A machine learning model trained on over 1.7 million young adults in South Korea can predict early-onset liver cancer with impressive accuracy. The AI-powered boosted survival model flagged key risk factors — including viral hepatitis, cirrhosis, and abnormal liver and kidney tests — even in those without classic warning signs. It could help identify high-risk individuals who would otherwise be missed by conventional screening approaches.
Researchers in South Korea have developed a machine learning model that can predict the risk of early-onset hepatocellular carcinoma (HCC) — liver cancer — in adults aged 20–39, a group largely overlooked by existing risk tools. Using nationwide health screening data from over 1.75 million young adults, the team built and validated both traditional and AI-based models, with a generalized boosted survival model coming out on top.
The winning model identified viral hepatitis, prior non-liver cancer, cirrhosis, and abnormal liver and kidney function tests as the strongest predictors. Crucially, it maintained predictive power even among individuals without established major risk factors — meaning it could catch people who'd typically be missed by conventional screening approaches.
By the Numbers
Why it matters: Early-onset liver cancer is rare but rising, and most prediction tools aren't built with younger patients in mind. This AI model could enable earlier, more targeted screening — potentially saving lives by identifying at-risk young adults before symptoms appear.
