Imaging characteristics and discrimination model development for early gastric cancer using multi-spectral CT

Early detection of gastric cancer is vital for improving patient outcomes. While endoscopy with biopsy remains the gold standard for diagnosis, its invasiveness hinders its use in population screening. Computed tomography (CT), though widely available, has limited sensitivity for early gastric cancer (EGC). This exploratory study aimed to identify characteristic imaging indicators of EGC on multi-spectral CT (MSCT).

This retrospective study analyzed MSCT data from 141 patients (144 lesions) to identify and analyze imaging characteristics of EGC. Receiver operating characteristic analysis was employed to evaluate the discrimination performance of various parameters, and a least absolute shrinkage and selection operator (LASSO) regression model was constructed for discrimination. Subsequently, data from lesions and adjacent normal gastric mucosa in 51 EGC patients were used for external validation of the model's discrimination efficacy.

MSCT imaging revealed that the mean CT and effective atomic number (Effective-Z) values of EGC lesions were significantly higher than those of adjacent normal mucosa in both arterial and venous phases. Although multiple parameters showed significant differences, the discrimination efficacy of any single indicator was limited. The multivariate discrimination model constructed with LASSO regression demonstrated excellent performance, achieving area under the curve (AUC) values exceeding 0.90 in the training set, internal validation set, and external independent validation.

MSCT shows significant advantages in the imaging evaluation of EGC. The identification model constructed by integrating multiple imaging parameters may effectively improve the discrimination rate of EGC, providing a new reference for clinical practice.