Machine Learning Prediction of Prevertebral Soft Tissue Swelling after Single-Level Anterior Cervical Surgery : A...

Area of Science:

• Neurosurgery
• Medical Informatics
• Machine Learning

Background:

• Prevertebral soft tissue swelling (PSTS) is a serious complication of anterior cervical spine surgery (ACSS).
• PSTS can cause dysphagia, dysphonia, and airway obstruction.
• Predicting PSTS is crucial for patient management.

Purpose of the Study:

• Develop and validate an interpretable machine learning model for predicting significant PSTS after single-level ACSS.
• Utilize a small, single-center dataset for model development.
• Assess model interpretability and clinical utility.

Main Methods:

• Retrospective analysis of 62 patients undergoing single-level ACSS.
• Defined significant PSTS as postoperative swelling > 7.0 mm.
• Developed an elastic net regularized logistic regression model with nested cross-validation and bootstrap validation.
• Assessed interpretability using SHAP and clinical utility via decision curve analysis.

Main Results:

• 16 out of 62 patients (25.8%) developed significant PSTS.
• The model achieved a bootstrap-validated AUC of 0.84.
• Key predictors identified: low preoperative serum albumin, surgery above C5, and male sex.
• Decision curve analysis showed clinical utility across probability thresholds of 15-45%.

Conclusions:

• The machine learning model effectively predicts significant PSTS risk after ACSS, even with a small sample size.
• Findings support AI-based risk stratification for PSTS prevention.
• External validation is recommended to enhance clinical relevance.