Comparison of Diagnostic Accuracies of Radiomics and Non-radiomics Model for Preoperative Prediction of Macrotrabecular-Massive Hepatocellular Carcinoma: A Systematic Review and Meta-analysis.

Rationale And Objectives: The diagnostic value of traditional imaging methods and radiomics in predicting macrotrabecular-massive hepatocellular carcinoma (MTM HCC) is yet to be ascertained. Therefore, this meta-analysis aims to compare the diagnostic performance of radiomics and conventional imaging techniques for MTM HCC.

Materials And Methods: Comprehensive publications were searched in PubMed, Embase, Web of Science, and Cochrane Library up to 28 February 2025. Pooled sensitivity, specificity, positive likelihood ratio (PLR), and negative likelihood ratio (NLR) with 95% confidence interval (CI) were calculated for radiomics and non-radiomics methods using a bivariate random-effects model. Summary receiver operating characteristic (SROC) curves with the area under the curve (AUC) were generated to evaluate overall diagnostic performance. Meta-regression analysis was performed to explore potential sources of heterogeneity. Sensitivity analysis was conducted in radiomics model group, excluding training cohorts. Pairwise comparisons of sensitivity and specificity were performed between radiomic and non-radiomics groups. The 95% CIs of the AUCs of radiomics and non-radiomics groups were compared.

Results: A total of 20 eligible studies, including 1145 MTM HCC lesions and 2839 non-MTM HCC lesions were analyzed. The pooled sensitivity, specificity, and AUC were 0.87 (95%CI: 0.81-0.91, I = 63.60%), 0.78 (95%CI: 072-0.83, I = 77.28%), and 0.89 (95% CI: 0.86-0.91) for radiomics models. Conversely, the non-radiomics models had sensitivity and specificity of 0.62 (95%CI: 0.54-0.70, I = 71.78%) and 0.88 (95%CI: 0.82-0.92, I = 90.09%) and an AUC of 0.80 (95%CI: 0.77-0.84). Meta-regression analysis demonstrated that the tumor size and imaging modality affected the heterogeneity (P<0.001). Comparative analysis indicated significantly higher sensitivity for radiomics model group (P<0.001) and significantly higher specificity for non-radiomics model group (P=0.02). Radiomics model groups demonstrated a significant higher AUC than non-radiomics group due to nonoverlapping 95% CIs.

Conclusion: Radiomics methods have potential advantages in preoperative MTM HCC prediction than non-radiomics methods due to higher sensitivity and AUC. Prospective multicenter validation and standardization of radiomics pipelines are needed to ensure its clinical application.