An AI model developed at Mayo Clinic shows that routine electrocardiograms (ECGs) can play an unexpected but valuable role in the early detection of advanced chronic liver disease. By analysing subtle patterns in heart signals, the model helped clinicians identify twice as many patients with serious liver disease before symptoms appeared, opening the door to earlier intervention and better outcomes.
Advanced chronic liver disease and cirrhosis are on the rise, driven in part by increasing rates of obesity, hypertension, type 2 diabetes and sleep apnoea. Diagnosis often comes late, when patients present with complications such as gastrointestinal bleeding, ascites or jaundice. Signs that the disease has already progressed to an advanced and sometimes irreversible stage.
“Chronic liver disease is progressive, but early stages are often silent,” says Doug Simonetto, MD, transplant hepatologist at Mayo Clinic and lead author of the study published in Nature Medicine. “The earlier we can detect it, the greater the opportunity to slow or even halt progression, and potentially avoid the need for a liver transplant.”
Linking heart signals to liver health
The approach builds on the close physiological relationship between the heart and the liver. Liver scarring can increase vascular resistance and pressure, which in turn may subtly influence cardiac electrical activity. These changes are not easily recognised by clinicians, but they can be detected by AI models trained on large datasets.
Researchers trained the AI model using ECG data from 11,513 Mayo Clinic patients who underwent routine heart tests. The algorithm searched for patterns associated with advanced liver disease and flagged patients who had not yet been diagnosed through standard care pathways. Diagnoses were subsequently confirmed using validated imaging and blood-based tests.
The results were striking: the AI-enabled ECG analysis identified approximately twice as many cases of advanced chronic liver disease compared with usual clinical practice alone, many in patients who were entirely asymptomatic. “As a family physician, I’ve seen how liver disease can go unnoticed until it’s too late,” says David Rushlow, MD, a Mayo Clinic Health System physician and co-author. “Several patients identified by the AI model had no idea they were living with advanced disease. Detecting it earlier allowed us to intervene at a point where treatment can really change the trajectory.”
Tested in real-world clinical practice
The model was evaluated in a randomised clinical trial involving 248 clinicians across Mayo Clinic in Rochester and the wider Mayo Clinic Health System. The study assessed not only diagnostic accuracy, but also whether the technology could be integrated into everyday clinical workflows.
“What made this study compelling is that it tested AI in a real-world setting,” says Rushlow. “A simple, non-invasive and low-cost test like an ECG is already widely available. Using it to identify patients at risk of serious liver disease could have a meaningful impact at scale.”
Patients identified through the AI-ECG pathway were referred for further evaluation and appropriate management, including lifestyle interventions, medical therapy and specialist follow-up where needed.
Toward preventive, data-driven care
The research is part of Mayo Clinic’s broader Precure initiative, which focuses on predicting and intercepting disease processes before they become clinically apparent or difficult to treat. By embedding AI into commonly used diagnostic tools, the initiative aims to support more proactive, personalised care.
Researchers will continue to follow newly diagnosed patients over the next two years to assess long-term outcomes and the impact of earlier intervention. “We’re only beginning to see what AI-enabled diagnostics can do,” Simonetto concludes. “This work shows how existing data and routine tests, combined with advanced analytics, can help clinicians move from reacting to disease to preventing its most serious consequences.”
