Dual-layer detector spectral CT-derived extracellular volume fraction combined with clinical parameters for predicting lymphovascular and perineural invasion in gastric cancer.

[OBJECTIVE] This study aimed to evaluate the feasibility and diagnostic performance of extracellular volume fraction (ECV), derived from iodine concentration (IC) values obtained through dual-layer detector spectral CT, in combination with clinical parameters, for preoperative prediction of lymphovascular invasion and/or perineural invasion (LVI/PNI) in gastric cancer.

[METHODS] This retrospective study included clinical and pathologic data of 124 patients diagnosed with GC at our hospital between September 2023 and February 2025. The patients were stratified into LVI/PNI-positive (n = 85) and LVI/PNI-negative (n = 39) groups based on their postoperative histopathologic findings. All patients underwent contrast-enhanced dual-layer detector spectral CT. Preoperative hematocrit levels were obtained, and IC values within the tumor were measured during the delayed phase. Concurrently, IC in the abdominal aorta at the same anatomical level was measured to calculate the ECV using the formula: ECV = (1 - hematocrit) × (IC_tumor / IC_blood). Clinical variables and spectral CT parameters were analyzed using univariate analysis, LASSO selection, and multivariable logistic regression to identify independent predictors of the composite LVI/PNI endpoint. The diagnostic performance of individual parameters and the combined model for preoperatively predicting vascular and neural invasion was assessed using receiver operating characteristic (ROC) curve analysis.

[RESULTS] ECV, CT-N stage, and Borrmann classification were identified as independent predictors of the composite LVI/PNI endpoint in patients with GC (all P < 0.05). Among the candidate models, the combined model incorporating ECV, CT-N stage, and Borrmann classification achieved the best predictive performance, with AUCs of 0.920 in the training cohort and 0.917 in the validation cohort. DeLong analysis showed that the combined model significantly outperformed the CT-N stage model and the Borrmann classification model in both cohorts. However, its improvement over the ECV-only model was not statistically significant.

[CONCLUSIONS] Delayed-phase dual-layer detector spectral CT-derived ECV is a promising noninvasive biomarker for identifying the composite LVI/PNI endpoint in gastric cancer. A combined model integrating ECV, CT-N stage, and Borrmann classification provides the best overall diagnostic performance, although its incremental benefit over ECV alone appears limited.