Preoperative Prediction of Lymphovascular Space Invasion in Endometrial Cancer Using Diffusion MRI-Derived Vessel...

To evaluate the feasibility of diffusion-derived vessel density (DDVD) for the preoperative prediction of lymphovascular space invasion (LVSI) in endometrial cancer (EC) and to explore whether the integration of DDVD with intravoxel incoherent motion (IVIM)-derived parameters and clinicopathologic (CP) variables could improve the prediction.

This exploratory retrospective study included 194 patients with histopathologically confirmed EC (126 LVSI-negative and 68 LVSI-positive) who underwent preoperative multi-b-value diffusion-weighted magnetic resonance imaging. The apparent diffusion coefficient, true diffusion coefficient (D), pseudodiffusion coefficient (D*), perfusion fraction (f), and DDVD_b0b10 (derived from b = 0 and 10 s/mm² images) were calculated. Univariable and multivariable logistic regression analyses were performed to identify independent predictors of LVSI and to develop predictive models. The predictive performance was assessed using receiver operating characteristic (ROC) curve analysis. Differences in areas under the ROC curves (AUCs) among models were evaluated using DeLong's test, with the Bonferroni correction applied to adjust for multiple comparisons.

The f, DDVD_b0b10, and three CP factors, including the depth of myometrial invasion, histologic grade, and tumor stage, emerged as independent predictors of LVSI. DDVD_b0b10 yielded a higher AUC than f (0.772 vs. 0.692), although the difference was not statistically significant after Bonferroni correction (adjusted P = 1.000). Predictive performance improved significantly when DDVD_b0b10 was combined with the retained CP factors (AUC, 0.895 vs. 0.772; P < 0.001). The model combining the f, DDVD_b0b10, and CP predictors achieved the highest AUC of 0.904, although it was not significantly different from that of the DDVD_b0b10 + CP model after Bonferroni correction (adjusted P = 1.000).

DDVD_b0b10 may be a feasible imaging biomarker for the preoperative prediction of LVSI in EC. Multivariable approaches integrating DDVD_b0b10 with CP features and IVIM-derived f may enhance preoperative risk stratification in EC. Further prospective multicenter validation is warranted.