Development and validation of machine learning models for predicting STAS in stage I lung adenocarcinoma with
Qing-Lin Ren1, Liu Lin2, Kai Chu3
1Department of Graduate School, Dalian Medical University, Dalian, China.
Frontiers in Oncology
|November 14, 2025
Summary
Machine learning accurately predicts spread through air spaces (STAS) in early lung cancer. This XGBoost model aids surgeons in treatment decisions for stage I lung adenocarcinoma patients.
Area of Science:
- Oncology
- Radiology
- Data Science
Background:
- Spread Through Air Spaces (STAS) is a prognostic factor in lung adenocarcinoma.
- Preoperative prediction of STAS is crucial for surgical planning in stage I lung cancer.
- Distinguishing between part-solid and solid nodules requires accurate STAS assessment.
Purpose of the Study:
- To develop and validate machine learning models for preoperative prediction of STAS in stage I lung adenocarcinoma.
- To identify key clinical features predictive of STAS.
- To provide a tool for optimizing surgical strategies and patient counseling.
Main Methods:
- Retrospective analysis of 473 patients with stage I lung adenocarcinoma.
- Feature selection using mRMR and LASSO algorithms.
- Development and evaluation of seven machine learning models, including XGBoost, using ROC curves, calibration plots, and DCA.
- SHAP analysis for feature importance and construction of a web-based nomogram.
Main Results:
- STAS prevalence was 44.76% (training) and 50.83% (validation).
- The XGBoost model achieved high predictive performance (AUC 0.889 training, 0.856 validation).
- Key predictors included CEA, vascular convergence, proGRP, age, AFP, smoking history, and CTR.
Conclusions:
- The XGBoost model offers robust preoperative STAS prediction for stage I lung adenocarcinoma.
- This tool can assist clinicians in optimizing surgical approaches.
- Accurate STAS prediction enhances patient counseling and treatment planning.
