Skip to main content

Table 2 Imaging features selected for the construction of combined model

From: Prediction of prognosis of immune checkpoint inhibitors combined with anti-angiogenic agents for unresectable hepatocellular carcinoma by machine learning-based radiomics

Phase

Features number

Features selected

Importance

Arterial phase

12

wavelet.LLH_glszm_LargeAreaLowGrayLevelEmphasis

0.027912

wavelet.HLL_glcm_Autocorrelation

0.016868

wavelet.LLL_glszm_GrayLevelNonUniformity

0.018511

wavelet.HHL_firstorder_Maximum

0.01982

wavelet.LLH_glcm_MaximumProbability

0.020024

wavelet.LHH_glcm_ClusterProminence

0.008468

wavelet.HLH_gldm_LargeDependenceHighGrayLevelEmphasis

0.011627

wavelet.LLL_glcm_Imc2

0.010737

wavelet.HHH_firstorder_Median

0.005421

wavelet.HLH_glszm_LargeAreaLowGrayLevelEmphasis

0.00187

wavelet.LLL_glrlm_GrayLevelNonUniformity

0.017364

wavelet.LHH_glcm_Imc2

0.002767

Portal venous phase

10

wavelet.LLL_glcm_Imc2

0.177879

wavelet.HHH_firstorder_Mean

0.007811

log.sigma.3.0.mm.3D_glcm_ClusterShade

0.013591

wavelet.HLH_glszm_LargeAreaLowGrayLevelEmphasis

0.002256

wavelet.LLH_firstorder_Kurtosis

0.006625

wavelet.LLL_glszm_LargeAreaHighGrayLevelEmphasis

0.000742

original_glszm_LargeAreaHighGrayLevelEmphasis

0.003997

log.sigma.5.0.mm.3D_glcm_Imc2

0.002265

wavelet.LLL_gldm_GrayLevelNonUniformity

0.003511

wavelet.LHH_glrlm_ShortRunEmphasis

0.003647

Delayed phase

10

wavelet.LLL_glcm_Imc2

0.030602

wavelet.LLH_firstorder_Kurtosis

0.006942

log.sigma.3.0.mm.3D_glcm_ClusterShade

0.014806

log.sigma.5.0.mm.3D_firstorder_90Percentile

0.00555

wavelet.HHH_glcm_Imc2

0.003582

wavelet.HLL_glcm_Correlation

0.010021

wavelet.HHH_glszm_LargeAreaLowGrayLevelEmphasis

0.006388

wavelet.LLL_gldm_LargeDependenceHighGrayLevelEmphasis

0.007395

log.sigma.3.0.mm.3D_gldm_SmallDependenceHighGrayLevelEmphasis

0.001644

wavelet.LLL_glszm_GrayLevelNonUniformity

0.01647

  1. Wavelet Transform Features: These features are extracted using wavelet transform, a mathematical tool that allows for the analysis of data at different scales and orientations. “Wavelet. LLH_glszm_LargeAreaLowGrayLevelEmphasis” specifically measures the emphasis of large areas with low gray-level intensities within the tumor, which may indicate regions of necrosis or hypoxia and are associated with tumor aggressiveness and treatment resistance
  2. GLSZM (Gray Level Size Zone Matrix) Features: Derived from the gray level size zone matrix, these features describe the distribution of pixel intensities within zones of uniform intensity. They provide information about the spatial arrangement of tissue characteristics within the tumor and can reflect tumor heterogeneity
  3. GLCM (Gray Level Co-occurrence Matrix) Features: Calculated from the gray level co-occurrence matrix, these features quantify the spatial relationship between pixel intensities. Features such as “Wavelet. HLL_glcm_Autocorrelation” reflect the similarity of pixel values across the image, indicating the uniformity of the tumor texture
Back to article page

BMC Cancer

ISSN: 1471-2407

Contact us