Abstract:PURPOSE: To assess the continuous-time random-walk (CTRW) model's diagnostic value in breast lesions and to explore the associations between the CTRW parameters and breast cancer pathologic factors.METHOD: This retrospective study included 85 patients (70 malignant and 18 benign lesions) who underwent 3.0T MRI examinations. Diffusion-weighted images (DWI) were acquired with 16b-values to fit the CTRW model. Three parameters (Dm, α, and β) derived from CTRW and apparent diffusion coefficient (ADC) from DWI were compared among the benign/malignant lesions, molecular prognostic factors, and molecular subtypes by Mann-Whitney U test. Spearman correlation was used to evaluate the associations between the parameters and prognostic factors. The diagnostic performance was assessed by the area under the receiver operating characteristic curve (AUC) based on the diffusion parameters.RESULTS: All parameters, ADC, Dm, α, and β were significantly lower in the malignant than benign lesions (P < 0.05). The combination of all the CTRW parameters (Dm, α, and β) provided the highest AUC (0.833) and the best sensitivity (94.3%) in differentiating malignant status. And the positive status of estrogen receptor (ER) and progesterone receptor (PR) showed significantly lower β compared with the negative counterparts (P < 0.05). The high Ki-67 expression produced significantly lower Dm and ADC values (P < 0.05). Additionally, combining multiple CTRW parameters improved the performance of diagnosing molecular subtypes of breast cancer. Moreover, Spearman correlations analysis showed that β produced significant correlations with ER, PR and Ki-67 expression (P < 0.05).CONCLUSIONS: The CTRW parameters could be used as non-invasive quantitative imaging markers to evaluate breast lesions.

Evaluation of breast cancer malignancy, prognostic factors and molecular subtypes using a continuous-time random-walk MR diffusion model