Preoperative prediction of HER2 expression in gastric cancer based on an enhanced CT radiomics model.

Background: Accurate assessment of human epidermal growth factor receptor 2 (HER2) status is particularly significant for gastric cancer patients. This study aimed to explore the application value and internal validation of constructing an imaging omics model based on portal vein phase enhanced computed tomography (CT) for the preoperative prediction of HER2 status in gastric cancer. Methods: A total of 100 gastric cancer patients (HER2 negative: n=79; HER2 positive: n=21) who underwent curative surgery for gastric cancer and underwent postoperative pathological immunohistochemistry (IHC) were retrospectively included and were randomly divided into a training cohort (n=70) and a validation cohort (n=30). The differences in preoperative clinical indicators between the two groups identified through univariate analysis were compared, and variables with P<0.05 were incorporated into multivariate logistic regression analysis to screen for independent risk factors and establish a clinical model. Radiomics features were extracted from portal phase (PP) images, and minimum absolute shrinkage and selection operators were used in Spearman correlation analysis to screen features and establish a radiomics prediction model. A combined model was constructed by combining clinical independent risk factors with radiomics features to visualize the results as a column chart and calculate the area under the curve (AUC) and the receiver operating characteristic (ROC). The AUC values were used to evaluate the predictive performance of the clinical radiomics combined model, radiomics model, and clinical model. Calibration curves were used to estimate the fitting degree of the column chart model in the training and validation and decision curve analysis (DCA) was used to evaluate the application value of the column chart. Results: After dimensionality reduction and screening, eight imaging omics features related to the status of HER2 expression were retained, and a clinical imaging omics combined model was established by combining clinical independent risk factors (serosal invasion) with imaging omics features. The AUC values for the training and validation cohorts of clinical models were 0.932 [95% confidence interval (CI): 0.846-1.000] and 0.944 (95% CI: 0.884-1.000), respectively, and the AUC values for the training and validation cohorts of radiomics models were 0.823 (95% CI: 0.708-0.939) and 0.790 (95% CI: 0.568-1.000), respectively. The AUC value for the combined model in the training cohort was 0.986 (95% CI: 0.849-1.000), and that for the validation cohorts was 0.938 (95% CI: 0.849-1.000). The performance of the combined model with respect to predicting the HER2 expression status was superior to that of imaging omics models and clinical models. The calibration curve also showed good calibration performance. DCA showed that the imaging omics column chart had good application value. Conclusions: The imaging omics model based on enhanced CT imaging features exhibits good performance for predicting HER2 expression status in gastric cancer patients. The predictive model established in combination with clinical independent risk factors has improved performance.