Table 5 Summary of the studies on radiomics analysis of HCC
Authors | Objectives | Study | Significant features/model | Phase | Summary |
|---|---|---|---|---|---|
Oh et al. [69] | Predict tumor grade and DFS | CECT | SD, MPP and skewness | AP | AP based CCR model correlated well with tumor grade and DFS after resection |
S. Song et al. [70] | Differentiate hypervascular lesions | CECT | Histogram, GLCM and GLRLM | AP | AP features characterized hypervascular liver lesions |
Mokrane et al. [71] | Verify indeterminate liver nodules | CECT | Radiomic signature using KNN, SVM, and RF | AP and PVP | Machine-learning-identified feature diagnosed HCC in patients with indeterminate liver nodules |
Huang et al. [72] | Characterization of HCC based on gene expression | Gd-EOB-DTPA MRI | GLCM, GLRLM and GLSZM-based signature computed using SVM | AP, PVP, DP, and HBP | A radiomic model predicted DPHCC preoperatively |
Ma and Peng et al. [56] | Prediction of MVI | CECT | Radiomic signature computed with SVM and LASSO | PVP | CCR model was useful in preoperative and individualized prediction of MVI |
Yang et al. [41] | Prediction of MVI | Gd-EOB-DTPA MRI | Radiomic signature computed with LASSO | HBP, T1W and HBP T1 map | HBP T1W and HBP T1 maps radiomic signature were independent predictors of MVI |
Zhu et al. [58] | Preoperative prediction of MVI | MRI | Uniformity, CP, CS and LRLGLE in CCR | AP | CCR model predictive of MVI |
Zhang et al. [59] | Prediction of ER | Gd-EOB-DTPA MRI | Histogram, GLCM, HGLRE in CCR computed with LASSO | T2W, AP, HBP | CCR had a better predictive ability of ER |
Zhou et al. [60] | Prediction of ER | CECT | Histogram and GLCM radiomic signature computed with LASSO | AP, PVP | AP and PVP based CCR was a significant predictor of ER |
Zhang et al. [22] | Prediction of ER | MRI | Uniformity, entropy, and skewness | AP | AP features were independent predictors of ER. |
Brenet Defour et al. [73] | Prediction of OS | CECT | Skewness | PVP | Skewness associated with OS and useful for selecting best candidates for resection. |
Zheng et al. [74] | Prediction of OS and TTR | CECT | GLCM radiomic signature computed with LASSO | AP | Low rad-score correlated with aggressive tumor phenotypes and predictive of postoperative outcome |
Song et al. [75] | Prediction of RFS | MRI | Histogram, GRLM, GLCM, GLSZM based signature computed with LASSO | PVP | Preoperative estimation of RFS |
Kim et al. [42] | Prediction of survival | CECT | Histogram, GLCM, GLSZM, and 2 shape-based features incorporated into CCR using LASSO | AP | A CCR nomogram performed better in survival prediction |
Fu et al. [63] | Treatment and prediction of TTP and OS | CECT | Gabor filter and wavelet transform | PVP | Appropriate selection of HCC’s for TACE plus sorafenib |
Kloth et al. [65] | Response assessment after TACE | CECT/pCT | Entropy, mean heterogeneity, uniformity, and skewness | AP/PVP | Significant correlation between texture features and pCT parameters in prediction of response |