Riham H. El Khouli MD, PhD, Presenter: Nothing to Disclose

Katarzyna J. Macura MD, PhD, Abstract Co-Author: Nothing to Disclose

Ihab R. Kamel MD, PhD, Abstract Co-Author: Nothing to Disclose

David A. Bluemke MD, PhD, Abstract Co-Author: Research support, Siemens AG

Michael Anthony Jacobs PhD, Abstract Co-Author: Nothing to Disclose

To assess the value of multiparametric breast MRI (including morphology, DCE MRI and DWI with Apparent Diffusion Coefficient (ADC) mapping) at 3T in distinguishing among DCIS, Luminal A and B, HER2 positive, and Triple Negative breast cancer phenotypes

Our institutional review board approved the study. We included 219 patients with 234 lesions patients who underwent bilateral breast MRI at 3T (mean age 53+11.5 year). Both high temporal (15 sec) DCE and high spatial resolution (0.5 mm2 voxel size) MRI were acquired along with DWI with ADC mapping. Regions of interest were drawn on the ADC maps of breast lesions and normal appearing glandular tissue (GT). Morphologic features, DCE-MRI results (kinetic curve type), GT and lesion absolute and normalized ADC values were included in multivariate models for prediction of breast cancer histological subtypes. Area under ROC curve analysis was performed

Of 234 breast cancer lesions, 12% of were DCIS, 47% Luminal A, 22.2% Luminal B, 4.3% HER2 positive, and 14.5% triple negative. Lesion morphology (combining type of lesion with margin/distribution), Kinetic curve type, time to peak enhancement, and both absolute and normalized ADC values were univariate predictors of breast cancer phenotypes with an AUC 0.61-0.79. Combining lesion volume, morphology, kinetic curve type, internal enhancement, and normalized ADC value showed the best accuracy in predicting estrogen receptor expression, while combining lesion diameter, morphology and ADC value showed the best diagnostic accuracy in predicting progesterone receptors expression, and combining lesion diameter, morphology, and normalized ADC value showed the best accuracy in predicting the HER2 receptor expression. For the phenotypes characterization, the multivariate diagnostic model combining lesion morphology, kinetic curve type, and normalized ADC value showed the best diagnostic accuracy (AUC 0.83)

Multiparametric MRI including morphology, DCE and DWI can characterize breast cancer phenotypes with a very good diagnostic accuracy (AUC =0.83) at 3T

Breast cancer tumors with the same histological characteristic may carry different prognosis and response to treatment due to the difference at the molecular level. In vivo identification of different breast cancer phenotypes can improve our ability to detect more aggressive regions within the tumor and evaluate treatment response

El Khouli, R, Macura, K, Kamel, I, Bluemke, D, Jacobs, M, Prediction of Breast Cancer Phenotypes Using Multiparametric MRI of the Breast with Dynamic Contrast Enhancement and Diffusion Weighted Imaging at 3T.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14015707.html