Impact of real-time artificial intelligence ultrasound system based on breast density in C4 breast lesions.

Abstract

e12549 Background: Artificial intelligence–based computer-aided diagnosis (AI-CAD) systems are increasingly used in breast ultrasonography; however, their diagnostic performance may vary according to breast density. Given that dense breasts are highly prevalent among Asian women, understanding this relationship is important for optimizing AI-assisted imaging strategies. This study evaluated how breast density affects the diagnostic accuracy of an AI-CAD ultrasound system in BI-RADS category 4 (C4) breast lesions. Methods: 110 consecutive BI-RADS C4 lesions were assessed by board-certified breast radiologists between January and December 2023. CadAI-B (BeamWorks Inc., Daegu, Republic of Korea) automatically generated BI-RADS categories and probability of malignancy (POM) values using static ultrasound images. Pathology served as the reference standard, with atypia and malignancy grouped as non-benign. Diagnostic performance—sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy—was analyzed by breast density (BI-RADS B–D), determined using AI-assisted mammography. Results: Overall sensitivity and NPV were 81.3% and 87.5%, whereas specificity and PPV were lower (53.8% and 41.9%). All diagnostic indices improved with increasing breast density. In the density D group, sensitivity (92.3%), specificity (61.5%), NPV (96.0%), and accuracy (69.2%) were highest. Concordance between AI-assigned BI-RADS category and pathological diagnosis also increased with density (B: 50.0%, C: 57.5%, D: 67.3%). Non-benign lesions consistently showed higher POM values across all densities. Conclusions: Breast density significantly influences the diagnostic performance of AI-CAD ultrasound in BI-RADS C4 lesions. The AI system demonstrated superior accuracy and concordance in dense breasts, suggesting enhanced interpretive stability in high-density environments. Diagnostic performance of AI ultrasound in C4 breast lesions based on breast density. Breast density Total B C D BI-RADS (n, %) C1/2 C3 ≥C4 Total C1/2 C3 ≥C4 Total C1/2 C3 ≥C4 Total Disease category Benign 78 (70.9) 2 (100.0) 2 (50.0) 7 (58.3) 11 (61.1) 10 (90.9) 4 (66.7) 14 (60.9) 28 (70.0) 12 (92.3) 12 (100.0) 15 (55.6) 39 (75.0) Atypia 7 (6.4) 0 (0.0) 1 (25.0) 0 (0.0) 1 (5.6) 0 (0.0) 1 (16.7) 2 (8.7) 3 (7.5) 1 (7.7) 0 (0.0) 2 (7.4) 3 (5.8) Malignancy 25 (22.7) 0 (0.0) 1 (25.0) 5 (41.7) 6 (33.3) 1 (9.1) 1 (16.7) 7 (30.4) 9 (22.5) 0 (0.0) 0 (0.0) 10 (37.0) 10 (19.2) Total 110 2 4 12 18 11 6 23 40 13 12 27 52 Sensitivity (%) 81.3 72.4 75.0 92.3 Specificity (%) 53.8 36.4 50.0 61.5 Positive predictive value (%) 41.9 41.7 39.1 44.4 Negative predictive value (%) 87.5 66.7 82.4 96.0 Accuracy (%) 61.8 50.0 52.5 69.2