Multiparametric MRI for non-mass enhancement breast lesions: a prospective diagnostic accuracy study

Objectives: To evaluate the diagnostic accuracy of multiparametric breast MRI combining dynamic contrast-enhanced imaging with diffusion-weighted imaging (DWI) for differentiating benign from malignant non-mass enhancement (NME) breast lesions. Materials and methods: This prospective diagnostic accuracy study recruited patients from seven tertiary centers with centralized imaging over a 20-month period. The index test was multiparametric MRI, including morphological assessment, kinetic curve analysis, and DWI with apparent diffusion coefficient (ADC) measurements. The reference standard was histopathological confirmation. Nine experienced radiologists from participating centers independently evaluated examinations while blinded to clinical history and outcomes. Inclusion criteria: female patients aged ≥ 18 years with BI-RADS 3–5 NME lesions. Statistical analyses included ROC curve analysis, logistic regression, and inter-reader agreement (IRA) assessment. Results: A total of 351 female patients (mean age, 44.2 ± 11.1 years; 189 malignant, 162 benign lesions) were enrolled. Consensus-based multiparametric MRI achieved 98.4% (95% CI: 95.6–99.7) sensitivity, 86.4% (95% CI: 80.2–91.3) specificity, 89.4% (95% CI: 84.5–92.9) positive predictive value, 97.9% (95% CI: 94.0–99.6) negative predictive value (NPV), and 92.9% (95% CI: 89.6–95.3) accuracy (AUC = 0.94; 95% CI: 0.91–0.96). Individual reader performance averaged 97.1% sensitivity and 84.8% specificity. Five independent predictors were identified: Type III washout curves (OR = 26.84; p = 0.001), segmental distribution (OR = 9.12; p = 0.001), Type II plateau curves (OR = 6.58; p = 0.008), ADC values ≤ 1.3 × 10⁻³ mm²/s (OR = 6.24; p = 0.001), and regional distribution (OR = 5.34; p = 0.032). IRA was almost perfect (κ = 0.847; 95% CI: 0.801–0.893). High NPV was maintained across all BI-RADS categories (96.8–98.5%). Conclusion: Multiparametric breast MRI demonstrates superior diagnostic accuracy for NME lesions with high NPV, supporting potential reduction of unnecessary biopsies while maintaining excellent sensitivity. Key Points: Question Can multiparametric MRI combining dynamic contrast-enhanced imaging with diffusion-weighted imaging improve diagnostic accuracy for non-mass enhancement breast lesions compared to individual imaging parameters? Findings Multiparametric MRI achieved 92.9% accuracy, 98.4% sensitivity, and 86.4% specificity with five independent malignancy predictors, significantly outperforming individual parameters, with almost perfect inter-reader agreement. Clinical relevance Multiparametric breast MRI demonstrates superior diagnostic performance for non-mass enhancement lesions, potentially reducing unnecessary biopsies while maintaining excellent sensitivity for malignancy detection. © The Author(s), under exclusive licence to European Society of Radiology 2026.