Spectral computed tomography-based quantitative parameters combined with extracellular volume fraction to predict lymph node metastases in gastric cancer.

OBJECTIVES

Preoperative prediction of lymph node metastasis (LNM) is important for gastric cancer (GC) diagnosis, treatment and prognosis. This study aimed to predict LNM risk in GC using quantitative parameters and extracellular volume fraction (ECV%) derived from spectral computed tomography (CT).

MATERIALS AND METHODS

Data from 230 lymph nodes (LNs) (97 nonmetastatic, 133 metastatic) were collected from 70 GC patients and were randomly divided into a training cohort and a test cohort (6:4 ratio). LN qualitative features (including edge, shape and degree of enhancement), spectral CT-derived quantitative parameters and ECV% were assessed. Multivariate logistic regression analysis with the forward variable selection method was used to build 3 models: Model 1 (traditional features: LN edge, short axis diameter), Model 2 (spectral CT parameters: iodine concentration in arterial and delayed phases), and Model 3 (spectral CT parameters and ECV%). Diagnostic performance was evaluated using AUC and compared with the Delong test.

RESULTS

In both cohorts, a significant difference in ECV% was observed between positive and negative LNs (p < 0.001), and the diagnostic efficacy of ECV% (AUC = 0.823 and 0.803, respectively, both p < 0.001) was higher than that of other parameters. Model 3 demonstrated significantly higher diagnostic efficacy than Models 1 and 2 in both cohorts (AUC = 0.858 and 0.881, respectively; both p < 0.001).

CONCLUSION

ECV% can help diagnose LNM in GC, and combining the spectral CT quantitative features with ECV% can further improve diagnosis. This finding enables accurate preoperative prediction of LNM and the GC prognosis so that patients receive personalized treatment.

KEY POINTS

Question Predicting lymph node metastasis (LNM) based on the LN remains a challenge; can spectral CT combined with the Extracellular volume fraction (ECV%) model predict regional LNM? Findings Based on the LN level, the spectral CT combined ECV% model could more accurately predict regional LNM in the training and test cohort. Clinical relevance Accurate and non-invasive preoperative prediction of LNM in each region is important for the individualized treatment and prognosis of gastric cancer. Assisting physicians in selecting the most appropriate treatment approaches to optimize patient outcomes.