Exploratory preoperative assessment of lymphovascular space invasion in cervical cancer: comparative insights from intravoxel incoherent motion and diffusion-relaxation correlation spectrum imaging

OBJECTIVES: To compare diffusion-relaxation correlation spectrum imaging (DR-CSI) and intravoxel incoherent motion (IVIM) for exploratory preoperative assessment of lymphovascular space invasion (LVSI) in cervical cancer. METHODS: This exploratory retrospective study included 67 patients with pathology-confirmed cervical cancer, including 31 LVSI-positive and 36 LVSI-negative cases. Patients underwent preoperative 3.0 T MRI with DR-CSI and IVIM. DR-CSI-derived spectral fractions (f-A to f-F) and IVIM-derived parameters (D, D*, and f) were extracted from tumor volumes. Variables with p < 0.10 in univariable logistic regression were entered into separate DR-CSI and IVIM multivariable models. ROC analysis was performed in the same cohort, without internal or external validation; therefore, AUCs were interpreted as apparent performance estimates. RESULTS: In the final DR-CSI model, f-C (p < 0.001) and f-F (p = 0.001) were retained as LVSI predictors, yielding an apparent AUC of 0.866 (95% CI: 0.760-0.937). Among IVIM-derived parameters, D was retained in the final IVIM model (p = 0.029), with an apparent AUC of 0.630 (95% CI: 0.503-0.745). The DR-CSI model showed substantially higher apparent discrimination performance than the IVIM model (p = 0.004). CONCLUSION: DR-CSI-derived spectral fractions showed preliminary value for preoperative assessment of LVSI-related tumor aggressiveness in cervical cancer, with higher apparent discrimination than IVIM. Larger validated studies are needed to confirm clinical utility. ADVANCES IN KNOWLEDGE: DR-CSI-derived spectral fractions, particularly f-C and f-F, may represent candidate MRI biomarkers for preoperatively assessing LVSI-related tumor features in cervical cancer, with higher apparent discrimination than IVIM in this exploratory cohort, although further validation is required.