Intravoxel Incoherent Motion Histogram Parameters Combined with Peripheral Blood Markers for Diagnosing the Pathological Differentiation Degree of Liver Cancer

Abstract

Background and Objectives: This retrospective observational study evaluated the diagnostic value of integrating intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) histogram parameters with peripheral blood biomarkers for the non-invasive preoperative assessment of hepatocellular carcinoma (HCC) differentiation. Methods: A total of 156 patients with pathologically confirmed HCC who underwent preoperative liver magnetic resonance imaging (MRI), including IVIM-DWI, and blood biomarker testing were retrospectively analyzed. The intravoxel incoherent motion (IVIM) parameters, particularly the 5th percentile of the true diffusion coefficient (D) and skewness, were assessed alongside serum alpha-fetoprotein (AFP), neutrophil-to-lymphocyte ratio (NLR), and C-reactive protein (CRP). A multivariate logistic regression model was constructed to differentiate poorly differentiated tumors from well or moderately differentiated tumors. Statistical analysis included receiver operating characteristic (ROC) curve comparison and 5-fold cross-validation. Results: Poorly differentiated tumors exhibited significantly lower 5th percentile D-values (0.68 ± 0.10 × 10-3 mm2/s vs. 0.82 ± 0.11, P < 0.001), higher D-skewness, and reduced perfusion fraction. Elevated AFP, NLR, and CRP levels were also strongly associated with poor differentiation (all P < 0.001). The combined model achieved an area under the curve (AUC) of 0.917 with 86.5% sensitivity and 88.4% specificity, outperforming the IVIM-only (AUC = 0.791) and biomarker-only (AUC = 0.811) models. Conclusions: Combining IVIM histogram parameters with inflammatory and tumor-related biomarkers significantly enhances the accuracy of preoperative HCC differentiation grading. This multiparametric, non-invasive approach has strong potential for guiding clinical decision-making and warrants prospective multicenter validation.