Construction of a prediction model for local tumor progression in patients with hepatocellular carcinoma after RFA
10.3760/cma.j.cn113884-20241217-00380
- VernacularTitle:肝细胞癌患者RFA术后局部肿瘤进展预测模型的构建
- Author:
Hongfang WANG
1
;
Guanhua YANG
;
Minglei WANG
;
Ziyu WANG
;
Yong CHEN
Author Information
1. 宁夏医科大学第一临床医学院，银川 750004
- Publication Type:Journal Article
- Keywords:
Carcinoma, hepatocellular;
Radiofrequency ablation;
Magnetic resonance imaging;
Local tumor progression;
Radiomics
- From:
Chinese Journal of Hepatobiliary Surgery
2025;31(8):567-573
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To construct a prediction model for local tumor progression (LTP) in patients with hepatocellular carcinoma (HCC) after radiofrequency ablation (RFA) based on the radiomics features of enhanced MRI.Methods:Clinical data of 120 patients with HCC undergoing RFA in the General Hospital of Ningxia Medical University from June 2017 to June 2022 were retrospectively analyzed, including 90 males and 30 females, aged (58.2±8.2) years. The patients were divided into training set ( n=84) and validation set ( n=36) in a ratio of 7∶3. According to whether LTP occurred within 2 years after RFA, the patients in training set were divided into LTP positive group ( n=32) and LTP negative group ( n=52). Logistic regression analysis was performed to analyze the risk factors for LTP after RFA in patients with HCC in training set. In the advanced arterial phase of preoperative enhanced MRI, the region of interest of tumor and peritumoral 5 mm area were mapped, and the radiomics features were extracted. The maximum correlation-minimum redundancy algorithm, the minimum absolute value shrinkage and selection operator algorithm were used to screen the radiomics features closely related to LTP, and the radiomics score was established. A nomogram model was constructed by combining the radiomics score with clinical tumor characteristics. The predictive performance and clinical practical value of different models were compared by the area under the receiver operating characteristic curve, calibration curve, clinical decision curve analysis (DCA) and clinical impact curve (CIC). Results:Tumor located around the blood vessels ( OR=4.574, 95% CI: 1.454-14.393, P=0.009) and ablation margin <5 mm ( OR=5.724, 95% CI: 1.996-16.420, P=0.001) were independent risk factors for LTP in patients with HCC after RFA. Five higher-order radiomics features were extracted and screened, including three tumoral features (glrlm_ShortRunHighGrayLevelEmphasis, ngtdm_Complexity and glcm_Imc1) and two peritumoral features (firstorder_Mean and glszm_SmallAreaHighGrayLevelEmphasis). Delong test showed that the area under curve of the combined model was higher than that of the radiomics model ( Z=2.90, P=0.004) and the clinical tumor characteristic model ( Z=2.56, P=0.010). Calibration curves, DCA and CIC curves all show that the combined model had a better clinical net benefit. Conclusion:Combining the radiomics features extracted from enhanced MRI images with clinical tumor characteristics can effectively predict the risk of LTP in patients with HCC after RFA.
