Radiomics-Based Delineation of Gross Tumor Volume and Normal Liver Tissue in Hepatocellular Carcinoma using
Luzhou Peng1, Changqin Lu1, Wenbo Zhong2
1Department of Radiology, Shanghai East Hospital, Tongji University School of Medicine, Shanghai 200120, China.
Current Medical Imaging
|June 2, 2026
Summary
Machine learning models using CT radiomics can accurately differentiate hepatocellular carcinoma (HCC) gross tumor volume from normal liver tissue, aiding radiotherapy planning and reducing variability.
Area of Science:
- Radiology
- Medical Imaging
- Machine Learning
Background:
- Accurate radiotherapy target delineation in hepatocellular carcinoma (HCC) requires precise discrimination between tumor and normal liver tissue.
- Current methods can be limited by inter-observer variability.
Purpose of the Study:
- To develop and compare machine learning models using contrast-enhanced CT radiomic features.
- To distinguish HCC gross tumor volume (GTV) from normal liver parenchyma.
Main Methods:
- Retrospective analysis of 82 HCC patients with contrast-enhanced CT scans.
- Extraction and selection of 1,209 radiomic features.
- Evaluation of seven machine learning classifiers using cross-validation.
Main Results:
- Twelve core radiomic features were identified.
- The CatBoost classifier achieved the highest performance with an AUC of 0.98.
- The model demonstrated clinical utility and good separability between tumor and normal tissue.
Conclusions:
- CT-based radiomics and machine learning offer objective differentiation for HCC radiotherapy.
- The CatBoost radiomics model shows potential as a decision-support tool for HCC radiotherapy planning.
