1279-Noninvasive Phenotyping for Liver Cancer Prognosis

Paper Talk · Jul 14, 2026 · 17m

This study introduces MTV-Net, a novel artificial intelligence framework designed to analyze the vascular microenvironment of primary liver cancers through routine CT scans. By extracting quantitative vascular features, the researchers developed two noninvasive biomarkers that accurately identify tumor subtypes and predict post-surgical recurrence risk. This multi-task learning approach proved effective across different patient groups, offering a way to reclassify biologically complex tumors like combined hepatocellular-cholangiocarcinoma. Radiogenomics analysis further revealed that these…