AI offers a route to early detection of fatty liver disease—a condition affecting over a billion people

More than a billion people worldwide are quietly developing fatty liver disease, a condition where excessive fat accumulates in the organ—often reaching 5 to 10 percent or more of its weight. The disease typically remains silent until it's dangerously advanced, with about three-quarters of cirrhosis cases only diagnosed once they've become life-threatening.

Yet early detection could be transformative. When caught in its initial phases, the condition is largely reversible through lifestyle modifications like weight loss, reduced alcohol consumption, and exercise. Newer medications such as semaglutide and resmetirom have also shown strong results even in moderate to advanced cases.

The detection gap#

The real challenge isn't a lack of diagnostic tools—it's that physicians rarely deploy them. Existing blood tests like the Fib-4 index and enhanced liver fibrosis test can identify at-risk patients efficiently and are often part of standard medical workups. Yet primary care providers facing mounting workload pressures rarely apply these assessments systematically, even for patients with clear risk factors like obesity and type 2 diabetes.

"You have to have something that can run in the background or it's easy to just hit a button and do it," says Jonathan Dranoff, a Yale University medicine professor.

Where AI could intervene#

Researchers envision AI automating what physicians already do—but doing it everywhere, all the time. Jeffrey Lazarus at the CUNY Graduate School of Public Health and Health Policy suggests that machine learning could scan electronic health records en masse to flag high-risk patients. AI could also integrate into routine imaging workflows, analyzing chest X-rays incidentally captured during standard care to spot liver problems.

A team at Osaka Metropolitan University demonstrated this approach in research published last year, developing an AI model that identified fatty liver disease from routine chest X-rays with 82 percent accuracy. Lazarus imagines such algorithms automatically recommending specialist referrals: "The AI could pick up cases of excess liver fat, check for other risk factors such as if the person is overweight, has high cholesterol or type 2 diabetes, and then make a recommendation to the doctor."

New algorithms in the pipeline#

Several AI-based tools are advancing beyond research. The Danish startup Evido has created LiverPRO, now being commercialized with pharmaceutical partner Roche, which uses nine routine blood biomarkers to assess fibrosis risk. Testing on over 470,000 middle-aged people showed it outperformed the conventional Fib-4 score.