Table 2 The image features used in this study. For the column of “image modality”, the term “PET/CT” means the feature is calculated for both PET and CT
Feature | Image modality | Spatial dimension | Definition |
|---|---|---|---|
D short | PET/CT | 1D | Diagnostic feature, maximum short diameter of the axial section |
Area | PET/CT | 2D | Diagnostic feature, area of the axial section |
Volume | PET/CT | 3D | Diagnostic feature, volume of the lymph node |
CTmean | CT | 2D/3D | Diagnostic feature, mean CT value inside the lymph node |
CTcontrast | CT | 2D/3D | Diagnostic feature, the difference between CTmean and the mean CT value of a 2-mm-thick tissue layer surrounding the lymph node. |
SUVmean | PET | 2D/3D | Diagnostic feature, mean SUV inside the lymph node |
SUVmax | PET | 2D/3D | Diagnostic feature, max SUV inside the lymph node |
SUVstd | PET | 2D/3D | Diagnostic feature, standard deviation of SUV inside the lymph node |
1st-order texture features | PET/CT | 3D | Six texture features calculated based on the pixel intensity histogram, see the supplementary material of [22] for detailed definition |
2nd-order texture features | PET/CT | 2D | Nineteen texture features calculated based on gray-level co-occurrence matrix (GLCM), see the supplementary material of [22] for detailed definition |
High-order texture features | PET/CT | 3D | Five texture features calculated based on neighborhood gray-tone difference matrix (NGTDM) and 11 texture features calculated based on gray-level zone size matrix (GLZSM), see the supplementary material of [22] for detailed definition |