Table 1.
Patients’ characteristics.
Fig 1.
Example of ROI (Region of Interest) in 2D and 3D images for a patient in the adaptive group (lung window).
Fig 2.
Example of ROI (Region of Interest) in 2D and 3D images for a patient in the non-adaptive group (mediastinal window).
Fig 3.
The chart shows the occurrence percentage of each feature obtained during the feature selection procedure.
Due to the limited space, we report only those features selected in more than the 3% of the folds of the feature selection procedure for a total of 41 features. Moreover, we highlighted in bold style the names of the descriptors constituting the final signature. The different colors in the histogram represent the different features, as explained in the legend in the upper right corner, whilst a vertical line indicates the threshold used for defining the final signature.
Fig 4.
ROC curve of the proposed system.
Table 2.
Performance of the radiomic approach.
Table 3.
Leave-one-out cross validation and Bootstrap .632+ estimator errors per features set.