Liver cancer is one of the deadliest forms of cancer worldwide. Hundreds of thousands of people die from it every year, often because the disease is detected too late. The reason is simple. For a long time, it causes no symptoms. By the time patients notice anything, the tumor is often already advanced.
Early screening could save lives. But it is often unclear who actually needs it. This is where a new AI model comes in, developed by researchers at the Else Kröner Fresenius Center for Digital Health (EKFZ) at TU Dresden together with RWTH Aachen University Hospital.
Open to all to advance research
The model assigns individuals to different risk groups. Those at higher risk could be invited for targeted ultrasound screening. This could make preventive care more efficient. All models, the code and a web-based risk calculator are freely available, allowing other researchers to validate and further develop the approach. “In the future, algorithms such as PRE-Screen-HCC could be linked directly to patient records, for example within the European Health Data Space,” explains Prof. Jakob N. Kather. This would be a major step forward. The AI could immediately indicate whether screening is advisable.
Researchers worldwide are working on using AI to detect cancer risks earlier. Many models rely on genetic data or specific biomarkers. The Dresden approach shows that routine clinical data can also be sufficient, which makes implementation in practice easier. At the same time, challenges remain. The model was developed using data from the UK and the US and needs to be validated in other populations. It is also important to assess how well such predictions perform in real-world settings and whether all patient groups are equally represented..
The study was funded by German Cancer Aid and published in the journal Cancer Discovery.
