Abstract Background: Real-time artificial intelligence-based computer-aided detection/diagnosis (AI-CAD) for breast ultrasound enables immediate assessment of suspicious lesions with probability of malignancy (POM) outputs, potentially aiding in the differentiation between ductal carcinoma in situ (DCIS) and invasive ductal carcinoma (IDC). Identifying ultrasonographic features with significant POM differences between DCIS and IDC may refine clinical triage. Methods: We retrospectively analyzed 34 cases (DCIS=12, IDC=22) assessed by a real-time AI solution (CadAI-B for Breast cancer) during breast ultrasound at a tertiary center. POM outputs and BI-RADS categorization were compared across ultrasonographic features including shape, margin, echo pattern, and orientation. Features demonstrating significant POM differences between DCIS and IDC were identified, and comparative analyses were performed to evaluate CadAI-B’s differential diagnostic utility. Results: Overall, IDC demonstrated a higher mean POM (0.365 ± 0.306) compared to DCIS (0.126 ± 0.196, p0.001). Ultrasonographic features with significant POM differences included irregular shape (IDC: 0.344 vs DCIS: 0.135, p=0.002), indistinct margins (IDC: 0.318 vs DCIS: 0.012, p=0.014), microlobulated margins (IDC: 0.419 vs DCIS: 0.132, p=0.031), hypoechoic pattern (IDC: 0.345 vs DCIS: 0.135, p=0.018), and parallel orientation (IDC: 0.362 vs DCIS: 0.044, p=0.003). These features were associated with larger POM differentials between IDC and DCIS, suggesting enhanced discrimination capacity. Under CadAI-B, IDC cases were more frequently categorized into higher BI-RADS categories, with IDC cases showing a substantial shift towards BI-RADS 4B, 4C, and 5 with high POM outputs. Notably, DCIS cases retained lower POM even when categorized into BI-RADS 4A and 4B. Conclusions: CadAI-B may help distinguish IDC from DCIS during real-time breast ultrasound examinations by reflecting differences in probability of malignancy. This may assist clinical decision-making during scanning by providing consistent, objective data when evaluating suspicious breast lesions. Further studies are needed to confirm its diagnostic value in broader clinical settings. Citation Format: J. Moon, J. Lee, B. Kang, W. Kim, J. Kim, J. Baek, H. Park, H. Kim, G. Baek. Using a Real-Time Artificial Intelligence Ultrasound System with Computer-Aided Detection and Diagnosis to Distinguish Ductal Carcinoma In Situ and Invasive Ductal Carcinoma abstract. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-04-09.
Moon et al. (2026) studied this question.