# Radiomics for Predicting the Efficacy of Immunotherapy in Hepatocellular Carcinoma: A Systematic Review and Radiomics Quality Score Assessment

## TL;DR

This review explores how radiomics can predict immunotherapy success in liver cancer, highlighting the need for better standardization and data sharing.

## Contribution

The paper systematically evaluates radiomics models for immunotherapy prediction in HCC and identifies key methodological gaps.

## Key findings

- Radiomics models perform better for short-term responses than long-term outcomes in HCC immunotherapy.
- Combining radiomic features with clinical data improves prediction accuracy.
- Standardization and open data sharing are critical for clinical translation of radiomics.

## Abstract

Radiomics shows strong potential to predict immunotherapy efficacy in hepatocellular carcinoma, whether used alone or with immune checkpoint inhibitors. Current models perform better for short-term responses (mRECIST/RECIST 1.1) than for long-term outcomes (overall survival/progression-free survival). Integrating radiomic features with clinical characteristics markedly improves prediction. Major challenges persist: heterogeneous imaging and protocols, limited external generalizability, weak biological interpretability, suboptimal clinical applicability, and poor data sharing. This review synthesizes current evidence and recommends prioritizing standardization, multimodal and clinical data fusion, prospective multicenter validation, and the adoption of open, FAIR-compliant datasets to facilitate the translation of radiomics into reliable decision-support tools for personalized immunotherapy in hepatocellular carcinoma.

Background/Objectives: Hepatocellular carcinoma (HCC) immunotherapy provides limited clinical benefits, partly due to the lack of reliable efficacy biomarkers. Radiomics, which non-invasively analyzes tumor heterogeneity, shows promising potential for predicting treatment outcomes. Methods: The present study systematically evaluated the predictive performance and methodological quality of radiomics models for assessing immunotherapy efficacy in patients with HCC. A literature search was conducted in PubMed, Web of Science, Embase, and the Cochrane Library for studies published up to 21 June 2025, which developed CT- or MRI-based radiomics models to predict immunotherapy efficacy in HCC. Study quality was assessed using the radiomics quality score (RQS) and the METhodological RadiomICs Score (METRICS). Results: A total of 11 studies were included and categorized by immunotherapy regimen: ICIs alone (1/11), ICIs combined with targeted therapy (6/11), and ICIs combined with targeted therapy plus locoregional therapy (4/11). The models primarily predicted treatment response (7/11), overall survival (OS) (4/11), or progression-free survival (PFS) (4/11). In the ICI monotherapy cohort, AUC values for predicting treatment response ranged from 0.705 to 0.772. In the ICI plus targeted therapy cohorts, AUC or concordance index (C-index) values for predicting the above efficacy endpoints were 0.792–0.956, 0.63–0.77, and 0.54–0.837, respectively. In the combination therapy cohorts incorporating locoregional treatment, predictive models showed AUC or C-index values of 0.721–0.92, 0.817–0.838, and 0.59. Quality assessment revealed a median RQS of 15 (range: 11–19) and a median METRICS of 72.5% (range: 56.0–79.5%) across all studies. Conclusions: CT/MRI-based radiomics uses routine imaging to non-invasively quantify whole-tumor phenotype and heterogeneity, enabling repeatable, longitudinal assessment in hepatocellular carcinoma. Evidence suggests that it can help to identify patients likely to benefit from immunotherapy before treatment. However, clinical implementation requires standardized imaging and analysis protocols, external validation, and transparent reporting.