Habib Rahbar1,2, Savannah Partridge1,2, Wendy DeMartini1,2, Franklin Liu1,2, Robert Gutierrez1,2, Kimberly Allison3, Sue Peacock1,4, Constance Lehman1,2
1Radiology, University of Washington, Seattle, WA, USA; 2Radiology, Seattle Cancer Care Alliance, Seattle, WA, USA; 3Pathology, University of Washington, Seattle, WA; 4Radiology, Seattle Cancer Care Alliance, Seattle , WA, USA
DCIS is a heterogeneous malignancy with variable invasive potential. There are currently no imaging models to assess in vivo DCIS grade. We investigated the feasibility to assess DCIS grade in 55 consecutive pure DCIS lesions using both univariate and multivariate logistic regression modeling based on DCE and DWI MRI characteristics. We found that DCIS grade could be accurately differentiated by both DWI and DCE features and that a model incorporating maximum lesion size and DWI CNR best discriminated DCIS by grade. Our findings suggest that MRI has the capability to contribute to risk stratification of DCIS.