Predicting response to CCRT for esophageal squamous carcinoma by a radiomics-clinical SHAP model

BMC Medical Imaging

Table 4 Rank of the top radiomics features in frequency through sample bootstrapping

Rank	Radiomics features	Frequency
CTV
1	wavelet-HLH_firstorder_Mean	247
2	original_shape_MajorAxisLength	98
3	wavelet-LLL_glcm_Imc1	92
4	square_glcm_InverseVariance	90
5	wavelet-HLH_firstorder_Median	84
6	log-sigma-1–0-mm-3D_gldm_DependenceVariance	72
7	wavelet-HHH_firstorder_Skewness	63
8	wavelet-LHH_gldm_DependenceVariance	57
9	gradient_firstorder_10Percentile	51
10	squareroot_glrlm_LongRunLowGrayLevelEmphasis	39
GTV
1	wavelet-LHH_gldm_LargeDependenceHighGrayLevelEmphasis	151
2	original_shape_Elongation	112
3	logarithm_gldm_LargeDependenceHighGrayLevelEmphasis	95
4	square_gldm_LargeDependenceHighGrayLevelEmphasis	92
5	gradient_glszm_SizeZoneNonUniformityNormalized	65
6	wavelet-HHH_glcm_Imc1	63
7	gradient_firstorder_Minimum	59
8	wavelet-LHL_ngtdm_Contrast	55
9	original_shape_Flatness	45
10	wavelet-HHL_ngtdm_Strength	45

ISSN: 1471-2342