Table 1.
Patient and breast carcinoma characteristics (n = 54).
Table 2.
Textural and SUV indices as a function of pathological characteristics of breast cancer at univariate logistic regression.
Figure 1.
Illustration of tumor heterogeneity at FDG-PET/CT in the different subtypes of breast.
Axial PET fusion images (left) and 3D-view of 3 orthogonal plans from the tumor volume (right) after voxel resampling are displayed. Two histologic types of breast tumor are displayed: triple negative (top) and luminal B (bottom) tumors. Both tumors exhibit intense FDG uptake with a central hypometabolic area. The triple negative breast tumor exhibits higher SUVmax and higher textural heterogeneity than the luminal B tumor (right). This example illustrates higher FDG uptake and higher texture heterogeneity in TNBC compared to non-TNBC.
Figure 2.
Axial PET fusion images (left) and 3D-view of 3 orthogonal plans from the tumor volume (right) after voxel resampling are displayed.
The triple negative breast tumor (top) exhibits lower SUVmax but higher textural heterogeneity than the HER-2 positive breast tumor (bottom). This example illustrates the ability of the HGRE textural index to identify higher heterogeneity despite lower FDG uptake in triple negative breast tumors compared to the non-triple negative breast tumors.
Figure 3.
ROC curves analysis. AUC of SUVmax, SUVmax associated with homogeneity and SUVmax with HGRE for identifying TNBC.
Figure 4.
Plots of the patients with TNBC (left) and non-TNBC (right) illustrating the effect of including a texture parameter (HGRE) in addition to SUVmax for classification of breast cancer.
In both tumor groups (TNBC and non-TNBC), adding HGRE to SUVmax improves the classification of the tumor if the patient is located above the diagonal. The probability of correct classification increased in 77% of TNBC (10/13) and in 71% (29/41) of non-TNBC (NRI = 0.95, p = 0.003). A perfect combination of indices would reclassify 100% of patients above the diagonal (NRI equal to 2).