A Novel Preoperative Scoring System to Accurately Predict Cord-Level Intraoperative Neuromonitoring Data Loss During...

Area of Science:

• Neurosurgery
• Orthopedic Surgery
• Machine Learning in Medicine

Background:

• Accurate prediction of intraoperative neuromonitoring (IONM) data loss is crucial for spinal deformity correction.
• No existing prediction tool addresses the risk of cord-level IONM data loss.

Purpose of the Study:

• To develop and validate a machine learning-derived scoring system to predict cord-level IONM data loss in spinal deformity surgery.
• To aid informed decision-making for patients undergoing spinal deformity correction.

Main Methods:

• A machine learning (ML) approach, including random forest (RF) analysis and multivariable logistic regression, was applied to 1,106 patients.
• Patients were divided into training (75%) and testing (25%) groups.
• A scoring system was developed using eight key predictive features identified through ML.

Main Results:

• The scoring system incorporates eight features, including sagittal deformity angular ratio (sDAR), spinal cord shape, conus medullaris level, and largest thoracic Cobb angle.
• Higher cumulative scores significantly correlated with increased rates of cord-level IONM data loss, ranging from 0.9% (score ≤2) to 86% (score ≥7).
• The system demonstrated high predictive performance with 93% accuracy, 75% sensitivity, 94% specificity, and an AUC of 0.898 in the testing group.

Conclusions:

• This study presents the first ML-derived preoperative scoring system for predicting cord-level IONM data loss in spinal deformity surgery.
• The developed scoring system achieves over 90% accuracy, offering a valuable tool for surgical planning and risk assessment.