Can new ultrasound imaging techniques improve breast lesions characterization?: prospective comparison between ultrasound BI-RADS, histological result, semi-automatic software—“Live BreastAssist”, strain elastography, shear wave elastography and microvascular flow

BackgroundMultiparametric ultrasound (mpUS) integrates BI-RADS (Breast Imaging Report And Data System) morphological assessment with biomechanical biomarkers, including strain elastography (SE) and shear-wave elastography (SWE), along with microvascular flow (M-Flow). Furthermore real-time computer-aided diagnosis (CAD/AI) systems, including Live Breast Assist and S-Detect are employed to standardize image interpretation, thereby reducing operator dependence.PurposeTo evaluate the diagnostic accuracy and reproducibility of mpUS supported by CAD/AI versus B-mode ultrasound alone, assessing their concordance with histopathology for the benign/malignant dichotomy and the incremental contribution of SE, SWE, and M-Flow. Additionally inter-operator agreement was quantified.Materials and methodsIn this single-center prospective study 78 consecutive patients presenting with 78 breast lesions, underwent B-mode/BI-RADS evaluation, Color Doppler, M-Flow (percentage of vascularization), SE (strain ratio), and SWE (quantitative stiffness in kPa). Live Breast Assist was semi-automatically employed during data acquisition, while S-Detect provided focal lesion characterization when deemed necessary. Histopathological examination served as the gold standard.ResultsHistology identified 60 malignancies (76.9%) and 18 benign lesions (23.1%).The highest diagnostic performance was shown by M-Flow (AUC 0.926), reaching 95.1% of sensitivity and 82.4% of specificity. SE also showed strong discriminative ability (AUC 0.885). SWE exhibited lower accuracy (AUC 0.658). B-mode alone achieved an AUC of 0.864. Although with a higher false-negative rate compared with the expert operator, Live Breast Assist demonstrated high specificity (88.2%) and an excellent PPV (96.4%). Inter-operator agreement was substantial (kappa = 0.63).ConclusionmpUS with CAD/AI support enhanced diagnostic accuracy and reproducibility for benign/malignant distinction, providing robust biomechanical and vascular biomarkers. This approach may reduce unnecessary biopsies in indeterminate BI-RADS lesions, though tumor grading remains the purview of pathology.