Table 3 Numerical values of key radiomic features in arterial, venous, and combined analyses.

From: Preoperative arterial and venous CT radiomics for survival prediction after pylorus preserving pancreatoduodenectomy in pancreatic head cancer

Most important arterial features

Feature importance scores

Shape: sphericity

0.26

Gray level size-zone non-uniformity (GLSZM): size-zone non-uniformity

0.18

Shape: surface volume ratio

0.07

Gray level run length matrix (GLRLM): long run low gray level emphasis

0.05

Shape: maximum three-dimensional (3D) diameter

0.05

GLSZM: gray level non-uniformity

0.03

Shape: maximum two-dimensional (2D) diameter slice

0.02

Gray level co-occurrence matrix (GLCM): cluster shade

0.02

First order: 90th percentile

0.02

First order: 10th percentile

0.02

GLSZM: size-zone non-uniformity

0.13

Shape: major axis length

0.09

First order: mean

0.07

Shape: surface volume ratio

0.06

GLSZM: zone entropy

0.05

Shape: flatness

0.05

Shape: sphericity

0.04

First order: kurtosis

0.03

First order: 90th percentile

0.03

GLCM: cluster shade

0.03

Shape: sphericity (arterial)

0.25

GLSZM: size-zone non-uniformity (arterial)

0.12

Shape: surface volume ratio (arterial)

0.06

First order: total energy (venous)

0.04

Shape: maximum 3D diameter (arterial)

0.04

GLRLM: long run low gray level emphasis (arterial)

0.03

GLCM: inverse difference normalized (venous)

0.03

Shape: major axis length (arterial)

0.02

GLSZM: gray level non-uniformity (arterial)

0.02

GLSZM: size-zone non-uniformity (venous)

0.02

  1. Per analysis, the 10 most important radiomic features and their corresponding feature importance scores are indicated. The features are indicated as “Feature group: feature name”.