Related Experiment Video
Updated: Sep 2, 2026
Visualization, Quantification, and Mapping of Immune Cell Populations in the Tumor Microenvironment
Published on: March 25, 2020
Explainable Radiomics Model Based on Intratumoral and Peritumoral CEMRI Features: Predicting Macrotrabecular-Massive
Yanxi Xiong1, Ying Zhao2, Xingwu Xie3
1Department of Radiology, the First Affiliated Hospital of Dalian Medical University, Dalian, Liaoning, PR China (Y.X., Y.Z., A.L.); Department of Radiology Imaging Center, Renmin Hospital, Hubei University of Medicine, Shiyan, Hubei, PR China (Y.X., X.X., Y.Z., X.X., X.T.).
Rationale And Objectives:
Macrotrabecular-Massive Hepatocellular Carcinoma (MTM-HCC) is characterized by strong invasiveness and poor prognosis, necessitating noninvasive preoperative identification methods.
Purpose:
To construct and validate a radiomics model for MTM-HCC prediction based on contrast-enhanced magnetic resonance imaging (CEMRI), and to explore its molecular and immune-related features.
Methods:
A retrospective analysis was performed on CEMRI data of 209 pathologically confirmed hepatocellular carcinoma (HCC) patients (including 88 MTM-HCC cases), divided into a training set (146 cases) and a test set (63 cases) at a 7:3 ratio. For the first time, six radiomics models were constructed based on portal venous phase images obtained before and after super-resolution (SR) reconstruction, covering three feature combinations: intratumoral features, intratumoral + 5 mm peritumoral features, and intratumoral + 10 mm peritumoral features. The optimal model was integrated with clinical factors to construct a nomogram. Based on radiomics signature labels, risk stratification, molecular analysis, and immune infiltration analysis were conducted using transcriptomic data of 32 HCC patients from The Cancer Genome Atlas (TCGA) and Gene Set Variation Analysis (GSVA).
Results:
The fusion model of intratumoral + 5 mm peritumoral features after super-resolution reconstruction (sup_IntraPeri5mm) showed the best performance, with area under the curve (AUC) values of 0.892 in the training set and 0.812 in the test set. The nomogram integrating the radiomics model with Child-Pugh class/score and protein induced by vitamin K absence or antagonist-II (PIVKA-II) further improved predictive efficacy (AUC = 0.906 in the training set, AUC = 0.853 in the test set). The high-risk group was enriched in angiogenesis, epithelial-mesenchymal transition (EMT), and inflammatory pathways, with increased infiltration of resting mast cells, suggesting a distinct immunosuppressive tumor microenvironment.
Conclusion:
This study establishes a noninvasive radiogenomic framework that can accurately predict MTM-HCC and reflect its molecular and immune characteristics, providing new insights for individualized diagnosis and therapeutic stratification.
