Predicting Postoperative Neurological Outcomes in Metastatic Spinal Tumor Surgery Using Machine Learning

Area of Science:

• Neurosurgery
• Oncology
• Medical Informatics

Background:

• Spinal metastases are increasing, necessitating surgical intervention for instability and neurological deficits.
• Accurate prediction of postoperative neurological status (Frankel classification) is crucial for surgical planning and patient counseling.
• Traditional prognostic models struggle to capture the complexity of neurological recovery.

Purpose of the Study:

• To develop machine learning models for predicting one-month postoperative neurological outcomes in patients with metastatic spinal tumors.
• To identify key factors influencing neurological recovery after surgery for spinal metastases.
• To enhance surgical decision-making and patient counseling through improved outcome prediction.

Main Methods:

• Retrospective analysis of data from 244 patients across 38 institutions.
• Development of predictive models using Random Forest, XGBoost, LightGBM, and CatBoost algorithms.
• Feature selection via Boruta algorithm and Variance Inflation Factor analysis to identify significant predictors.

Main Results:

• The proportion of ambulatory patients (Frankel grades D or E) increased from 36.8% to 63.1% postoperatively.
• The Random Forest model demonstrated the highest predictive accuracy with an AUC-ROC of 0.8516.
• Preoperative Frankel classification, transfer ability, inflammatory markers, and surgical timing were identified as key predictors.

Conclusions:

• Machine learning models offer robust prediction of postoperative neurological status in metastatic spinal tumor patients.
• Preoperative neurological function, functional status, and inflammation markers are critical determinants of patient outcomes.
• These findings can refine surgical strategies, set realistic expectations, and improve patient care through precise outcome prediction.