Skip to main content

Table 4 Top 10 features in each of the tumour detection models ranked from most to least important

From: Detecting localised prostate cancer using radiomic features in PSMA PET and multiparametric MRI for biologically targeted radiation therapy

Rank

PET model

mpMRI model

mpMRI + PET model

Image

Feature

Image

Feature

Image

Feature

1

PET

3D LoG σ = 3 mm minimum

–

PSA

PET

3D LoG σ = 3 mm minimum

2

–

Uptake time

ADC

NGTDM Coarseness

PET

Gradient Magnitude Energy

3

PET

SUVmax

ADC

10th percentile

PET

LBP 3D m = 1 maximum

4

–

PSA

TTP

10th percentile

ADC

NGTDM Coarseness

5

PET

LBP 3D m = 1 maximum

ADC

GLCM Correlation

ADC

10th percentile

6

PET

SUV

ADC

Entropy

PET

SUVmax

7

PET

Gradient Magnitude Energy

–

Age

ADC

GLCM Correlation

8

–

Age

ADC

GLDM LDLGLE

ADC

Entropy

9

–

Injected activity

TTP

90th percentile

–

PSA

10

–

Grade Group other

TTP

GLDM LDLGLE

PET

GLDM LDLGLE

  1. LoG Laplacian of Gaussian, LBP Local binary pattern, NGTDM Neighbouring Grey Tone Difference Matrix, GLDM Grey Level Dependence Matrix, LDLGLE Large Dependence Low Grey Level Emphasis