Table 2 List of computed features from T2-weighted images and ADC maps, and their association with prostate cancer aggressiveness.

From: Utility of T2-weighted MRI texture analysis in assessment of peripheral zone prostate cancer aggressiveness: a single-arm, multicenter study

Feature type

List of computed features

Correlation (ρ) with cancer grade group

Differentiation of low versus I/H aggressive cancers: p values

ADC

T2W

ADC

T2W

Intensity histogram

Kurtosis

− 0.15

0.14

0.133

0.283

Maximum

− 0.37*

− 0.23*

0.015

0.283

Mean

− 0.45*

− 0.34*

0.001

0.016

Median

− 0.44*

− 0.34*

0.002

0.015

Minimum

− 0.50*

− 0.46*

< 0.0001

< 0.0001

10th percentile

− 0.51*

− 0.39*

< 0.0001

0.002

25th percentile

− 0.47*

− 0.35*

0.001

0.006

75th percentile

− 0.41*

− 0.33*

0.005

0.027

90th percentile

− 0.40*

− 0.31*

0.006

0.048

Skewness

− 0.06

0.08

0.597

0.135

Standard deviation

0.14

− 0.04

0.152

0.422

Textural (GLCM & GLRLM) features

ASM

 

− 0.34*

 

0.043

Autocorrelation

 

− 0.02

 

0.597

Cluster Prominence

 

− 0.19

 

0.048

Cluster shade

 

0.02

 

0.422

Contrast

 

− 0.18

 

0.019

Correlation

 

0.03

 

0.597

Difference entropy

 

− 0.09

 

0.084

Difference variance

 

− 0.19

 

0.019

Dissimilarity

 

− 0.17

 

0.027

Entropy

 

0.38*

 

0.028

IMC 1

 

0.51*

 

< 0.0001

IMC 2

 

− 0.48*

 

< 0.0001

IDM

 

0.13

 

0.048

Maximum probability

 

− 0.32*

 

0.048

Sum average

 

− 0.02

 

0.495

Sum entropy

 

0.11

 

0.940

Sum variance

 

− 0.23*

 

0.027

Variance

 

− 0.25*

 

0.012

Grey level non-uniformity

 

0.52*

 

< 0.0001

HGLRE

 

− 0.06

 

0.364

LGLRE

 

− 0.18

 

0.345

Long run emphasis

 

0.17

 

0.046

Long run high GLE

 

− 0.02

 

0.651

Long run low GLE

 

− 0.14

 

0.557

RLNU

 

0.50*

 

0.001

Run percentage

 

− 0.17

 

0.046

Short run emphasis

 

− 0.15

 

0.051

Short run high GLE

 

− 0.07

 

0.293

Short run low GLE

 

− 0.18

 

0.330

  1. Bolded p values indicate features differed significantly between low- (grade group 1) and I/H-aggressive (grade group ≥ 2) prostate cancers.
  2. ASM angular second moment, IMC information measure of correlation, IDM inverse difference moment, HGLRE high grey level run emphasis, LGLRE Low grey level run emphasis, GLE grey level emphasis, RLNU run length non-uniformity, I/H intermediate/high, ρ spearman correlation coefficient between features and cancer grade groups.
  3. *Indicates feature correlated significantly with cancer grade group. The correlation coefficients were calculated for the entire grade groups (i.e. 1–5) and not for the dichotomized aggressive classes (i.e. grade groups 1 and ≥ 2).
Back to article page