Table 2 Selected features for prediction of targetable pathways.

From: Imaging genotyping of functional signaling pathways in lung squamous cell carcinoma using a radiomics approach

Pathway

 

Clinical features

Global variables

Histogram-based features

Lung cancer-specific features

Shape features

Local features

Regional features

Emphysema features

Redox stress

Variables

 

Volume

Energy

 

Maximum 3D diameter

Cluster prominence

Size-zone variability

 
 

Odds ratio (95% CI)

 

1.64 (0.32–8.30)

0.15 (0.02–1.05)

 

0.88 (0.20–3.82)

1.28 (0.64–2.56)

2.16 (0.66–7.06)

 
 

p value

 

0.55

0.06

 

0.86

0.49

0.20

 

Apoptosis

Variables

 

Mass

Range

 

Maximum 3D diameter

Cluster shade

Intensity variability

 
 

Odds ratio (95% CI)

 

23.94 (0.54- > 999.99)

0.08 (0.01–0.82)

 

2.06 (0.16–26.46)

0.69 (0.13–3.80)

0.35 (0.06–2.26)

 
 

p value

 

0.10

0.03

 

0.58

0.67

0.27

 

Proliferation

Variables

  

HU at the 97.5th percentile

MPP

 

Cluster prominence

 

Right lung volume

 

Odds ratio (95% CI)

  

1.83 (0.69–8.61)

1.33 (0.23–7.56)

 

1.77 (0.82–3.81)

 

0.35 (0.13–0.90)

 

p value

  

0.44

0.75

 

0.15

 

0.03

Differentiation

Variables

Smoking pack-years

Minimum

  

Spherical disproportion

 

Intensity variability

 
 

Odds ratio (95% CI)

1.01 (0.99–1.04)

1.21 (0.59–2.47)

  

0.68 (0.33–1.43)

 

0.73 (0.32–1.68)

 
 

p value

0.26

0.60

  

0.31

 

0.46

 

Chromatin remodelers

Variables

Smoking pack-years

Energy

   

Maximum probability

  
 

Odds ratio (95% CI)

1.02 (0.99–1.04)

1.73 (0.94–3.19)

   

0.54 (0.28–1.05)

  
 

p value

0.11

0.08

   

0.07

  
  1. Data are features selected as variables using univariate analysis by category.
  2. Bold are features selected as independent predictive factors on multivariate analysis.
  3. CI, confidence limits; MPP, mean value of positive pixe.