Related Experiment Video
Updated: Sep 16, 2025
Advanced Animal Model of Colorectal Metastasis in Liver: Imaging Techniques and Properties of Metastatic Clones
Published on: November 30, 2016
Multi-modality model predicts lung metastasis of hepatocellular carcinoma using MRI and pathological: a retrospective
Kaicai Liu1, Xiaomin Zheng1, Shiyu Wang2
1Department of Radiology, The First Affiliated Hospital of Anhui Medical University, Hefei, 230000, Anhui, China.
Background And Purpose:
Lung metastasis directly affects the prognosis of patients with hepatocellular carcinoma, How to predict the occurrence of lung metastasis and assist in advance can improve the efficacy of surgical treatment for hepatocellular carcinoma. This study aimed to develop and validate a model for predicting future lung metastasis of hepatocellular carcinoma (HCC) to assist in timely intervention.
Methods:
We retrospectively collected MRI (T2W, CE-T1W, and DWI) images of liver tumors and their full slide images of H&E stained biopsy sections for annotation and feature extraction. The Multi-modality model comprehensive prediction model was established based on three feature sets related to lung metastasis: Radiological features, Radiomics features and Pathomics features. The accuracy of the model in predicting lung metastasis in hepatocellular carcinoma was validated using area under curve (AUC), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).
Results:
The Multi-modality model had favorable accuracy for the prediction of lung metastasis in the training cohort (AUC:0.877[95 %CI0.832-0.921], Sensitivity:0.885(0.830-0.923), Specificity:0.755(0.659-0.831), NPV:0.926(0.891-0.982), PPV:0.654(0.524-0.73)), and internal validation cohort (AUC:0.839(0.765-0.913), Sensitivity:0.853(0.781-0.851), Specificity:0.909(0.788-0.964), NPV:0.906(0.867-0.957), PPV:0.841(0.752-0.896)) and external validation cohort (AUC:0.835(0.779-0.890), Sensitivity:0.876(0.751-0.886), Specificity:0.868(0.753-0.911), NPV:0.903(0.852-0.942), PPV:0.845(0.761-0.902)). The Multi-modality model also significantly outperformed any single-modality prediction models: AUC of 0.608 [0.538-0.679] for the Radiological model, 0.759 [0.699-0.818] for the Radiomics model, and 0.808 [0.755-0.862] for the Pathomics model.
Conclusion:
The Multi-modality model proposed in our study could predict the risk of lung metastasis in hepatocellular carcinoma with high accuracy, and may be served as a predictive tool to guide patients in early systematic treatment.
