A machine learning model may enable liver cancer risk prediction with routine clinical information
| Source: EurekAlert! | Original article
A team of researchers from the University of Helsinki has unveiled a machine‑learning model that predicts a patient’s risk of developing hepatocellular carcinoma (HCC) using only data already collected in routine care. The algorithm ingests age, sex, body‑mass index, diagnostic codes, medication histories and a standard panel of blood‑test results such as liver enzymes, platelet count and alpha‑fetoprotein. In a retrospective cohort of more than 120,000 Swedish and Finnish patients, the model achieved an area‑under‑the‑receiver‑operating‑characteristic curve of 0.89, correctly flagging 89 % of individuals who later received an HCC diagnosis while maintaining a low false‑positive rate.
The breakthrough matters because HCC is the world’s fastest‑rising cancer and is usually caught at an advanced stage, when curative options are limited. Current screening programmes rely on ultrasound and biomarker testing but are restricted to patients with known cirrhosis or chronic viral hepatitis, leaving a large proportion of at‑risk individuals unscreened. By leveraging information that primary‑care physicians already have, the new model could expand risk‑based surveillance to a broader population, potentially catching tumours when they are still amenable to surgery or ablation. Early detection also promises to reduce the heavy economic burden of late‑stage treatment on Nordic health systems.
The next step is external validation in diverse ethnic groups and prospective trials that embed the algorithm into electronic health‑record workflows. Regulators will need to assess the model’s safety and bias profile before it can be rolled out as a decision‑support tool. Observers will watch for partnerships with health‑tech firms and for pilot programmes in Finnish and Swedish primary‑care clinics, which could set the template for AI‑driven cancer screening across Europe.
Sources
Back to AIPULSEN