Risk Factors and an Interpretable Machine Learning Model for Predicting Spinal Epidural Lipomatosis: A Multicenter...

Area of Science:

• Spinal health and neurosurgery
• Medical imaging and diagnostics
• Machine learning in healthcare

Background:

• Spinal Epidural Lipomatosis (SEL) is a significant, underdiagnosed cause of lumbar spinal stenosis.
• Current tools for early SEL identification and risk stratification are insufficient.
• Developing predictive models is crucial for timely diagnosis and management.

Purpose of the Study:

• To identify independent risk factors for Spinal Epidural Lipomatosis (SEL).
• To develop and validate an interpretable machine learning-based predictive model for SEL.
• To create a clinically accessible tool for risk stratification.

Main Methods:

• Retrospective multicenter study involving 774 patients with low back and leg pain.
• LASSO regression for variable selection and development of a predictive nomogram.
• Four machine learning models were constructed and evaluated using AUC, calibration, and decision curve analysis.

Main Results:

• Seven independent predictors for SEL were identified: elevated random blood glucose, blood type B, atherosclerosis index, body mass index, uric acid, obstructive sleep apnea, and age.
• The XGBoost model showed superior predictive performance (AUC: 0.726) in the validation set.
• Interpretability analysis highlighted glucose, age, and uric acid as key contributors to risk prediction.

Conclusions:

• An interpretable prediction model integrating clinical factors and an XGBoost algorithm was developed and validated.
• The model provides an actionable nomogram to assist clinicians in early SEL detection.
• This tool supports risk-stratified management and targeted interventions for underdiagnosed SEL.