Application of large models in imaging diagnosis and prognostic analysis in hepatocellular carcinoma

Abstract

Hepatocellular carcinoma (HCC) remains a leading cause of cancer-related mortality worldwide, with high incidence and death rates. Despite significant progress in conventional diagnostic methods such as imaging studies and biomarkers, inherent limitations hinder their effectiveness. The rapid development of large model techniques has unveiled considerable potential for improving imaging-based diagnosis and prognostic evaluation of HCC. This review highlights recent advances in applying large models to HCC, emphasizing developments in deep neural network architecture and multimodal data integration. It examines how these models enhance early diagnosis accuracy through automated feature extraction and explores their role in integrating clinical variables, radiomics, genomics, and pathology data, offering novel perspectives for prognosis assessment. Despite their promise, challenges such as data quality, model interpretability, and generalization capacity remain. The review concludes by discussing the future potential of large models in HCC diagnosis and prognosis, addressing key challenges and ethical considerations for clinical adoption.