CT Radiomics Model Predicts Pain Relief After Radiotherapy in Bone Metastases

Overview

This dual-center study developed and validated a CT-based radiomics model using machine learning to predict pain relief following palliative radiotherapy in bone metastasis patients. Among 11 algorithms tested, the k-nearest neighbors (KNN) model demonstrated the best predictive performance with AUCs above 0.81 across training, internal validation, and external test sets.

Background

Bone metastases frequently cause severe pain and skeletal-related events, significantly impairing quality of life in advanced cancer patients. Palliative radiotherapy is a standard treatment to alleviate pain, but response rates vary widely, with approximately 40% of patients not achieving satisfactory relief. Conventional clinical predictors have limited accuracy in forecasting individual pain response. Radiomics, which extracts quantitative imaging features from CT scans, offers a promising approach to capture tumor heterogeneity and improve prediction of treatment outcomes. However, prior radiomics studies in this context have been limited by small cohorts and lack of external validation.

Data Highlights

Key Findings

• A total of 134 patients with bone metastases were analyzed, with 53 achieving pain relief and 81 not.
• Seven radiomic features were selected from CT images for model construction.
• Eleven machine learning classifiers were compared; the KNN model showed superior predictive accuracy.
• KNN model AUCs were consistently above 0.81 across training, internal validation, and external test cohorts.
• Decision curve analysis demonstrated the model's favorable net clinical benefit for predicting pain relief.

Clinical Implications

The CT radiomics-based KNN model can assist clinicians in identifying bone metastasis patients likely to benefit from palliative radiotherapy in terms of pain relief. This predictive tool may guide personalized treatment planning and optimize resource allocation by targeting radiotherapy to patients with higher probability of response. Integration of such models into clinical workflows could improve patient quality of life by anticipating and managing pain outcomes more effectively.

Conclusion

This study establishes a robust, externally validated CT radiomics model using KNN to predict pain relief after radiotherapy in bone metastases, highlighting its potential utility in personalized palliative care. Further prospective studies are warranted to confirm clinical benefits and facilitate implementation.

References

1. A Radiomics Model Utilizing CT Imaging to Forecast Pain Relief Following Radiotherapy in Bone Metastasis Patients: Findings from a Dual-Center Investigation