TLDR: Researchers at Rutgers Health and RWJBarnabas Health have pioneered an AI technology that translates standard electrocardiogram (ECG) readings into detailed heart motion signals, traditionally obtained through more expensive echocardiograms. This generative AI approach, trained using generative adversarial networks, accurately detects heart dysfunctions like diastolic and systolic dysfunction earlier than conventional methods, potentially reducing the need for costly imaging tests and improving patient outcomes.
A groundbreaking artificial intelligence (AI) technology, developed by researchers from Rutgers Health and RWJBarnabas Health, promises to transform the early detection and monitoring of heart disease. This innovative system converts basic electrocardiogram (ECG) electrical activity into sophisticated heart motion signals, typically acquired via echocardiograms, which are significantly more expensive and require specialized expertise.
Dr. Partho Sengupta, Henry Rutgers Professor and chief of cardiology at Rutgers Robert Wood Johnson Medical School and Robert Wood Johnson University Hospital, highlighted the cost-effectiveness of ECGs, noting their widespread availability, even in devices like the Apple Watch. In contrast, an echocardiogram can be 5 to 10 times more expensive due to the need for an expert sonographer.
The patent-pending technology leverages generative AI to analyze the speed of cardiac tissue during a heartbeat from ECG signals. It then generates a speed waveform that precisely mimics those measured by Doppler ultrasound imaging during echocardiography. These waveforms are crucial for doctors to assess the heart’s pumping and relaxing efficiency.
This innovation addresses a critical gap in healthcare by providing an affordable and accessible tool for early detection of heart dysfunction, enabling timely referrals to specialists for more advanced imaging when truly necessary. The research team utilized generative adversarial networks (GANs) to train their AI models, allowing them to produce synthetic heart motion waveforms from electrical data.
Rigorous testing across multiple clinical sites in the United States and Canada confirmed the technology’s high accuracy in identifying both diastolic dysfunction (problems with heart relaxation) and systolic dysfunction (problems with heart contraction).
Dr. Sengupta explained that conventional ECG measurements often detect changes in heart muscle function only at later stages, when risk factors like high blood pressure, diabetes, or coronary artery disease have already caused significant impact. The synthetic ultrasound, however, can detect subtle longitudinal changes in heart function long before the ejection fraction (pumping fraction) decreases, focusing on more nuanced tissue motion rather than just overall pumping capacity.
To validate their system, the researchers conducted blind tests where board-certified echocardiographers were unable to differentiate between real and AI-generated waveforms. Furthermore, the synthetic measurements demonstrated physiological variations consistent with patient age and blood pressure, mirroring real-world data.
Crucially, the synthetic ultrasound proved predictive of patient outcomes. In a study involving a large cohort of patients in South America with long-term ECG follow-up, the AI was able to predict mortality in survival analysis significantly earlier than standard ECG analysis indicated problems.
Beyond early detection, this technology could reduce unnecessary medical tests. Analysis showed it could decrease the number of echocardiograms by 64.3% for detecting left ventricular systolic dysfunction and by 69.9% for diastolic dysfunction, with a low miss rate of 1.4% and 6.5% of cases, respectively.
The clinical applications extend to monitoring cancer patients undergoing cardiotoxic therapies and individuals with hypertrophic cardiomyopathy on new muscle-altering medications. Dr. Sengupta emphasized the proactive potential, questioning why healthcare waits for symptoms when disease onset occurs much earlier, envisioning a future where heart disease screening is as routine as a colonoscopy.
Also Read:
- Advancements Unveiled: Novel Approaches Enhance Generative AI Capabilities
- TU Delft Spearheads ‘Responsible AI in Health Care’ Conference, Bridging Principles to Practical Application
Developed by a multidisciplinary team funded by the National Science Foundation’s Bridges to Digital Health program, this technology was built from scratch without relying on commercial AI solutions. Dr. Sengupta also envisions a future with digital ‘twins’ of patients’ hearts, allowing virtual testing of treatments before application to real patients, akin to NASA’s simulations for Mars landings.
