Prediction of false-negative extramural venous invasion in patients with rectal cancer using multiple mathematical models of diffusion-weighted imaging

Purpose: To investigate the parameters from mono-exponential, stretched-exponential, and intravoxel incoherent motion diffusion-weighted imaging (DWI) models for evaluating false-negative extramural venous invasion (EMVI) on conventional magnetic resonance imaging (MRI) in rectal cancer patients. Material and methods: Seventy-two rectal cancer patients with negative EMVI on conventional MRI who underwent direct surgical resection were enrolled in this prospective study. The apparent diffusion coefficient (ADC), true diffusion coefficient (D), pseudo-diffusion coefficient (D*), perfusion fraction (f), distributed diffusion coefficient (DDC), and water molecular diffusion heterogeneity index (alpha) values within the whole tumor were obtained to identify the patients with false-negative EMVI. Receiver operating characteristic (ROC) curves were applied to evaluate the diagnostic performance. Multivariate binary logistic regression analysis was conducted to determine the independent risk factors. Results: The DDC, D*, f, and alpha values were significantly different in the EMVI-positive and EMVI-negative groups (P = 0.018, and P < 0.001, respectively). The D*, f, and alpha values demonstrated good diagnostic performance with area under the ROC curve (AUC) of 0.861, 0.824, and 0.854, respectively. The combined model, including D*, alpha, and tumor location, proved superior diagnostic performance with the AUC, sensitivity, specificity, and accuracy of 0.971, 0.917, 0.967, and 0.931, respectively. The AUC of the combined model was significantly higher than that of the D*, f, and DDC (P = 0.004, 0.045, and 0.002, respectively). Conclusion: Multi-b-value DWI may be a potential tool for identifying micro-EMVI in rectal cancer. The combination of DWI parameters and tumor location leads to superior diagnostic performance.