The Breast Imaging Reporting and Data Systems (BI-RADS) MR lexicon helps quantify the likelihood of malignancy for MR-detected lesions, according to a study published online May 15 in Radiology.
Breast MRI provides high sensitivity, but it is somewhat hampered by relatively low specificity as malignant and benign lesions have overlapping features.
Mary C. Mahoney, MD, director of breast imaging at University of Cincinnati, and colleagues sought to determine the positive predictive value of BI-RADS categories for breast MR and determine features most predictive of malignancy.
The researchers prospectively enrolled 969 participants undergoing breast MR at 25 sites from April 1, 2003, to June 10, 2004. A total of 691 participants had BI-RADS category 1 or 2 lesions and required no further evaluation.
Of the remaining women, 106 had a category 3 finding, and one malignancy was diagnosed in this group for a positive predictive value (PPV) of 0.009.
A total of 97 women had a BI-RADS category 4 finding, and 72 women underwent biopsy, with 17 malignancies reported. The PPV was 0.205.
Another 14 women had category 5 results, with malignancies reported in 10 of these women, yielding a PPV of 0.714.
Of the 75 women with a category 0 result, malignancies were found in three women, for a PPV of 0.040.
Mahoney and colleagues reported that masses of irregular shape and those with irregular or spiculated margins had the highest PPVs for cancer. “By understanding which of these features is most predictive of malignancy, the specificity of breast MR may be improved,” the researchers wrote.
Ductal enhancement was the non-masslike enhancement most likely to represent cancer, and rapid or initial enhancement was more likely to represent cancer than slow enhancement. In addition, plateau and washout patterns of delayed enhancement were linked with cancer.
“In conclusion, the standardized BI-RADS lexicon for breast MR imaging provides descriptors and assessment categories that can be used to predict the likelihood of malignancy for breast MR lesions,” the researchers wrote.