Assessing microvascular heterogeneity in esophageal squamous cell carcinoma and peritumoral tissue using DCE-MRI pharmacokinetic parameters

OBJECTIVE: To develop DCE-MRI parameters based logistic regression models to evaluate microvascular heterogeneity between esophageal squamous cell carcinoma (ESCC) and cephalic peritumoral esophageal segmental tissues. METHODS: 133 consecutive ESCC patients receiving preoperative DCE-MRI from two centers were randomly divided into the training, internal validation and external validation cohorts at a ratio of 8:2:2. Mean, standard deviation (SD), coefficient of variation (CV) of DCE-MRI derived reflux rate (Kep), endothelial transfer constant (Ktrans) and fractional extravascular extracellular space volume (Ve) were compared between tumor (T) and five cephalic peritumoral esophageal segmental tissues at incremental 1-5 cm (recorded as PT1-PT5), and the statistical differential parameters were selected to develop logistic regression models to improve assessment of microvascular heterogeneity between ESCC and cephalic peritumoral esophageal segmental tissues. Model's performance was evaluated using area under the receiver operating characteristic curve (AUC) with DeLong's test. RESULTS: All parameters showed differences among T, PT1, PT2 and PT3 except among PT3-PT5. Ve_CV and Kep_mean showed discriminatory ability among T, PT1, PT2 and PT3-5 (P-values < 0.05). Multivariate models by combining Ve_CV and Kep_mean outperformed Ve_CV or Kep_mean to show microvascular heterogeneity in T vs. PT2 (model AUC: 0.830, 0.917, and 0.818), and in T vs. PT3-5 (model AUC: 0.921, 0.845, and 0.857), while Ve_CV outperformed model for T vs. PT1 or Kep_mean (Ve_CV AUC: 0.716, 0.872, and 0.787) in the training, internal and external validation cohorts, respectively (P-values < 0.05). CONCLUSION: Our multivariate models help discriminate ESCC vs. PT2 or PT3-5 whereas Ve_CV can discriminate ESCC vs. PT1. ADVANCES IN KNOWLEDGE: In this study on DCE-MRI, after consolidating peritumoral tissues with 3-5 cm from the cephalic tumor margin of esophageal squamous cell carcinoma into a single group due to no significant differences among them, significant differences were identified among the tumor, peritumoral tissues 1 cm, 2 cm, and 3-5 cm from the margin, reflecting progressive microvascular heterogeneity with the incremental distances. Model by combination of mean reflux rate and coefficient of variation of fractional extravascular extracellular space volume (Ve_CV) help discriminate microvascular heterogeneity between tumor and cephalic peritumoral tissues except that Ve_CV help discriminate tumor vs. 1 cm from the tumor margin.