Related Experiment Video
Updated: May 16, 2026
07:13
Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025
Multi-strategy feature selection and multi-model machine learning for prognostic prediction in primary gastric
Jingjie Lin1, Hanlei Wang1, Huirong Lin2
1Department of Gastrointestinal Surgery, The First People's Hospital of Wenling, Affiliated Wenling Hospital, Wenzhou Medical University, Wenling, China.
Digital Health
|May 15, 2026
Summary
A new machine learning model accurately predicts outcomes for primary gastric diffuse large B-cell lymphoma (PG-DLBCL) patients. It integrates clinical, treatment, and socioeconomic factors for better risk stratification and personalized care.
Area of Science:
- Hematology
- Oncology
- Medical Informatics
Background:
- Primary gastric diffuse large B-cell lymphoma (PG-DLBCL) presents heterogeneous outcomes.
- Existing prognostic systems inadequately capture PG-DLBCL's unique features.
- There is a need for improved risk stratification and individualized management strategies.
Purpose of the Study:
- To develop a robust and interpretable prognostic model for PG-DLBCL.
- To enhance risk stratification for patients with PG-DLBCL.
- To guide individualized treatment decisions for PG-DLBCL.
Main Methods:
- Utilized SEER database data from 3773 PG-DLBCL patients (2000-2021).
- Employed four feature selection strategies (LASSO, Boruta, backward stepwise, BSR) and four ML algorithms (logistic regression, SVM, k-NN, XGBoost).
- Evaluated model performance using AUC, calibration, decision curve analysis, SHAP for interpretability, and internal validation.
Main Results:
- Identified age, stage, chemotherapy, marital status, and income as key prognostic factors.
- The XGBoost model, using BSR-selected predictors, demonstrated superior performance, calibration, and clinical benefit.
- SHAP analysis revealed older age, advanced stage, and lack of chemotherapy as risk factors, while marital status and higher income were protective.
Conclusions:
- Developed an interpretable, high-performing ML-based prognostic model for PG-DLBCL.
- The model integrates clinical, treatment, and socioeconomic factors for precise risk stratification.
- A web-based tool was created for individualized risk assessment, aiding therapeutic decisions.