Groundbreaking AI tool could save lives from sudden cardiac arrest

A groundbreaking AI tool could soon help prevent sudden cardiac deaths by accurately identifying high-risk patients with a dangerous heart condition—before it's too late. The new model, developed by researchers at Johns Hopkins University, is proving to be far more precise than traditional methods, thanks to its ability to read hidden details in heart images that most doctors can't interpret.

The new artificial intelligence model is called MAARS, which stands for Multimodal Artificial Intelligence for Ventricular Arrhythmia Risk Stratification. This tool uses a deep learning system designed to forecast the risk of sudden cardiac death in people with hypertrophic cardiomyopathy, or HCM—a common genetic heart disease that affects roughly one in every 200 to 500 individuals around the world. It is one of the leading causes of unexpected cardiac death in young athletes and otherwise healthy adults.

HCM is often misunderstood. Many who have it live normal, healthy lives. But for a smaller group, the risk of fatal heart rhythm problems is much higher. The challenge for doctors has always been figuring out who’s actually at risk. According to cardiologist and lead researcher Natalia Trayanova, the standard methods currently in use—both in the U.S. and Europe—have about a 50% success rate in predicting who’s in danger. “Not much better than throwing dice,” she said.

MAARS, on the other hand, changes the game completely. It learns from a wide range of patient data—including electronic health records, doctor notes, and imaging results—then finds patterns that help predict whether someone is at risk for sudden cardiac death. Most important of all, it uses contrast-enhanced MRI scans of the heart, which offer unique insights that were rarely used before in risk predictions.

People with HCM often develop scar tissue inside the heart, a process known as fibrosis. This scarring raises the risk of sudden cardiac death but is difficult for doctors to measure using MRI images. The AI model doesn’t face the same challenge when reading these images. “People have not used deep learning on those images,” said Trayanova. “We are able to extract this hidden information in the images that is not usually accounted for.” MAARS focuses on this unseen data, highlighting scarring patterns that signal the highest risk. It does this more effectively than the human eye.

The model uses a transformer-based neural network—a modern form of AI. It can learn from several types of patient data at the same time. This includes test results, patient history, and MRI findings. The model identifies links between them to make accurate predictions.

The results are impressive and hard to ignore. Researchers tested MAARS using real patients at Johns Hopkins and Sanger Heart & Vascular Institute in North Carolina. The model reached 89% predictive accuracy across all test patients. In people aged 40 to 60—the highest risk group—it scored 93% accuracy.

This tool is also consistent across a range of patient groups. Current clinical guidelines can miss the mark for certain age or racial groups. MAARS performs well no matter the patient’s age, sex, or race. That means more people receive the correct care.

Side-by-side comparisons showed clear differences. MAARS outperformed standard guidelines across all metrics. In internal tests, it achieved a 0.89 AUC score—well above the current 0.62. In external tests, MAARS scored 0.81, showing continued strong performance.

Clinical guidelines don’t deliver this level of accuracy. Even worse, they often recommend defibrillators that may not be necessary. These implants are invasive, expensive, and impact quality of life. MAARS can help reduce these unnecessary surgeries.

“Currently we have patients dying in the prime of their life because they aren’t protected,” Trayanova said. “And others who are putting up with defibrillators for the rest of their lives with no benefit.” The model could help fix both problems at once.

AI in medicine is often called a “black box” because it doesn’t explain its reasoning. MAARS changes that by showing why a patient is at high risk. This lets doctors understand and trust the system's predictions. They can also use the details to plan personalized care.

“Our study demonstrates that the AI model significantly enhances our ability to predict those at highest risk,” said Jonathan Crispin. He’s a cardiologist and co-author of the study. “This model has the power to transform clinical care,” he added.

MAARS also helps scientists discover new risk factors. It highlights subtle patterns in heart images and medical histories. These patterns could reveal early signs of disease that doctors often miss. Such discoveries may lead to better treatments in the future.

The model now focuses on patients with hypertrophic cardiomyopathy. But the Johns Hopkins team has bigger plans for MAARS. They aim to expand it to cover more heart diseases. That includes cardiac sarcoidosis and arrhythmogenic right ventricular cardiomyopathy. These conditions can also cause sudden cardiac death if mismanaged. MAARS could be adapted to help predict risk in those patients too.

The research was published in Nature Cardiovascular Research. It marks a major advance in predicting heart risk. With more testing, MAARS could be part of everyday medical practice. The model was trained using federally funded data. That shows the value of public investment in medical research.

Tools like this one could soon save countless lives. Researchers now hope to move the field forward. Their goal is smarter tools and targeted treatments. They want decisions to be based on solid, data-driven insights—not guesswork.

Note: The article above provided above by The Brighter Side of News.

Like these kind of feel good stories? Get The Brighter Side of News' newsletter.