Skip to main content
. 2019 Nov 25;14(11):e0225242. doi: 10.1371/journal.pone.0225242

Table 4. Radiomics features decided by Spearman’s rank correlation and backward stepwise elimination for use in multivariable regression analysis.

Classification of matrix Features selected by Spearman’s correlation Features selected by(Spearman’s correlation+backward stepwise elimination)
Conventional Indices

  • SUVmax

  • MTV

  • SUVmax
Indices based on intensity histogram

  • Skewness
Indices based on shape

  • Sphericity
GLCM based on gray level co-occurrence matrix

  • Correlation
NGLDM based on gray level neighborhood matrix

  • Contrast

  • Busyness
GLRLM based on gray level homogeneous run size matrix

  • GLNU
GLZLM based on gray level homogeneous zone size matrix

  • SZLGE

  • SZLGE