Explainable machine learning for predicting aspiration pneumonia after radiotherapy in oral cavity cancer
Contact us if these videos are not relevant.
Contact us if these videos are not relevant.
View abstract on PubMed
Summary
This summary is machine-generated.Explainable AI models can predict aspiration pneumonia after oral cavity cancer radiotherapy. The best model identified mean radiation doses to swallowing structures, enabling personalized treatment to reduce patient risk.
Area Of Science
- Oncology
- Radiotherapy
- Machine Learning
- Medical Informatics
Background
- Aspiration pneumonia is a significant late complication following radiotherapy for oral cavity cancer (OCC).
- Predicting this severe event is crucial for patient management and treatment planning.
Purpose Of The Study
- To develop and validate explainable machine learning (ML) models for predicting aspiration pneumonia in OCC patients.
- To identify key clinical and dosimetric predictors of aspiration pneumonia.
Main Methods
- Trained and validated Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Categorical Boosting (CatBoost) models on 880 OCC patients.
- Utilized clinical and dosimetric features, evaluating performance with the area under the curve (AUC).
- Applied SHapley Additive exPlanations (SHAP) for feature importance and model interpretability.
Main Results
- The RF model demonstrated superior performance in external validation (AUC = 0.966).
- Mean radiation doses to the superior pharyngeal constrictor muscle (PCM), middle PCM, and supraglottic larynx were the top predictors.
- Identified a nonlinear dose-toxicity relationship and determined a risk-lowering threshold dose for swallowing structures.
Conclusions
- Explainable ML models can effectively predict aspiration pneumonia post-radiotherapy for OCC.
- Individualized predictions facilitate tailored radiotherapy plans to minimize aspiration pneumonia risk.
Contact us if these videos are not relevant.
Contact us if these videos are not relevant.