Diagnostic Efficacy of Multimodal Imaging in Non-mass-like Breast Lesions and Correlation with Pathological Findings

YUHAN WANG^1,2XIULIANG WEI²MEI WU^2*

• ¹Shandong University, Jinan, China
• ²The Second Hospital of Shandong University, Jinan, China

Objective: This study aimed to identify key imaging features related to malignancy in Non-mass breast lesions (NMLs) via ultrasonography, systematically evaluate the diagnostic performance of ultrasound (US), digital mammography (DM), and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). Additionally, it assessed the diagnostic independence and complementarity of these modalities within a multimodal framework to inform individualized imaging strategies. Methods: This retrospective study included female patients who underwent breast US between January 2017 and March 2025. .Imaging data (including US, DM, and DCE-MRI) and clinicopathological records were collected. Chi-square and Fisher's exact tests were used to analyze the association between US features and pathological outcomes, while multivariate binary logistic regression was employed to construct a predictive model. Receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) was calculated for each modality. Lastly, DeLong's test was used to compare diagnostic efficacy, and the clinical added value of multimodal diagnosis was evaluated. Results: A total of 235 female patients (mean age, 48.06 ± 12.46 years) with 240 ultrasound-defined non-mass-like breast lesions were included, comprising 115 benign (47.9%) and 125 malignant (52.1%) cases. Among the clinical and ultrasound manifestations, age ≥45 years, palpable mass, lesion diameter ≥1.5 cm, abundant blood flow and microcalcifications, structural distortion, and abnormal axillary lymph nodes demonstrated a strong correlation with the risk of malignancy of non-massive lesions of the breast in both univariate and multivariate analyses. Significant differences in diagnostic performance metrics (accuracy, sensitivity, specificity, positive predictive value) were noted among US, DM, and DCE-MRI (all P<0.05). Although combined multimodal approaches (serial or parallel) improved certain diagnostic parameters (e.g., sensitivity or specificity, P<0.05), their AUC values were non-inferior to individual modalities (P>0.05). Finally, US had a significantly higher AUC than DM (P=0.0155), while DCE-MRI and combined methods did not outperform US alone (all P>0.05). Conclusion: US showed balanced diagnostic performance, comparable to DCE-MRI and superior to DM for initial evaluation of non-mass-like breast lesions. A refined classification system for NMLs based on imaging-pathology correspondence should be developed in the future.