The Breast Imaging Reporting and Data System (BI-RADS) terminology employed by radiologists to classify breast imaging results is useful in predicting malignancy in breast lesions detected with magnetic resonance imaging (MRI), a recent study demonstrated.
BI-RADS, published by the American College of Radiology in collaboration with other health care organizations, is a quality assurance tool used to standardize reporting for breast imaging studies. Initially developed for mammography, the system was expanded in 2003 to include both MRI and ultrasound breast imaging. The imaging studies are assigned an assessment of 0 to 6 based on the radiologist’s interpretation of the images and his or her characterization of any lesions noted.
To evaluate the positive predictive values of BI-RADS assessment categories for breast MRI and to identify the BI-RADS MRI lesion features most predictive of cancer, a team led by Mary C. Mahoney, MD, director of breast imaging at the University of Cincinnati (Ohio) Medical Center, conducted a multicenter study involving 969 women. The participants had recently received a breast cancer diagnosis in one breast and had undergone MRI on the other breast at one of 25 participating imaging sites.
As the investigators reported in the journal Radiology, the analysis showed that a BI-RADS assessment of 5, defined as highly suggestive of malignancy, and the identification of masses (three-dimensional lesions) with irregular shape, irregular margins, spiculated margins (margins characterized by spikes or points), and marked internal enhancement (a very bright image with contrast agent) were most predictive of malignancy. For lesions that were not three-dimensional, known as nonmasslike enhancements (NMLEs), location in a milk duct or clumped enhancement were the features most frequently seen with malignancy. Morphologic features were more predictive of malignancy than were kinetic enhancement features.
Mahoney and colleagues concluded that BI-RADS assessment categories and morphologic descriptors for masses and NMLEs were useful in estimating the probability of cancer.