Radiomics-based assessment of HER2 status and prognosis in gastric cancer: a retrospective dual-center CT study.

Objective: This research investigated the potential of CT-based radiomics analysis for predicting human epidermal growth factor receptor 2 (HER2) status and assessing the prognosis of patients with gastric cancer (GC).

Methods: 431 patients with GC from two medical centers were included in this retrospective study, with patients allocated to a training cohort (n = 221), a testing cohort (n = 94), and an external validation cohort (n = 116). Radiomics features and clinical variables associated with HER2 status were identified, and the radiomics score was subsequently derived. A radiomics model was constructed using the radiomics score, and a nomogram was developed by integrating related variables. The predictive accuracy of models was assessed via receiver operating characteristic curves, with the area under the curve (AUC) being computed. Prognostic significance was assessed by exploring the association between nomogram-predicted HER2 status and overall survival (OS).

Results: The radiomics model yielded AUCs of 0.801, 0.793, and 0.784 for the training, testing, and external validation cohorts, respectively. A nomogram that integrated sex, CA72-4 levels, and radiomics score exhibited enhanced predictive accuracy, achieving AUCs of 0.847, 0.836, and 0.828 across the cohorts. Decision curve analysis highlighted the clinical utility of the nomogram, illustrating its ability to deliver a higher net benefit. In addition, survival analysis indicated that individuals with nomogram-predicted HER2 positivity experienced significantly shorter OS, providing robust risk stratification and prognostic insights.

Conclusions: The CT-based radiomics nomogram demonstrated the ability to non-invasively predict preoperative HER2 status and stratify prognostic risk in this GC cohort.