Giá trị của phân loại BI-RADS trên nhũ ảnh cắt lớp trong chẩn đoán ác tính các tổn thương vú ở bệnh nhân có mô vú đặc

Abstract

This analytical cross-sectional study was conducted in 91 patients with 103 breast lesions in dense breasts to evaluate the diagnostic performance of BI-RADS classification on digital breast tomosynthesis (DBT) for malignancy, using histopathological findings as the reference standard. The results showed that malignancy rates increased progressively with higher BI-RADS categories; BI-RADS 1–3 lesions were predominantly benign, whereas malignancy rates increased from BI-RADS 4A to 4C and were highest in BI-RADS 5 (p for trend < 0.001). Across the evaluated thresholds, BI-RADS ≥ 4A demonstrated the highest sensitivity (0.976), while BI-RADS ≥ 5 showed the highest specificity and positive predictive value, at 0.984 and 0.964, respectively. The BI-RADS ≥ 4B threshold yielded the highest accuracy and AUC in the study sample (0.871). BI-RADS classification on DBT is valuable for malignancy risk stratification in breast lesions among patients with dense breasts. In this context, the BI-RADS ≥ 4A threshold is clinically meaningful for minimizing missed malignant lesions, whereas the BI-RADS ≥ 4B threshold should be regarded only as a statistical reference point and should not replace clinical assessment or biopsy indications according to current BI-RADS recommendations.