of HCM. To confirm the accuracy of these predictions, the researchers conducted an extensive review of patient records and imaging data to establish definitive diagnoses of HCM. The findings showed that the calibrated AI model effectively aligned its predicted probabilities of HCM with the actual diagnoses, confirming its usefulness in identifying patients at risk. By utilizing this refined risk model, clinicians can prioritize patient care based on their individual risk levels, thus improving clinical workflows. This approach enhances the healthcare experience by enabling more targeted and informed discussions during patient consultations.

“This is an important step forward in translating novel deep-learning algorithms into clinical practice by providing clinicians and patients with more meaningful information. Clinicians can improve their clinical workflows by ensuring the highest-risk patients are identified at the top of their clinical work list using a sorting tool,” said Joshua Lampert, Director of Machine Learning at Mount Sinai Fuster Heart Hospital.