Skip to main content

Table 2 Overview of the image filters

From: The effect of preprocessing filters on predictive performance in radiomics

Filter

Parameters

N

Original

105

Exponential

91 (196)

Gradient

91 (196)

Laplacian-of-Gaussian

Sigma 1.0, 2.0, 3.0, 4.0, 5.0 mm

455 (560)

Local binary pattern

Two-dimensional, three-dimensional

273 (378)

Logarithm

91 (196)

Square

91 (196)

Square root

91 (196)

Wavelet

Directions HHH, HHL, HLH, HLL, LHH, LHL, LLH, LLL

728 (833)

  1. Original denotes the set of extracted features without any image processing, which also contains of shape features. Parameters denote the chosen extraction parameters for the preprocessing filter, if applicable. HHH, HHL etc. denote the frequency components of the wavelet filters (H = high, L = low). N denotes the number of features that were generated after applying the filter; in parenthesis is the sum of the number of original features and the features after applying the corresponding filter