Development and verification of a radiomics model to forecast Ki67 index and prognosis in advanced gastric tubular adenocarcinoma

Abstract

Background

The purpose of this research is to evaluate the predictive capabilities of a radiomics model for the Ki67 index and its correlation with prognosis in advanced gastric tubular adenocarcinoma patients.

Methods

Clinical data from 213 patients were analyzed, categorizing patients into high and low Ki67 index groups. The radiomic features of 192 patients were selected by lasso method and the model was constructed, which was validated using the TCIA dataset. Univariate and multivariate Cox regression analyses were used to further analyze clinical features associated with the prognosis of gastric cancer, radiomic models are also used to assess patient outcomes.

Results

The radiomics model demonstrated moderate accuracy, with AUC values of 0.634, 0.666, and 0.602 for the training, validation 1, and validation 2 sets, respectively. Additionally, a significant correlation was found between the Ki67 index and radiomics scores, a higher Ki67 index was associated with improved outcomes. Kaplan-Meier analysis showed distinct survival differences between patients with high and low radiomics scores, indicating that higher scores predict better prognosis.

Conclusions

The radiomics model accurately predicts the Ki67 index and correlates with prognosis in advanced gastric tubular adenocarcinoma, offering valuable insights for clinical decision-making and personalized treatment strategies.