Hasty Briefsbeta

Bilingual

Artificial intelligence for the prediction of prognosis in colorectal cancer patients using routine blood indices - PubMed

5 hours ago
  • #prognostic prediction
  • #artificial intelligence
  • #colorectal cancer
  • Aim: Develop explainable ML models using routine blood indices to predict prognosis in colorectal cancer patients across multicenter cohorts.
  • Method: Used seven time-to-event models and SHAP for interpretation, with training cohort (850 patients from Union) and two external validation cohorts (Hefei: 403 patients, Shihezi: 217 patients).
  • Result: Random Survival Forest (RSF) algorithm showed superior performance, with high AUCs for 1-, 2-, and 3-year overall survival in training and validation cohorts.
  • SHAP analysis identified key predictors: CEA, CA125, age, MPV, CA19-9, INR, and monocytes, contributing to accurate RSF model predictions.
  • Conclusion: The RSF model provides an innovative, convenient strategy for accurate survival prediction, helping identify high-risk CRC patients early and supporting personalized treatment.