Prediction of Response to Spinal Cord Stimulation Using Machine Learning Based on Radiomics and Patient-Reported
Eung-Joo Lee1, Meghan L Edgerton2, Barbara Buccilli3
1Department of Electrical and Computer Engineering, University of Arizona College of Engineering, Tucson , Arizona , USA.
Neurosurgery
|August 29, 2025
Summary
Machine learning models integrating radiomics and clinical data accurately predict spinal cord stimulation (SCS) outcomes. This approach improves patient selection for SCS, enhancing treatment effectiveness and reducing healthcare costs.
Area of Science:
- Pain Medicine
- Medical Imaging
- Artificial Intelligence
Background:
- Chronic pain impacts a significant patient population, leading to high healthcare expenditures.
- Spinal cord stimulation (SCS) is an FDA-approved treatment for chronic pain, with increasing utilization.
- High SCS failure rates necessitate improved patient selection criteria.
Purpose of the Study:
- To develop machine learning (ML) models for predicting patient response to SCS.
- To integrate spinal imaging radiomics and clinical data for enhanced predictive accuracy.
- To identify patients most likely to benefit from SCS, optimizing outcomes and reducing costs.
Main Methods:
- Development of ML models using the largest US SCS database.
- Integration of spinal imaging radiomics with patient clinical data.
- Validation of predictive models for SCS response.
Main Results:
- The integrated ML model achieved 90.00% accuracy and 91.40% AUC for predicting "50% Responder" status.
- For the "70% Responder" target, the model showed 90.00% accuracy and 86.11% AUC.
- Radiomic features combined with clinical variables significantly improved predictive capabilities.
Conclusions:
- ML models incorporating radiomics and clinical data are valuable for predicting SCS outcomes.
- Systematic feature selection enhances model interpretability and robustness.
- Integrating diverse data sources is crucial for improving SCS patient selection and treatment efficacy.
