Machine learning model predicts liver cancer risk with high accuracy
Machine learning model predicts liver cancer risk with high accuracy
Publish Date: 2026-03-26 23:05:00
Source Domain: www.news-medical.net
-
High-Accuracy Model for HCC Risk Assessment: The study published in Cancer Discovery developed a machine learning model that predicts the risk of hepatocellular carcinoma (HCC), the most common type of liver cancer, with high accuracy using patient demographics, electronic health records, and routine blood test results.
-
Study Leadership: The study was co-led by Carolin Schneider, MD, from RWTH Aachen University, and Jakob Kather, MD, MSc, professor of clinical artificial intelligence at the Technical University of Dresden, Germany.
-
Existing Screening Limitations: Current guidelines focus narrowly on high-risk groups, missing many individuals with undiagnosed risk factors such as liver cirrhosis who could benefit from earlier liver cancer screenings.
-
Comprehensive Data Analysis: Researchers utilized data from over 500,000 UK Biobank participants to train their models, with an additional external validation using the All of Us registry data in the United States, encompassing diverse and underrepresented populations.
-
Efficient Predictive Model: A model combining demographics, electronic health records, and blood test results produced the best prediction performance (AUROC of 0.88), demonstrating the model’s potential to work with readily available clinical data without expensive genetic sequencing.
-
Overperforming Existing Models: The developed model outperformed existing liver cancer risk prediction models, such as FIB-4, APRI, NFS, and aMAP scores, by identifying more true HCC cases with fewer false positives.
-
Practical Applications: After reduction to as few as 15 routinely collected clinical features, the simplified version of the model remained superior to prior prediction models, making it practical for clinical settings.
-
Study Challenges Ahead: The study’s retrospective nature and limited representation of patients with viral hepatitis highlight the need for further validation across different populations to confirm the model’s generalizability.
