Skip to main content
. 2020 Nov 11;20:81. doi: 10.1186/s40644-020-00359-2

Table 3.

The selected feature sets with LASSO and LASSO with SVMSMOTE

Selected Features
LASSO (n = 22) original_shape_LeastAxis
original_shape_Elongation
original_shape_Flatness
logarithm_firstorder_Kurtosis
logarithm_glrlm_HighGrayLevelRunEmphasis
square_firstorder_10Percentile
square_glrlm_ShortRunHighGrayLevelEmphasis
exponential_glcm_Imc1
exponential_glrlm_LongRunEmphasis
exponential_glrlm_LongRunLowGrayLevelEmphasis
wavelet-LHL_firstorder_Skewness
wavelet-LHH_glcm_ClusterShade
wavelet-HLL_firstorder_Energy
wavelet-LLH_firstorder_Kurtosis
wavelet-LLH_glcm_ClusterProminence
wavelet-HHH_glszm_GrayLevelNonUniformity
wavelet-HHH_glszm_LowGrayLevelZoneEmphasis
wavelet-HHH_glszm_SmallAreaLowGrayLevelEmphasis
wavelet-HHL_firstorder_Skewness
wavelet-HHL_glrlm_ShortRunLowGrayLevelEmphasis
wavelet-LLL_firstorder_Kurtosis
wavelet-LLL_glcm_Correlation
LASSO with SVMSMOTE (n = 30) wavelet-HLL_firstorder_Energy
exponential_glrlm_LongRunEmphasis
wavelet-HHL_firstorder_Skewness
wavelet-HHL_glrlm_ShortRunLowGrayLevelEmphasis
wavelet-HLH_glrlm_RunPercentage
wavelet-LHL_firstorder_90Percentile
square_glrlm_LongRunEmphasis
logarithm_glrlm_ShortRunHighGrayLevelEmphasis
wavelet-LHL_glcm_MaximumProbability
wavelet-LLH_glszm_GrayLevelNonUniformity
wavelet-LHL_firstorder_Mean
wavelet-LHH_glcm_ClusterShade
wavelet-LHL_glszm_ZonePercentage
wavelet-HHL_glszm_GrayLevelVariance
original_shape_Elongation
wavelet-LLH_firstorder_10Percentile
square_glcm_Correlation
original_shape_Flatness
wavelet-HHH_glszm_LowGrayLevelZoneEmphasis
wavelet-LLL_glcm_Correlation
exponential_glrlm_ShortRunHighGrayLevelEmphasis
wavelet-LLL_firstorder_Kurtosis
wavelet-LHL_firstorder_InterquartileRange
wavelet-LLH_glcm_Contrast
wavelet-LLH_firstorder_Energy
wavelet-LLH_firstorder_Minimum
wavelet-LHL_glszm_ZoneEntropy
wavelet-HHH_glszm_GrayLevelNonUniformity
original_glszm_ZoneEntropy
original_shape_LeastAxis

Each feature is denoted as Filter_FeatureGroup_FeatureName and ‘Original’ indicates the radiomics features extracted from the original images without preprocessing