South Middlesex Times

Researchers from Rutgers Health and RWJBarnabas Health have introduced a groundbreaking technology harnessing artificial intelligence (AI) to improve heart disease detection and monitoring. This innovation involves transforming basic electrocardiogram (ECG) readings into detailed heart motion signals, a process traditionally requiring the more expensive echocardiogram.

The new technology, which is patent-pending, uses generative AI to analyze cardiac tissue speed during a heartbeat using ECG electrical signals. It subsequently translates these signals into speed waveforms resembling those obtained from Doppler ultrasound imaging during echocardiography.

“ECG is very inexpensive,” stated Partho Sengupta, senior study author and chief of cardiology at Rutgers Robert Wood Johnson Medical School and Robert Wood Johnson University Hospital. He emphasized the accessibility of ECG, pointing out that it's used by millions, including those using an Apple Watch, making it far less costly compared to echocardiograms.

The invention addresses an important healthcare need: early detection of heart dysfunction using cost-effective and accessible tools like the ECG, and directing appropriate patients to specialists for more advanced imaging tests.

The research team employed generative adversarial networks (GANs) to develop AI models capable of generating synthetic heart motion waveforms from electrical signals. In extensive testing conducted across various clinical sites in the U.S. and Canada, the AI's accuracy in detecting both diastolic and systolic dysfunction was confirmed.

“This approach can detect heart problems earlier than conventional methods,” Sengupta noted, explaining that traditional ECG changes appear late when risk factors like high blood pressure or diabetes impact the heart muscle. The synthetic technology identifies subtle longitudinal changes in heart function well before any reduction in pumping capacity, also known as ejection fraction.

Extensive testing sought to validate the AI-generated heart-motion waveforms. Certified echocardiographers were unable to distinguish between real and AI-generated waveforms, even in randomized tests. These synthetic measurements also varied in accordance with patient physiology, akin to real measurements.

Critically, the technology demonstrated predictive power. "We found a large subset of patients in South America... our AI was able to use their ECG data to predict death in the survival analysis long before standard ECG analysis indicated problems," shared Sengupta.

Moreover, the new technology could reduce unnecessary medical tests by predicting conditions more selectively. The synthetic TDI technique showed a potential reduction in echocardiograms by 64.3% for detecting systolic dysfunction and 69.9% for diastolic dysfunction, while only missing a small percentage of cases.

The potential applications span beyond screening, possibly enhancing monitoring for cancer patients undergoing cardiotoxic treatments or patients on new medication affecting heart muscle function. Sengupta posed a forward-looking question, “Why do we wait for symptoms to develop when the disease starts far earlier?”

This breakthrough originates from a multidisciplinary team backed by the National Science Foundation's Bridges to Digital Health program. The team independently crafted the AI systems rather than relying on commercial alternatives.

Looking ahead, Sengupta envisions a scenario where digital "twins" of patients' hearts are employed to test treatments virtually, offering a simulation approach akin to NASA's pre-physical reality simulations for Mars landings. This would enable cardiologists to select the most effective therapy based on a detailed "blueprint" of likely success.

Explore more of the ways Rutgers research is shaping the future.