Automatic Segmentation of Lung Carcinoma Using 3D Texture Features in 18-FDG PET/CT

Figure 5

Summary of performance by area under the receiver operation characteristic (ROC) curve (AUC) when using individual features to discern a given tissue from the tumor volume. The 𝑘 value of the KNN classifier was used to vary sensitivity and specificity and plot the curve. Lighter areas indicate better discrimination. Note that these data refer to discrimination of tumor ROIs from normal tissue ROIs, not individual voxel discrimination.