Patients with mammographically visible or palpable findings underwent T1- and fat-saturated T2 weighted spin-echo and contrast agent-enhanced fat-saturated gradient-echo MR imaging. Patients underwent subsequent excisional biopsy for histopathologic confirmation. An interpretational model was constructed using 98 cases and was tested prospectively using 94 different cases. Sensitivity, specificity, predictive values, and receiver operating characteristics curves were computed for all models. Individual features with highly predictive values were MR visibility, enhancement degree and pattern, focal mass border characteristics, and focal mass internal septations. The model was built in a stepwise fashion to result in a decision tree in which individual architectural features occupied decision nodes. The diagnostic performance of all models was assessed and comprised assessment of the performance of each model's individual terminal nodes (positive predictive value PPV and negative predictive value NPV), as well as performance of each model as a whole. Perhaps the most valuable aspect of our interpretation model is its ability to identify hierarchical combinations of features with high NPV's for malignancy. We were interested in individual features and feature combinations with a greater than 95% likelihood of being benign. If similar results are seen in prospective validation studies performed in other clinical settings, biopsy may not be necessary in patients with findings that fall into these five feature categories. In addition, the categories of ductal and regional enhancement, in particular, could benefit from further subcategorization. Thus far in our investigation, ductal enhancement has been sufficiently highly associated with malignancy, that if a case shows ductal enhancement the case is considered suspicious for malignancy. A larger sample size will allow the construction of an even more detailed model.
Showing the most recent 10 out of 414 publications
