Diagnostic value of super-resolution ultrasound imaging in differentiating benign and malignant BI-RADS-4 breast lesions

Xueqin HOUZhiming LIYibin LIUJunxi GAOTao SONG^*

• First Affiliated Hospital of Xinjiang Medical University, Urumqi, China

Objective: Benign and malignant breast tumors exhibit distinct microvascular morphological patterns and spatial distribution characteristics; however, current clinical imaging modalities cannot comprehensively assess their microvascular network architecture. Super-resolution ultrasound (SRUS) imaging addresses this critical gap by providing super-resolved visualization of microvascular topology. This study aimed to evaluate the utility of SRUS imaging in visualizing breast lesion microvasculature and establishing diagnostic models for BI-RADS 4 masses. Methods: A total of 120 breast lesions from 117 patients with conventional ultrasound-confirmed BI-RADS 4 lesions were prospectively enrolled in this study between July 2024 and January 2025. Preoperative conventional ultrasound and SRUS examinations were performed on all included patients. Based on pathological findings, the lesions were categorized as benign (n = 85) or malignant (n = 35). The benign group was further stratified into hypovascular (n = 42) and hypervascular (n = 43) subgroups based on the SRUS enhancement levels. Univariate analysis was performed to screen SRUS parameters, and variables with P < 0.05 were incorporated into multivariate logistic regression models to construct nomogram-based predictive models and validated using ROC analysis. Results: Max vel (OR = 1.848, 95% CI: 1.205–3.122), curvature (A/E) (OR = 2.162, 95% CI: 1.981–2.323), and complexity level (OR = 1.772, 95% CI: 1.608–1.942) independently predicted malignancy (all P < 0.05). Curvature (A/E) and complexity level were also independent markers for distinguishing malignant lesions from hypervascular benign lesions (curvature (A/E): OR = 1.808, 95% CI: 1.612– 1.987; complexity level: OR = 1.952, 95% CI: 1.804–2.181; both P < 0.05). The nomogram prediction models demonstrated high diagnostic efficacy, with AUC values of 0.899 (95% CI: 0.844– 0.953) (sensitivity = 94.3%, specificity = 76.5%) for benign compared to 0.816 (95% CI: 0.721– 0.910 (sensitivity = 65.1%, specificity = 88.6%) for malignant differentiation. The Hosmer-Lemeshow goodness-of-fit test indicated adequate model fit (all P > 0.05). The nomogram prediction model demonstrated superior net benefit in predicting breast cancer compared to alternative strategies. Conclusion: SRUS enables microvascular characterization of breast lesions, with validated nomograms demonstrating high diagnostic accuracy for early cancer detection.