KA-TMoE: A Deep Learning Method Using Time-Series CT Radiomics to Predict Post-Radiotherapy Rib Fractures in NSCLC
Yijun Chen1, Michael Farris1, Ariel R Choi1
1Department of Radiation Oncology, Wake Forest University School of Medicine, Winston-Salem, NC, United States.
International Journal of Radiation Oncology, Biology, Physics
|August 10, 2026
Summary
A novel deep learning model, KA-TMoE, accurately predicts rib fractures in non-small cell lung cancer (NSCLC) patients after stereotactic body radiotherapy (SBRT). This tool aids in personalized risk stratification and improved patient care.
Area of Science:
- Radiology
- Oncology
- Artificial Intelligence
Background:
- Rib fractures are a common complication for non-small cell lung cancer (NSCLC) patients undergoing stereotactic body radiotherapy (SBRT).
- Predicting fracture risk is crucial for managing patient quality of life and recovery.
Purpose of the Study:
- To develop and validate a deep learning model for predicting rib fractures after SBRT.
- To utilize time-series CT radiomics for enhanced fracture risk assessment.
Main Methods:
- A retrospective study involving 67 NSCLC patients and over 1600 ribs.
- Development of a Knowledge-aware Temporal Mixture of Experts (KA-TMoE) model integrating longitudinal CT radiomics and radiomic grouping knowledge.
- Evaluation of model performance and interpretability.
Main Results:
- The KA-TMoE model achieved an AUC of 0.792 in the validation cohort.
- Both longitudinal data and radiomics-grouping knowledge improved predictive performance.
- Model-predicted high-risk patients had a significantly higher fracture risk (HR=10.82, p < 0.001).
Conclusions:
- The KA-TMoE model shows promise as a risk-stratification tool for post-SBRT rib fractures in NSCLC patients.
- This framework can support personalized risk assessment, closer monitoring, and timely interventions.
