Mount Sinai Researchers Refine AI for Heart Disease Detection
Researchers at Mount Sinai have fine-tuned an artificial intelligence (AI) algorithm to better identify patients with hypertrophic cardiomyopathy (HCM)—a common but often undiagnosed heart disease. By calibrating the AI tool, called Viz HCM, they can now assign specific probabilities of disease risk to individual patients. This helps doctors prioritize those most at risk for early and tailored medical care.
Previously, the FDA-approved algorithm flagged patients as “suspected HCM” or “high risk.” Now, thanks to the study published in NEJM AI, the system offers more precise assessments. For instance, a patient might be informed they have a “60% chance of having HCM,” according to Dr. Joshua Lampert, Director of Machine Learning at Mount Sinai Fuster Heart Hospital.
Improved Accuracy for Better Patient Outcomes
As per Medical Xpress, this advancement enables patients—especially those not yet diagnosed—to understand their individual risk better. That understanding could lead to quicker evaluations, earlier treatment, and possibly the prevention of serious complications, such as sudden cardiac death, which can impact young patients in particular.0
“This marks a significant step forward,” says Dr. Lampert. “By providing clinicians with a calibrated risk score, we can ensure the highest-risk patients are prioritized. Meanwhile, patients receive clearer, more personalized information.”
Putting AI Into Practice at Scale
The Mount Sinai team tested the Viz HCM algorithm on nearly 71,000 patients who had undergone an electrocardiogram (ECG) between March 2023 and January 2024. The algorithm flagged 1,522 patients as potentially having HCM. Researchers then cross-referenced these results with patient records and imaging data to confirm diagnoses.
Following confirmation, they applied model calibration to determine if the algorithm’s estimated probabilities matched actual diagnoses. The results showed a high degree of accuracy, demonstrating the tool’s potential for practical use in everyday clinical settings.
Shifting from Alerts to Actionable Insights
This upgrade allows cardiologists not only to flag potential HCM cases but also to explain risk in meaningful terms. Instead of offering vague AI alerts, doctors can now guide patients with data-driven, personalized information—leading to earlier engagement and proactive care.
“This kind of granularity helps us rethink how we triage and counsel patients,” says Dr. Vivek Reddy, Director of Cardiac Arrhythmia Services at Mount Sinai. “We can now operationalize AI tools even for rare conditions like HCM.”
A Model for Future AI Integration in Medicine
Co-senior author Dr. Girish N. Nadkarni, Chair of the Windreich Department of Artificial Intelligence and Human Health, emphasizes the real-world application: “It’s not enough to build a high-performing algorithm—we must make sure it actually helps doctors make better decisions and improves patient outcomes.”
By embedding this AI tool into clinical workflows, Mount Sinai is showing how AI can enhance—not replace—medical decision-making. Clinicians gain support in identifying high-risk cases, while patients benefit from more informed care plans.
Looking Ahead: National Expansion of AI Calibration
The next phase of this research aims to expand the calibrated model to other health systems across the country. This move could make the technology accessible to a wider patient population, advancing the integration of AI in heart care on a national scale.