Skip to main content
. 2019 Dec 6;9:1338. doi: 10.3389/fonc.2019.01338

Table 2.

Selected features using distance correlation, LASSO, and GBDT.

Selection method Selected features
Distance Correlation HISTO_Kurtosis, HISTO_Entropy, HISTO_Energy, SHAPE_Volume,
GLCM_Energy, GLCM_Entropy_log10, NGLDM_Contrast, GLZLM_ZLNU
LASSO minValue, meanValue, stdValue, SHAPE_Volume (ml),
GLCM_Contrast, GLRLM_HGRE, GLRLM_LRHGE,
GLRLM_GLNU, GLRLM_RLNU, GLZLM_LZE, GLZLM_HGZE,
GLZLM_SZHGE, GLZLM_LZHGE, GLZLM_GLNU, GLZLM_ZLNU
GBDT minValue, HISTO_Skewness, SHAPE_Volume (ml),
GLCM_Homogeneity, GLCM_Energy, GLCM_Correlation,
GLCM_Entropy_log10, GLCM_Dissimilarity, GLRLM_LRLGE,
GLRLM_RLNU, NGLDM_Contrast, GLZLM_SZE, GLZLM_LZHGE,
GLZLM_ZLNU

LASSO, least absolute shrinkage and selection operator; GBDT, gradient boosting decision tree.