Correlation between quantitative parameters and apparent diffusion coefficient of 3.0T dynamic contrast-enhanced MRI and prognostic factors and molecular classification of breast cancer

Preoperative,image-only stratification of breast cancer prognostic factors and molecular subtypes remains a clinically important objective. Most existing reports have analysed dynamic contrast-enhanced MRI (DCE-MRI) and diffusion-weighted imaging (DWI) in separate cohorts, used heterogeneous protocols, and lacked formal multivariable modelling. The aim of the study was to investigate, in a single large Chinese cohort, the correlation between quantitative DCE-MRI parameters (Ktrans, Kep, Ve), the apparent diffusion coefficient (ADC) and breast-cancer prognostic factors (ER, PR, HER-2, Ki-67) and molecular subtypes; and to develop a multivariable logistic regression model for the preoperative identification of triple-negative breast cancer (TNBC). A retrospective analysis was conducted on 746 patients with breast cancer surgically and pathologically confirmed at our hospital between January 2016 and January 2019. All patients underwent preoperative 3.0T DCE-MRI and DWI; DWI was performed prior to contrast injection. Quantitative DCE-MRI parameters (Ktrans, Kep, Ve) were derived using the extended Tofts model on a vendor-neutral platform (Tissue 4D, Siemens). ROI placement reproducibility was quantified by intraclass correlation coefficients (ICCs) in 100 lesions independently re-measured by two readers and re-measured 4 weeks later by the senior reader. Distributions were assessed by Shapiro–Wilk and Levene's tests. Bonferroni correction was applied to the 16 Spearman correlation tests (corrected α = 0.003). AUCs were reported with 95% CIs (DeLong method). The cohort comprised 149 Luminal A, 400 Luminal B, 88 HER-2 overexpression and 109 TNBC lesions. After Bonferroni correction, ER, PR negatively correlated with Ktrans, Kep and ADC (all P < 0.001); HER-2 correlated positively with Ve and ADC and negatively with Kep (all P < 0.001); Ki-67 correlated positively with Ktrans and Kep (all P < 0.001); other correlations were non-significant. TNBC lesions had the highest Ktrans (2.45 ± 0.45 min⁻¹), highest Kep (6.40 ± 0.71 min⁻¹) and the lowest Ve (0.42 ± 0.15) of the four subtypes. Single-parameter AUC for TNBC was highest for Kep (0.826, 95% CI 0.789–0.863). Ve showed inverse discrimination (AUC 0.170; reversed AUC 0.830). A multivariable logistic regression model combining Ktrans, Kep and inverted Ve achieved AUC 0.891 (95% CI 0.857–0.926), significantly higher than any single parameter (DeLong P < 0.001). All ICCs were ≥ 0.85. At the population level, quantitative DCE-MRI parameters and ADC show weak but reproducible correlations with breast-cancer prognostic factors and molecular subtypes. A combined Ktrans+Kep+inverted-Ve logistic regression model identifies TNBC with substantially better performance than any single parameter. These findings should be regarded as hypothesis-generating; prospective external validation is required before clinical translation.