Contrast-Enhanced Cone-Beam Breast CT Features for Differentiation of Breast Lesions with a Ductal Carcinoma In Situ Component.

RATIONALE AND OBJECTIVE: Preoperative differentiation among ductal carcinoma in situ (DCIS), DCIS with microinvasion (DCISM), and DCIS with invasive ductal carcinoma (DCIS-IDC) remains challenging. We explored clinicopathological and contrast-enhanced cone-beam breast CT (CE-CBBCT) features associated with invasive status and developed exploratory stepwise models. MATERIALS AND METHODS: Patients with DCIS, DCISM, or DCIS-IDC undergoing preoperative CE-CBBCT from August 2018 to January 2026 were retrospectively included in the training cohort, whereas patients meeting the same inclusion criteria at the same institution from February 2026 to June 2026 comprised a temporally separated validation cohort. Clinicopathological and CE-CBBCT features were compared among groups. Logistic regression models were developed to distinguish DCIS from invasive lesions and DCISM from DCIS-IDC using CE-CBBCT features. RESULTS: 164 women were included in the training cohort and 58 in the validation cohort. In the training cohort, sentinel lymph node metastasis, human epidermal growth factor receptor 2 status, Ki-67 expression, and nuclear grade differed among groups (all P < 0.05). DCIS and DCISM differed in the maximum extent perpendicular to the ductal orientation (P = 0.014) and adjacent vessel sign (P = 0.005), whereas DCISM and DCIS-IDC differed in the degree of lesion enhancement and lesion type (both P < 0.001). The two models achieved areas under the receiver operating characteristic curves of 0.815 and 0.810 in the training cohort and 0.785 and 0.817, respectively, in the temporally separated validation cohort. CONCLUSION: CE-CBBCT features varied with invasive status, and the exploratory models showed preliminary discriminatory value. However, prospective multicenter external validation is needed before clinical implementation.