Machine Learning Algorithms Provide Greater Prediction of Response to SCS Than Lead Screening Trial: A Predictive...

Area of Science:

• Pain Management
• Medical Artificial Intelligence
• Neurosurgery

Background:

• Persistent pain post-spinal surgery is a significant challenge.
• Spinal cord stimulation (SCS) is a recommended treatment, typically preceded by a lead trial.
• Current lead trial methods have limitations, including similar patient outcomes and increased infection risk over time.

Purpose of the Study:

• To investigate the efficacy of machine learning (ML) models in predicting long-term spinal cord stimulation (SCS) success.
• To compare the predictive performance of various ML algorithms against traditional lead trial outcomes.
• To identify a robust ML model for optimizing clinical decisions in SCS therapy.

Main Methods:

• Development and validation of multiple ML algorithms (logistic regression, RLR, naive Bayes, ANNs, random forest, gradient-boosted trees).
• Utilized composite pain assessment data from 103 patients.
• Internal and external validation of models against 1-year composite outcomes and lead trial results.

Main Results:

• Most developed ML models outperformed traditional lead trialing in predicting SCS efficacy.
• The regularized logistic regression (RLR) model demonstrated an optimal balance of complexity and interpretability.
• AI-based predictive medicine shows promise in enhancing clinical decision-making for SCS.

Conclusions:

• Machine learning models offer a powerful tool for predicting long-term SCS efficacy.
• The RLR model is a strong candidate for clinical application due to its performance and interpretability.
• AI-driven predictive medicine can synergistically support clinicians in optimizing SCS treatment choices.