Jeff Wang, Presenter: Nothing to Disclose

Fumi Kato, Abstract Co-Author: Nothing to Disclose

Kohsuke Kudo MD, Abstract Co-Author: Nothing to Disclose

Hiroko Yamashita, Abstract Co-Author: Nothing to Disclose

Hiroki Shirato MD, PhD, Abstract Co-Author: Nothing to Disclose

Despite many efforts having studied lesion texture as imaging biomarkers of breast cancer (BC) subtypes, it appears none have yet been published assessing the same of background parenchymal enhancement (BPE). This study aims to determine the prognostic ability of BPE texture surrogates with molecular subtypes of BC.

Building evidence continues to show BC is a diverse disease. Molecular subtyping based on estrogen (ER), progesterone (PgR), and human epidermal growth factor 2 (HER2) receptor expression provides valuable information for treatment. Dynamic contrast-enhanced (DCE)-MRI is standard in diagnostic breast imaging, known for its high sensitivity. Increased BPE on DCE-MRI has been associated with higher rates of abnormal interpretation and obscured breast masses. There is also evidence it may provide insight with BC risk. This retrospective study included 64 women with 69 invasive mass carcinomas, who had DCE-MRI. ER, PgR, and HER2 receptor expression of the lesions were determined by immunohistochemistry in specimens. The cancers were also categorized triple-negative (TN) or Luminal A (LumA), as clinically significant. Segmentation of parenchyma tissue was performed from DCE-MRI of the affected breast and BPE texture was then quantified as first and second-order statistical features of pharmacokinetic parameter maps calculated from the tissue compartment. Logistic regression models were learned, using reduced BPE texture features to classify receptor status. Accuracy (ACC), sensitivity (TPR), specificity (TNR), and area under the ROC curve (AUC) of performance were calculated from leave-one-out cross-validation.

TN BC were classified with ACC of 95%, TPR of 89%, TNR of 97%, and AUC of 0.89. ER BC were classified with ACC of 88%, TPR of 67%, TNR of 96%, and AUC of 0.81. PgR BC were classified with ACC of 68%, TPR of 42%, TNR of 86%, and AUC of 0.61. HER2 BC were classified with ACC of 83%, TPR of 36%, TNR of 94%, and AUC of 0.63. LumA BC were classified with ACC of 61%, TPR of 65%, TNR of 57%, and AUC of 0.66.

BPE texture is demonstrated as able to predict TN and ER BC with great accuracy and discriminative ability; PgR, HER2, and LumA BC to lesser degrees.

BPE heterogeneity can extend the diagnostic ability of DCE-MRI, as it is strongly predictive of some molecular subtypes of breast cancer, particularly the more aggressive triple-negative subtype.

Wang, J, Kato, F, Kudo, K, Yamashita, H, Shirato, H, Heterogeneity of Background Parenchymal Enhancement on MRI Strongly Predictive of Breast Cancer Molecular Subtypes.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14010844.html