FDG-PET texture analysis in breast cancer according to molecular subtypes.

Objectives Breast cancer (BC) is the most common cancer in women in western countries. Previous studies have shown a significant correlation of SUV and more recently textural indices (TI) in BC to histological data mainly SBR grade, Triple negative (TN) status and Ki 67 labelling index (LI). However, an analysis of TI according to BC molecular subtypes is of interest as it follows a more clinical approach of BC.

Methods A cohort of women with BC who underwent a F18-FDG-PET/CT in our institution (GE Discovery LS scanner) was included in this study. SUV conventional indices and TI (homogeneity, entropy, LRE, SRE, LGZE and HGZE) were assessed in a tumor volume of interest (VOI) defined using a threshold (40% of SUVmax). Histological data were obtained using the initial diagnostic biopsy for each tumor. SUV and textural indices were compared as a function of SBR grade (II versus III), TN status, Ki 67 LI (20% cut off value), histological type and BC molecular subtypes according to the 2010 international classification. Other histological data such as the presence of carcinoma in situ (CIS), necrosis, vascular emboli and an inflammatory stroma were also analyzed. Data analysis was performed using a Kruskal-Wallis test for molecular subtypes and a Mann-Whitney test was used otherwise.

Results Fifty four women with different stages of BC and a total of 68 tumors VOI with histologically proven breast cancers were included. Significant differences in TI between tumors VOI according to the histological type, SBR grade, TN status, Ki 67 LI, the presence of CIS and BC subtypes were found (table 1). No difference related to the presence of necrosis, vascular emboli or an inflammatory stroma was found (p > 0.05). 

Conclusions Our study showed significant differences in TI for various histological data in BC especially for some breast cancer molecular subtypes and the presence of in situ carcinoma and sometimes with more significance than SUVmax or SUVpeak. Those findings suggest that TI might enrich the characterization of breast lesions from PET images, although interpretation rules for prospective studies still have to be identified.