Related Experiment Video
Updated: Jun 30, 2025
07:53
Author Spotlight: Advancing 3D Modeling for Enhanced Diagnosis and Treatment of Pulmonary Nodules in Early-Stage Lung Cancer
Published on: October 13, 2023
1.4K
Enhancing the prediction of symptomatic radiation pneumonitis for locally advanced non-small-cell lung cancer by
Yan Kong1, Mingming Su1,2, Yan Zhu1
1Department of Radiation Oncology, Affiliated Hospital of Jiangnan University, 1000 Hefeng Road, 214122, Wuxi, Jiangsu, China.
Summary
Deep learning (DL) imaging features from pre-radiotherapy CT scans can predict radiation pneumonitis (RP) in locally advanced non-small-cell lung cancer (LA-NSCLC) patients. Integrating these features with dose-volume metrics enhances predictive accuracy for better treatment outcomes.
Area of Science:
- Radiology
- Oncology
- Artificial Intelligence in Medicine
Background:
- Radiation pneumonitis (RP) is a significant toxicity in lung cancer radiotherapy.
- Accurate prediction of RP is crucial for personalized treatment planning and toxicity management in locally advanced non-small-cell lung cancer (LA-NSCLC).
Purpose of the Study:
- To evaluate the efficacy of deep learning (DL)-derived imaging features for predicting RP in LA-NSCLC patients.
- To assess the added value of combining DL features with clinical variables and dose-volume metrics for improved RP prediction.
Main Methods:
- A cohort of 149 LA-NSCLC patients was analyzed.
- 512 3D DL features were extracted from pre-radiotherapy planning CT scans (lung-PTV and PTV-GTV regions).
- LASSO regression and multilayered perceptron were used for feature selection and model building, with performance evaluated by ROC analysis.
Main Results:
- DL features from lung-PTV showed superior prediction of RP (AUC 0.921) compared to PTV-GTV (AUC 0.892) in the internal test set.
- Incorporating the V30Gy dose-volume metric improved prediction model performance (AUC increased from 0.835 to 0.881 in training, 0.690 to 0.746 in validation).
Conclusions:
- 3D DL-derived imaging features from pre-radiotherapy CT scans can effectively predict RP in LA-NSCLC patients.
- Combining DL features with dose-volume metrics offers a promising non-invasive approach for RP prediction, aiding risk stratification and personalized radiotherapy.
