Artificial intelligence has identified a previously unrecognized electrical signal in routine electrocardiograms (EKGs) that predicts sudden cardiac death.
By analyzing over 440,000 heart scans, researchers at UC Berkeley developed an algorithm capable of identifying high-risk patients who currently remain undetected by standard clinical measurements, potentially allowing for life-saving preventive interventions.
AI Identifies Higher Annual Death Risk Through EKG Data
- Routine Testing: The AI utilizes standard EKG data already collected in hospitals, requiring no new, invasive, or expensive diagnostic procedures for patients.
- Improved Accuracy: While current guidelines focus on how much blood the heart pumps with each contraction, the AI identifies a 7% annual death risk group—significantly higher than the 4.6% risk group identified by traditional methods.
- Early Intervention: By detecting subtle electrical patterns before a cardiac event, doctors may be able to offer preventative care, such as implantable cardioverter defibrillators, to patients who were previously considered low-risk.
AI Analysis of Global Heart Data
The research, published in Nature, represents a decade of data collection aimed at solving one of medicine’s most persistent challenges. Sudden cardiac arrest occurs when the heart’s electrical system fails, often resulting in death within minutes.
Unlike a heart attack, which involves blocked blood flow, cardiac arrest is an electrical malfunction. Because these events frequently strike without warning, clinicians have long struggled to identify candidates for implantable cardioverter defibrillators (ICDs).
To train the system, researchers used six years of EKG recordings from Sweden’s unified healthcare system, linking electrical patterns to death certificates. This allowed the AI to isolate unique markers associated with sudden cardiac mortality. The team subsequently validated the algorithm using separate datasets from hospitals in San Diego and Taipei to ensure the findings were consistent across diverse populations.
“Medical decisions are really hard, and I think that’s why AI is so exciting for me. We can not only make better decisions, but also start to understand what’s actually going on with these patients before their heart stops,” explained Ziad Obermeyer, who serves as the study’s primary author and an associate professor at the UC Berkeley School of Public Health.
Comparison of Risk Identification Methods
| Method | Metric | Annual Risk of Sudden Cardiac Death |
|---|---|---|
| Conventional Screening | How much blood the heart pumps with each contraction | 4.6% |
| AI-Enhanced Screening | Electrical Pattern Recognition | 7% |
Researchers Integrate the Algorithm into Hospital Databases
The implementation of this technology is currently moving into real-world hospital settings. Collaborators in Sweden, Taiwan, and the United States are working to integrate the algorithm into existing databases.
When the system identifies a high-risk EKG, physicians may offer patients the opportunity to wear a patch that continuously records heart activity. Such extended monitoring sessions might offer critical insights regarding the biological mechanisms triggering the warning indicators spotted by the AI.
The research involved contributions from two organizations Ziad Obermeyer helped establish, Dandelion Health and Nightingale Open Science, which work to advance medical research through better access to data and artificial intelligence. Ziad Obermeyer emphasized that the project is a slow burn that could have a big return.
The investigators have also launched an online portal where individuals curious about their personal risk status can submit basic demographic details and an email address to receive notifications regarding EKG screening once the tool sees broader implementation.
Contraindications & When to Consult a Doctor
This AI model is an investigative tool and does not replace the judgment of a cardiologist. Patients with known heart conditions, symptoms such as unexplained fainting, chest pain, or palpitations, or a family history of sudden cardiac death should always consult a medical professional for a standard clinical evaluation.
This technology is not currently a substitute for standard diagnostic procedures like echocardiograms or stress tests, and it does not provide immediate diagnostic results for acute cardiac emergencies.
The study highlights that many individuals currently labeled as low-risk may possess hidden vulnerabilities.