Table 2.

Features selected for predicting OS from CT images (N = 108).

Feature No.	Imaging modality	VOI of feature extraction	Filter type	Feature class	Statistic	Coefficients*
F1	Portal venous phase	Tumor	Wavelet_LLL	GLCM	IMC1	−0.1487
F2	Portal venous phase	Tumor	Wavelet_LLL	GLCM	IMC2	−0.0177
F3	Portal venous phase	Tumor	Wavelet_HLL	GLRLM	SRLGLE	−0.0282
F4	Arterial phase	Tumor	Wavelet_LHL	GLRLM	SRLGLE	−0.0600
F5	Arterial phase	Peritumoral region	Log.sigma.1.0.mm	GLDM	DNN	−0.1651
F6	Arterial phase	Peritumoral region	Wavelet_LHL	GLSZM	GLNN	−0.0571

OS, Overall survival; VOI, Volume of interest; GLCM, Gray Level Co-occurrence Matrix; GLRLM, Gray Level Run Length Matrix; GLCM, Gray Level Co-occurrence Matrix; GLDM, Gray Level Dependence Matrix; GLSZM, Gray Level Size Zone Matrix. IMC, Informational Measure of Correlation; SRLGLE, Short_Run_Low_Gray_Level_Emphasis; DNN, Dependence Non-Uniformity Normalized; GLNN, Gray Level Non-Uniformity Normalized.

^*Coefficients were derived from the LASSO Cox regression. Formula of the radiomics signature was as follows: radiomics signature = IMC1 × −0.1487 + IMC2 × -0.0177 + SRLGLE × −0.0282 + SRLGLE × −0.0600 +DNN × −0.1651 + GLNN × −0.0571.