Dual-layer spectral-detector CT-derived quantifications for preoperative identification of perineural invasion in...

To investigate the utility of dual-layer spectral-detector CT (DLCT)-derived quantifications for preoperative identification of perineural invasion (PNI) in gastric cancer (GC).

Between Feb 2023 and April 2024, 80 GC patients confirmed by surgical pathology who underwent preoperative multi-phase enhanced DLCT including arterial/venous/equilibrium phase (AP/VP/EP) were retrospectively included. Post-surgical pathology diagnosis of PNI was the reference standard, patients were divided into PNI-positive-or-negative accordingly. DLCT-based measurements, including iodine density (ID), normalized ID (nID), effective atomic number (Zeff), normalized Zeff (nZeff), arterial enhancement fraction (AEF1 based on AP/VP, AEF2 based on AP/EP), extracellular volume (ECV), were completed using two dimensional free-hand delineations at each phase and their differences were compared between different PNI groups. Independent predictors were screened and used to establish a combined parameter. Its performance was assessed using the receiver operating characteristic curve analysis and tested in an external dataset of 33 patients. Their associations with patient survival outcomes were explored by the Kaplan-Meier curve analysis in the primary dataset.

AEF1, nIDVP, and nZeffVP were independent predictors with similar AUCs of 0.666, 0.679, and 0.701, respectively. Their combination demonstrated superior performances with AUC achieving 0.865, 0.812, respectively for the primary and validation dataset. The risk-category defined by the combined parameter was associated with patient disease-free survival (χ² = 5.012, p = 0.025) with hazard ratio being 3.462 (95% CI: 1.198-10.008) in the primary dataset.

DLCT-derived quantifications of AEF1, nIDVP, and nZeffVP are equally useful for identifying PNI in GC, their combination demonstrated incremental benefit and was associated with patient survival outcomes.