Multimodal biomarkers for early detection and immunotherapy assessment in hepatocellular carcinoma: integrating liquid biopsy, AI-based imaging and tumor immune profiling.

Hepatocellular carcinoma (HCC) commonly develops in chronically diseased liver and remains challenging because of late detection, heterogeneity, recurrence and limited hepatic reserve. Immune checkpoint inhibitor (ICI)-based combinations improve outcomes in advanced HCC, but durable responses occur in only a subset of patients. This structured narrative review evaluates biomarkers by intended use: surveillance and early detection, pretreatment prognostic stratification, treatment selection, on-treatment monitoring, neoadjuvant response assessment and postoperative minimal residual disease (MRD) detection. Serum alpha-fetoprotein and clinical scores are accessible but largely reflect tumor burden, inflammation or liver function. Cell-free DNA methylation and fragmentomics provide early-detection and tissue-of-origin signals, while circulating tumor DNA tracks molecular burden and clonal evolution. Tissue and spatial profiling clarify immune responsiveness and resistance, whereas radiomics and deep learning noninvasively assess whole-tumor and peritumoral heterogeneity. These modalities are complementary but require distinct evidence standards. Most early-detection studies use case-control or enriched cohorts, and most immunotherapy associations are retrospective, prognostic or monitoring-related. No biomarker yet supports routine selection among ICI regimens. Unlike modality-specific summaries, this review aligns each biomarker with a defined clinical decision and evidentiary category and proposes a longitudinal framework integrating baseline blood, imaging and tissue features with on-treatment dynamics, preoperative reassessment and postoperative molecular surveillance. Clinical translation requires standardized assays, representative multicenter cohorts, locked external validation and biomarker-guided trials demonstrating value beyond established criteria.