ECG-based AI model doubles early liver disease diagnoses

As rates of obesity, high blood pressure, type 2 diabetes and sleep apnea increase, cases of advanced chronic liver disease and resulting liver scarring or cirrhosis also are rising. Patients often are diagnosed based on symptoms, such as gastrointestinal bleeding, fluid retention or jaundice, which happen when liver disease has progressed to a late stage. This problem led Mayo Clinic researchers to develop an artificial intelligence (AI) model that resulted in twice the number of advanced chronic liver disease diagnoses in patients without symptoms, helping physicians treat them before the disease had progressed.

"Chronic liver disease is a progressive condition, so the sooner we can diagnose it, the sooner we can stop it from advancing to irreversible stages. Early intervention may decrease the likelihood that a patient will need a liver transplant in the future," says Doug Simonetto, M.D., a Mayo Clinic transplant hepatologist and lead author of the study published in Nature Medicine.

The heart and liver are closely linked. Liver scarring may lead to increased local pressure that can affect the heart. As a result, a heartbeat test called an electrocardiogram (ECG) can capture electrical signal changes in the heart connected to advanced liver disease. Dr. Simonetto and colleagues developed an AI model to analyze data from 11,513 Mayo Clinic patients undergoing routine ECGs. The model looked for patterns connected to advanced liver disease in the ECG data, and it found twice the number of patients who were diagnosed by standard methods. The diagnosis was confirmed by validated imaging or blood tests.

As a family physician, I've often seen how advanced liver disease - which frequently has no symptoms until it becomes irreversible - can go undetected. Many patients identified through the AI-ECG model had no idea they were living with advanced liver disease. By identifying these cases earlier, we were able to connect them to the right treatment - at a time when intervention can truly make a difference. For these patients, the technology helped us not only to uncover a diagnosis, it created an opportunity for better health outcomes and, in some cases, may have saved lives."

David Rushlow, M.D., Mayo Clinic Health System family physician and study co-author

In this randomized clinical trial, 248 clinicians at Mayo Clinic in Rochester and throughout the Mayo Clinic Health System participated.

"The idea that a simple, noninvasive and inexpensive test could help identify patients at risk of developing advanced liver disease was very compelling. This study provided an opportunity to evaluate AI in our real-world clinical environment, where the true test of innovation is whether it improves care for patients in the community," says Dr. Rushlow. "We're only beginning to understand the full potential of AI-enabled tools like this and the promise they hold for preventive, personalized care."

In the next steps, the researchers will follow up with the patients, who were newly diagnosed with advanced liver disease, over the next two years.

The research is part of a larger effort at Mayo Clinic called the Precure initiative focused on developing tools that empower clinicians to predict and intercept biological processes before they evolve into disease or progress into complex, hard-to-treat conditions.