AI Could Help Catch Fatty Liver Disease Before It's Too Late

Fatty liver disease has become a silent epidemic affecting approximately 30 percent of adults worldwide, yet the condition remains poorly detected until late stages. The core problem, as the body describes, is a disconnect between what is possible and what is practiced: noninvasive blood tests and imaging tools exist and work, but they are underused because physicians lack the bandwidth or workflow incentive to deploy them routinely. This is where AI enters—not as a replacement for existing diagnostics, but as a tool that removes friction from the existing system.

The mechanism is straightforward. AI can automate the calculation of existing risk scores like Fib-4 by pulling from routine blood test results, or it can analyze x-ray images (taken for other reasons) to flag liver fat incidentally. Two specific examples show the promise: Evido's LiverPRO algorithm, validated across more than 470,000 middle-aged people in partnership with Roche, outperformed the standard Fib-4 test; and ALADDIN, evaluated in an international study earlier this year, similarly beat Fib-4 in identifying candidates for the drug resmetirom. An AI model from Osaka Metropolitan University achieved 82 percent accuracy in identifying fatty liver disease from routine chest x-rays.

The stakes are both humane and economic. Early intervention—through diet, exercise, coffee consumption, or medications like semaglutide and resmetirom—can reverse scarring and inflammation. By contrast, the body notes that three-quarters of people are diagnosed only when their condition has become life-threatening, and liver transplants are described as extraordinarily expensive. For health systems already burdened with chronic disease, identifying and treating patients earlier could save vast amounts of money. Research cited in the body also suggests that informing patients they have liver fibrosis increases their commitment to dietary and exercise regimes, creating an additional behavioral incentive.