Related Experiment Video
Updated: Aug 16, 2026
08:16
Intra-Operative Neural Monitoring of Thyroid Surgery in a Porcine Model
Published on: February 11, 2019
Predicting neurological outcomes following spinal surgery: A machine learning approach using intraoperative
Tamir Themans1, Valerie Ter Wengel2, Saskia van der Gaag3
1Department of Biomechanical Engineering, Delft University of Technology, Delft, The Netherlands.
North American Spine Society Journal
|August 14, 2026
Summary
Machine learning models using intraoperative neuromonitoring (IONM) data, including motor evoked potentials (MEPs) and somatosensory evoked potentials (SSEPs), can predict postoperative neurological outcomes after spinal surgery. Preoperative status and intraoperative signal latency are key predictors.
Area of Science:
- Neurosurgery
- Machine Learning
- Biomedical Engineering
Background:
- Intraoperative neuromonitoring (IONM) aids in reducing postoperative neurological complications.
- The precise value of IONM for predicting neurological outcomes requires further clarification.
- Machine learning (ML) offers an objective, real-time method for analyzing extensive IONM datasets.
Purpose of the Study:
- To develop and evaluate ML models for predicting postoperative neurological outcomes.
- To identify key predictive features from baseline characteristics and IONM data.
- To compare the predictive performance of different ML classifiers and IONM modalities (MEPs, SSEPs, combined).
Main Methods:
- Retrospective cohort study of 67 spinal surgery patients (2019-2023).
- Analysis of 260 IONM features and clinical data, including MEPs and SSEPs.
- Evaluation of four ML classifiers (SVM, KNN, Random Forest, XGBoost) using nested cross-validation.
Main Results:
- Extreme gradient boosting (XGBoost) demonstrated superior performance across all metrics.
- The combined MEP-SSEP model achieved the highest accuracy (87.1%), sensitivity (70.4%), and specificity (88.3%).
- Key predictors included preoperative neurological deficits (29%) and intraoperative signal latency changes (13.5%).
Conclusions:
- IONM features, particularly MEPs, significantly enhance the prediction of 3-month neurological outcomes.
- Accurate preoperative status documentation is crucial for model performance.
- Intraoperative signal latency change is a prominent predictive IONM feature.
