Predicting peritoneal recurrence and hyperthermic intraperitoneal chemotherapy benefits in T4 gastric cancer patients via contrast-enhanced CT radiomics.

OBJECTIVES

This study aimed to predict post-surgical peritoneal recurrence (PR) and the preventative effects of hyperthermic intraperitoneal chemotherapy (HIPEC) in T4 gastric cancer (GC) patients using CT radiomics and clinical features.

MATERIALS AND

METHODS

Patients with T4 GC who underwent radical gastrectomy were retrospectively enrolled. A total of 1343 radiomics features were extracted from each normalized contrast-enhanced CT series. The radiomics score (RadScore) was computed via an ensembled machine learning framework to predict the risk of PR. We also generated a combined risk score by integrating the RadScore and clinicopathological features. The performance for predicting PR was evaluated by the area under the curve (AUC), Kaplan-Meier analysis, c-index, and decision curve analysis. The performance for predicting HIPEC benefits was evaluated via Kaplan-Meier analysis and interaction analysis.

RESULTS

A total of 245 T4 GC patients (aged 60.54 ± 9.79 years, 177 men) were enrolled in this study. The RadScore could identify patients at high risk of PR with moderate accuracy (test cohort AUC: 0.726; [HR]: 3.088). The combined risk score outperformed the RadScore in predicting PR (test cohort AUC: 0.820, net reclassification index: 0.446, p < 0.01). T4 GC patients with high combined risk scores are more likely to benefit from prophylactic HIPEC ([HR]: 0.490).

CONCLUSION

The combined risk score integrating imaging biomarkers and clinicopathological traits could precisely identify T4 GC patients at high risk of PR. Patients with higher risk scores are more likely to benefit from prophylactic HIPEC.