CT Radiomics Model Predicts Microsatellite Instability and Immunotherapy Response in Gastric Cancer

Researchers at Lanzhou University have developed a CT radiomics model that can predict microsatellite instability (MSI) status and immunotherapy response in gastric cancer patients. The model demonstrates good predictive performance, achieving AUCs of 0.952, 0.835, and 0.879 in the training set and two external testing sets, respectively. The study revealed that radiomics scores (Radscores) were an independent predictor for progression-free survival (PFS) in the outcome cohort, with a hazard ratio of 0.145 (95% CI: 0.032-0.657, p = 0.012). The model also showed positive correlation with CD8+ T cells and negative correlation with M2-type macrophages.

Key Takeaways:

  • The CT radiomics model can effectively predict MSI status and immunotherapy outcomes in gastric cancer patients.
  • The model achieved AUCs of 0.952, 0.835, and 0.879 in the training set and two external testing sets, respectively.
  • Radiomics scores (Radscores) were an independent predictor for PFS in the outcome cohort with a hazard ratio of 0.145 (95% CI: 0.032-0.657, p = 0.012).
  • The model showed positive correlation with CD8+ T cells (R = 0.74, p = 0.013) and negative correlation with M2-type macrophages (R = -0.67, p = 0.028).
  • The radiomics model can serve as a noninvasive biomarker to identify gastric cancer patients who may benefit from immunotherapy.
  • The model can also help develop personalized treatment decisions for gastric cancer patients.

Statistics:

  • AUC in the training set: 0.952
  • AUC in the first external testing set: 0.835
  • AUC in the second external testing set: 0.879
  • Hazard ratio for PFS: 0.145 (95% CI: 0.032-0.657, p = 0.012)
  • Correlation coefficient with CD8+ T cells: 0.74 (p = 0.013)
  • Correlation coefficient with M2-type macrophages: -0.67 (p = 0.028)

Sources:

  • Computed tomography radiomics to predict microsatellite instability status and immunotherapy response in gastric cancer. Insights into Imaging, 2025;16(1):177.
  • Lanzhou University, School of Medicine, Lanzhou, People's Republic of China.
  • Springer Wien, Prinz-Eugen-Strasse 8-10, A-1040 Vienna, Austria.