Dynamic contrast-enhanced MRI, the standard imaging test for diagnosing hepatocellular carcinoma, may carry measurable traces of the immune cells surrounding the tumor, according to an exploratory study published in European Radiology Experimental. Researchers in Germany found that the strength of a tumor’s arterial wash-in on MRI was inversely correlated with the density of CD3-positive stromal infiltrating lymphocytes, one of the immune features linked to how liver cancers respond to immunotherapy. The association survived statistical correction for multiple testing, unlike several other imaging-histology correlations the team reported, which remained exploratory.
The retrospective study, conducted at the University of Leipzig with ethics approval number 159/25-ek, included 40 patients with histologically confirmed hepatocellular carcinoma: six women and 34 men with a mean age of 61.5 years and a standard deviation of 10.1 years. All patients were treatment-naive for the analyzed lesion and underwent major liver surgery, including liver resection, hemihepatectomy, or hepatectomy. Selection depended on the availability of a radiology-pathology match, so the cohort does not represent a consecutive surgical series. The study period ranged from 2013 to 2021, and nine patients were excluded because significant metallic or movement artifacts compromised their presurgical MR images.
Each patient underwent MRI on a clinical 1.5-T scanner with the hepatocyte-specific contrast agent Gd-EOB-DTPA before surgery. The researchers calculated quantitative parameters reflecting arterial wash-in, portal venous wash-out, and hepatobiliary uptake from signal intensity measurements taken in the native phase, late arterial phase, portal venous phase, transitional phase, and hepatobiliary phase 20 minutes after contrast injection. A trained radiology resident with 2 years of general experience and another with 4 years of experience measured signal intensities using a polygonal region of interest placed on the slice with the largest tumor extension, with supervision from a consultant radiologist with 10 years of abdominal and oncological experience. Inter-reader agreement was excellent for all investigated parameters, with the best value for native signal intensity measurement at an intraclass correlation coefficient of 0.988.
On the pathology side, a board-certified pathologist with 18 years of general experience evaluated formalin-fixed, paraffin-embedded tissue sections without knowledge of the patients or imaging data. The team stained specimens for programmed death ligand 1, glypican-3, CD3-positive tumor- and stromal-infiltrating lymphocytes, CD68, and CD34. CD3-positive cells were assessed separately in intratumoral and stromal zones, with tumors divided into high and low-medium subgroups using cut-offs of more than 10 percent positive cells for tumoral infiltration and more than 20 percent for stromal infiltration. PD-L1 was scored with the combined positive score, tumor proportion score, and immune cell score, and a combined positive score cutoff of 1 separated high from low expression.
The central result concerned the stromal compartment. CD3-positive stromal lymphocyte infiltration showed a moderate negative correlation with arterial wash-in, with a Spearman correlation coefficient of -0.47, an unadjusted p-value of 0.003, and an adjusted p-value of 0.025 after Benjamini-Hochberg correction for multiple testing. Relative arterial phase signal intensity correlated negatively as well, at r = -0.38 with an adjusted p-value of 0.044. Several post-arterial parameters showed moderate positive correlations with stromal infiltration: absolute portal venous wash-out at r = 0.39 with an adjusted p-value of 0.045, relative portal venous wash-out at r = 0.39 with an adjusted p-value of 0.044, and the signal intensity difference between the arterial and transitional phases at r = 0.37 with an adjusted p-value of 0.045. In practical terms, tumors that brightened less dramatically during the arterial phase and washed out more strongly afterward tended to harbor denser stromal lymphocyte infiltrates.
The discrimination analysis reinforced this pattern. Using the pathological cut-offs, 21 patients were classified as having high CD3-positive stromal infiltration and 19 as low. Wash-in discriminated between the two groups with a p-value of 0.006, and arterial relative signal intensity with a p-value of 0.028. Among post-arterial parameters, absolute portal venous washout showed a p-value of 0.017, portal venous washout a p-value of 0.022, and the arterial-to-transitional signal difference a p-value of 0.03. These comparisons suggest that a handful of routinely calculable MRI numbers can separate tumors with sparse stromal immune infiltration from those with denser infiltration, at least within this cohort.
The remaining findings did not withstand correction and are presented as hypothesis-generating. Wash-in was the only arterial parameter correlating with CD3-positive tumoral infiltration, at r = -0.35 with an adjusted p-value of 0.092, and transitional phase relative signal intensity, portal venous phase relative signal intensity, and hepatobiliary uptake measured as relative tumor enhancement also showed moderate negative correlations with tumoral infiltration that lost significance after adjustment. Native signal intensity showed a weak negative correlation with PD-L1 combined positive score, at r = -0.32 with an unadjusted p-value of 0.045 that rose to 0.45 after correction, and tumors with lower PD-L1 combined positive score expression had higher native signal intensity, at 1.45 versus 0.89 with a p-value of 0.012. The difference in signal intensity between the transitional and portal venous phases correlated weakly and positively with glypican-3 expression, at r = 0.33 with an unadjusted p-value of 0.04 that reached 0.404 after adjustment. No statistically significant associations emerged for CD68-positive cells or microvessel density, and no significant differences were identified according to glypican-3 expression in the discrimination analysis.
The authors place these results in the context of a growing effort to read tumor biology from standard imaging. Contrast enhancement characteristics are known to relate to hepatocellular carcinoma biology, reflecting histopathology including arterial supply and, to some degree, tumor grading, and radiomics and radiogenomics analyses have been shown to decipher aspects of the tumor microenvironment and microvascular invasion, though they face challenges of time-consuming workflows and limited standardization. Because biopsy sampling is not routinely performed in all patients, noninvasive MRI-based identification of histopathological features could offer a valuable contribution to treatment optimization. Prior work has also linked CD3-positive lymphocyte infiltration with dynamic contrast-enhanced imaging features, primarily derived from computed tomography, and high concentrations of CD3-positive lymphocytes in tumor or perivascular regions are associated with favorable outcomes in hepatocellular carcinoma.
The study’s limitations are substantial and shape how the findings should be read. This was a retrospective single-center analysis with a relatively small sample size, which limits statistical power and potentially introduces selection bias. The median time between MRI and histopathology was 94 days, with an interquartile range of 41 to 163 days and a range of 3 to 288 days, an interval the authors acknowledge as a meaningful source of uncertainty in interpreting imaging-histopathological correlations. Only curatively treated lesions were investigated, hindering translation to more advanced hepatocellular carcinoma stages, and the patients were not treated with immunotherapy, the cohort for which such associations would be most clinically relevant. Most correlations were weak to moderate, survival data were not available, the analysis did not characterize CD3-positive lymphocyte subpopulations, their functional states, exhaustion, or spatial distribution, and a single-slice region of interest approach may not capture the high heterogeneity of hepatocellular carcinoma. The use of Gd-EOB-DTPA also restricts generalizability, since signal intensities in dynamic post-contrast phases depend heavily on the agent’s pharmacological properties.
The authors conclude that parameters derived from dynamic contrast-enhanced MRI are moderately associated with stromal lymphocyte infiltration in hepatocellular carcinoma, while associations with tumor-infiltrating lymphocytes, PD-L1 expression, and glypican-3 expression were observed only at uncorrected thresholds and should be considered hypothesis-generating. They note that different parameters from the contrast media phases may be associated with distinct features of the tumor microenvironment, but it remains unclear to what extent imaging can predict different tumor-micromilieu aspects. External validation in larger, multicentric studies with survival endpoints is described as strictly necessary before any clinical application, and further work combining these dynamic parameters with diffusion-weighted imaging values, which have been associated with tissue cellularity and lymphocyte infiltration in oncological imaging, could provide complementary information about lymphocyte infiltration.
Subject of Research: Exploratory analysis correlating dynamic contrast-enhanced MRI parameters with immune and molecular features of the tumor microenvironment in hepatocellular carcinoma
Article Title: Dynamic contrast-enhanced MRI and tumor microenvironment in hepatocellular carcinoma: exploratory analysis of CD3 + TILs, PD-L1, and glypican-3
Article References: Romeo, D., Hassan, J. J., Pfister, F., Höhn, A.-K., Tautenhahn, H.-M., Seehofer, D., Scheuermann, U., van Bömmel, F., Denecke, T., & Meyer, H.-J. (2026). Dynamic contrast-enhanced MRI and tumor microenvironment in hepatocellular carcinoma: exploratory analysis of CD3 + TILs, PD-L1, and glypican-3. European Radiology Experimental, 10(1), Article 165. https://doi.org/10.1186/s41747-026-00834-8
Image Credits: AI Generated
DOI: 10.1186/s41747-026-00834-8
Keywords: hepatocellular carcinoma, DCE-MRI, tumor microenvironment, CD3 stromal infiltrating lymphocytes, PD-L1, glypican-3, Gd-EOB-DTPA, arterial wash-in, radiogenomics, liver cancer imaging, immunotherapy biomarkers, exploratory study
News Source: Nathaniel Bowman. (November 5, 2026). MRI signal patterns hint at immune landscape of liver cancer in exploratory study. Scienmag.
