OBJECTIVES: To determine the predictive values and the interobserver variability of the descriptors and diagnostic categories of the BI-RADS- Ultrasound system and its usefulness for predicting malignancy in solid breast nodules. MATERIAL AND METHODS: We evaluated 601 consecutive solid nodules in 554 patients studied with ultrasound. All ultrasound examinations were performed by one of the three radiologists that participated in the study and the static images were reviewed by all three radiologists independently; radiologists were blind to the clinical history and to the findings at mammography and at histological study. RESULTS: The descriptors that best predicted benignity were circumscribed margins and oval shape (NPV, > 96%), parallel orientation (NPV, 84%-91%), and abrupt interface (NPV, 81%-90%). The descriptors that best predicted malignancy were spiculated margins (PPV, 77%-85%), echogenic halo (PPV, 61%-71%), and nonparallel orientation (PPV, 53%-54%). Interobserver concordance was good for lesion shape (kappa=0.61), circumscribed margins (kappa=0.65), and calcifications (kappa=0.63). The descriptors that presented the highest prognostic values for malignancy were spiculated margins (OR=14.68-10.45) and nonparallel orientation (OR=3.95-6.17). Final assessment category 3 yielded an excellent NPV for all three radiologists (99%-100%). The interobserver concordance was good for category 5 (kappa=0.77) and for category 3 (kappa=0.68); it was moderate for category 4 (kappa=0.59). CONCLUSIONS: The BI-RADS- Ultrasound system helps differentiate benign breast lesions from malignant ones and has a good or moderate interobserver concordance. Final assessment category 3 yielded an excellent negative predictive value, making it possible to avoid biopsies in lesions in this category.
OBJECTIVES: To determine the predictive values and the interobserver variability of the descriptors and diagnostic categories of the BI-RADS- Ultrasound system and its usefulness for predicting malignancy in solid breast nodules. MATERIAL AND METHODS: We evaluated 601 consecutive solid nodules in 554 patients studied with ultrasound. All ultrasound examinations were performed by one of the three radiologists that participated in the study and the static images were reviewed by all three radiologists independently; radiologists were blind to the clinical history and to the findings at mammography and at histological study. RESULTS: The descriptors that best predicted benignity were circumscribed margins and oval shape (NPV, > 96%), parallel orientation (NPV, 84%-91%), and abrupt interface (NPV, 81%-90%). The descriptors that best predicted malignancy were spiculated margins (PPV, 77%-85%), echogenic halo (PPV, 61%-71%), and nonparallel orientation (PPV, 53%-54%). Interobserver concordance was good for lesion shape (kappa=0.61), circumscribed margins (kappa=0.65), and calcifications (kappa=0.63). The descriptors that presented the highest prognostic values for malignancy were spiculated margins (OR=14.68-10.45) and nonparallel orientation (OR=3.95-6.17). Final assessment category 3 yielded an excellent NPV for all three radiologists (99%-100%). The interobserver concordance was good for category 5 (kappa=0.77) and for category 3 (kappa=0.68); it was moderate for category 4 (kappa=0.59). CONCLUSIONS: The BI-RADS- Ultrasound system helps differentiate benign breast lesions from malignant ones and has a good or moderate interobserver concordance. Final assessment category 3 yielded an excellent negative predictive value, making it possible to avoid biopsies in lesions in this category.