Multiparametric diffusion‐weighted imaging in breast lesions: Association with pathologic diagnosis and prognostic factors

Abstract

Purpose To determine the utility of multiparametric diffusion‐weighted imaging (DWI) including monoexponential (apparent diffusion coefficient [ADC]), biexponential ( D f , D s , and f ), stretched‐exponential (distributed diffusion coefficient [DDC] and α), and kurtosis (mean diffusivity [MD] and mean kurtosis [MK]) models in the differentiation and characterization of breast lesions, and assess their associations with prognostic factors in invasive breast cancer. Materials and Methods This study included 101 patients (44 benign and 57 malignant lesions) who underwent 3T breast multi‐ b ‐value DWI. Diffusion model selection was investigated in benign and malignant lesions using the Akaike information criteria (AIC). Mann–Whitney U ‐test and receiver operating characteristic (ROC) curves were used for statistical analysis. Results Goodness‐of‐fit analysis showed that most benign lesion voxels (50.5%) were preferred by the kurtosis model, and most malignant lesion voxels (51.2%) by the stretched‐exponential model. All diffusion measures showed significant differences between benign and malignant lesions ( P < 0.05), and between in situ and invasive cancers ( P < 0.05) except MD ( P = 0.103). There were no significant differences in areas under the ROC curves (AUCs) between ADC and non‐monoexponential diffusion parameters ( P > 0.05), except D f and α, whose AUCs were significantly lower than AUC of ADC for differentiating benign from malignant lesions ( P = 0.03 and P < 0.01, respectively). In patients with invasive breast cancer, α was significantly correlated with tumor size ( P = 0.007) and Ki‐67 expression ( P = 0.012), D f was significantly correlated with lymph node metastasis ( P = 0.021) and Ki‐67 expression ( P = 0.042), and ADC, D s , f , DDC, and MD were significantly correlated with estrogen receptor status (all P < 0.05). Conclusion Multiparametric DWI shows relationships with pathologic outcomes and prognostic factors of breast lesions. Level of Evidence: 3 Technical Efficacy: Stage 2 J. MAGN. RESON. IMAGING 2017;46:740–750