To identify heart issues like heart attacks and irregular heartbeats, healthcare professionals usually depend on 12-lead electrocardiograms (ECGs). These complicated setups involve multiple electrodes and wires placed on the chest and limbs to monitor the heart’s electrical signals. Unfortunately, these ECGs need specialized tools and knowledge, which not all medical facilities have.
Recently, a group of scientists and medical experts from Scripps Research demonstrated that heart problems could be diagnosed with comparable accuracy using just three electrodes combined with an artificial intelligence (AI) system. In a study published on August 1, 2024, in npj Digital Medicine, the researchers indicated that their AI can generate complete 12-lead ECGs using data from merely three ECG leads. Additionally, doctors can accurately identify heart attacks by analyzing these AI-produced ECGs nearly as well as the standard 12-lead ECGs.
“This opens up opportunities for patients to receive high-quality and timely clinical information without needing to travel to facilities equipped with a 12-lead ECG,” states cardiologist Evan Muse, MD, PhD, who leads cardiovascular genomics at Scripps Research Translational Institute and is an assistant professor of Molecular Medicine at Scripps Research, and co-senior author of the paper. “This likely means increased accessibility to ECG technology, reduced costs, and enhanced patient safety.”
To create the new AI tool, the research team utilized data from over 600,000 12-lead ECGs collected from various patients. Approximately half of those ECGs showed normal heart rhythms, while the other half exhibited different heart conditions. Giorgio Quer, PhD, director of artificial intelligence at Scripps Research Translational Institute and co-senior author of the study, began to explore which combinations of just two or three electrodes could effectively allow AI to recreate the complete 12-lead data.
“We recognized that the leads have some interrelations. Deep learning algorithms enabled us to handle a vast dataset and comprehend these relationships among the leads, facilitating the reconstruction of the entire 12-lead output. We initially aimed to get a full reconstruction using just limb leads, as those are easiest for non-specialists to apply,” explains Quer. “However, we discovered that incorporating a chest lead significantly improved our results.”
The researchers conducted experiments with a group of 238 ECGs, half of which indicated signs of a heart attack. They presented the original 12-lead ECG and an AI-reconstructed ECG—using three selected leads—to cardiologists. The cardiologists were unable to differentiate between the two; they accurately identified heart attack indicators 81.4 percent of the time in the AI-generated ECGs, closely matching the 84.6 percent accuracy of the original 12-lead ECGs.
“It was crucial for us to demonstrate that this algorithm not only functions at a technical level but that the data it generates can be interpreted accurately by cardiologists,” Quer added.
The research team suggests that before this algorithm is utilized for clinical decisions, further studies involving diverse patient groups in various medical environments will be necessary. However, if the tool performs consistently well, it could lead to ECGs being conducted in settings with less specialized equipment and personnel, thus enabling quicker diagnoses and treatments for patients.
“This represents an ideal case for AI—condensing information from a few leads of the (12-lead) electrocardiogram to create a highly informative output, which could have significant practical consequences for patients in the future,” remarks Eric Topol, MD, director and founder of the Scripps Research Translational Institute and executive vice president of Scripps Research.
This research is part of a growing body of work aimed at enhancing the use of AI tools for screening and diagnosing heart-related conditions. In 2023, Quer’s team revealed that a single ECG patch worn over two weeks could help identify patients at the highest risk for atrial fibrillation.
“This latest research exemplifies how AI can facilitate operations that were previously unachievable,” concludes Quer.
