Improving Early Liver Disease Diagnosis With AI ECG, With Doug Simonetto, MD

Chronic liver disease continues to present a paradox in modern hepatology: it is both highly prevalent and frequently undiagnosed until late stages, when meaningful intervention is more limited. Despite growing awareness and advances in noninvasive assessment, there remains no standardized, scalable approach to identifying patients earlier in the disease course, particularly those who are asymptomatic.

“Unfortunately, we don't currently have a systematic way of catching these patients early, and most patients present to us or to their primary care physicians when they already have symptoms of advanced liver disease, and many times it's really too late for us to have effective interventions to alter the natural course of their diseases,” Doug Simonetto, MD, a Mayo Clinic transplant hepatologist, told HCPLive.

Current strategies rely largely on laboratory-based scoring systems and risk stratification tools, yet these approaches are not applied consistently, and universal screening is not recommended. As a result, detection often depends on clinical suspicion or the onset of complications, leaving a substantial proportion of patients unidentified.

The burden of early detection, Simonetto emphasized, must shift upstream. Primary care clinicians are often the first and only point of contact for patients in the asymptomatic phase.

“We need to move toward targeted or opportunistic screening in higher-risk populations,” he noted, underscoring the importance of identifying disease before patients reach hepatology clinics.

Against this backdrop, emerging technologies are beginning to reshape the conversation. In a recent study published in Nature Medicine, Simonetto and colleagues evaluated an artificial intelligence-enabled model applied to routine electrocardiograms (ECGs) to detect advanced chronic liver disease. The approach builds on longstanding observations that cirrhosis can produce detectable changes in cardiac electrophysiology, including alterations in heart rate variability.

Using deep learning, the model identified ECG-based signals associated not only with advanced disease but also with earlier, subclinical stages, prompting further evaluation in a primary care-based clinical trial. The findings suggest that routinely collected data may hold value as an opportunistic screening tool.

Still, significant hurdles remain. AI models, while powerful, introduce challenges related to interpretability, validation, and clinician trust.

“We need rigorous clinical studies to demonstrate that these tools improve outcomes,” Simonetto said, noting that premature implementation risks adding to clinician burden without clear benefit.

Looking ahead, the field appears to be moving toward a broader paradigm shift, one that some have termed “preventive hepatology.” Rather than reacting to advanced disease, future strategies may integrate ECG signals with biomarkers, imaging, and genetic risk to enable multimodal, earlier detection.

“The goal,” Simonetto said, “is to identify patients earlier and change the trajectory of disease—before they ever need transplant.”

Editors’ Note: Simonetto reports no relevant disclosures.

References

1. Simonetto DA, Rushlow D, Liu K, et al. Detection of undiagnosed liver cirrhosis via AI-enabled electrocardiogram: a pragmatic, cluster-randomized clinical trial. Nat Med. 2026;32(1):160-167. doi:10.1038/s41591-025-04058-y

2. Satake A. Mayo Clinic researchers develop AI-ECG model to diagnose liver disease earlier. Mayo Clinic. December 17, 2025. Accessed March 18, 2026. https://newsnetwork.mayoclinic.org/discussion/mayo-clinic-researchers-develop-ai-ecg-model-to-diagnose-liver-disease-earlier/?cjdata=MXxOfDB8WXww&cjevent=2c8878ed047d11f1805201900a82b838&cm_mmc=CJ-_-100357191-_-5250933-_-Evergreen+Link+for+Mayo+Clinic+Diet&utm_source=cj&utm_content=100357191&utm_capaign=3-months