Local Recurrence Prediction After Carbon-ion Radiotherapy for Early-stage Non-small Cell Lung Cancer Using Machine
Keishiro Mochida1, Yuhei Miyasaka2,3, Nobuteru Kubo1,3
1Department of Radiation Oncology, Gunma University Graduate School of Medicine, Maebashi, Japan.
Anticancer Research
|July 29, 2026
Summary
This study developed a machine learning model to predict local recurrence in non-small cell lung cancer (NSCLC) patients after carbon-ion radiotherapy (CIRT). The model shows potential for identifying high-risk patients for improved treatment strategies.
Area of Science:
- Oncology
- Radiotherapy
- Machine Learning
Background:
- Predicting local recurrence after carbon-ion radiotherapy (CIRT) for non-small cell lung cancer (NSCLC) is challenging.
- Early-stage peripheral NSCLC patients treated with CIRT require accurate risk stratification.
Purpose of the Study:
- To develop and validate a machine learning model for predicting local recurrence within 24 months after CIRT in early-stage peripheral NSCLC.
- To stratify patients into risk groups for 2-year local control.
Main Methods:
- Retrospective analysis of 124 patients treated with CIRT between 2010 and 2020.
- Development of an Extreme Gradient Boosting classifier using clinical parameters and nested threefold cross-validation.
- Evaluation using ROC-AUC, PR-AUC, survival analysis, and SHapley Additive exPlanations (SHAP) for interpretability.
Main Results:
- The prediction model achieved an ROC-AUC of 0.622 and a PR-AUC of 0.145.
- Identified low-risk and high-risk groups with 2-year local control rates of 94.0% and 65.0%, respectively (p<0.01).
- SHAP analysis indicated the importance of Brinkman Index, C-reactive protein, and solid tumor component diameter.
Conclusions:
- A machine learning model utilizing clinical parameters can predict 2-year local recurrence after CIRT for early-stage peripheral NSCLC.
- This model aids in identifying patients at higher risk for local recurrence, potentially guiding treatment decisions.
