Analysis of Risk Prediction Model for Recurrence of Trigeminal Neuralgia After Percutaneous Balloon Compression

TL;DR

This study develops a machine learning model to predict the recurrence of trigeminal neuralgia after a specific surgical treatment, helping improve clinical outcomes.

Contribution

A novel random forest-based risk prediction model for trigeminal neuralgia recurrence after percutaneous balloon compression is proposed and validated.

Findings

Random forest outperformed logistic regression and XGBoost in predicting TN recurrence.

Duration of disease, pain type, and balloon shape were significant predictors of recurrence.

The model showed strong external validation performance with an AUC of 0.835.

Abstract

Trigeminal neuralgia (TN) is a debilitating disorder characterized by severe facial pain. While percutaneous balloon compression (PBC) is an effective surgical treatment for TN, recurrence remains a significant concern, with varying reported rates. The identification of factors that contribute to recurrence after PBC is critical for improving treatment outcomes. However, existing predictive models for recurrence have limitations in accuracy and generalizability. This study aims to explore the influencing factors of TN recurrence after PBC and to construct a TN recurrence risk prediction model.

The clinical data of 448 TN patients treated for PBC were retrospectively analyzed and divided into a modeling group (n = 317) and a validation group (n = 131) in a ratio of 7:3. Patients were divided into two groups based on whether they experienced recurrence or not. Risk prediction models were constructed using three machine learning methods: logistic regression, random forest, and XGBoost. The area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity were used to evaluate the model performance.

Multivariate analysis showed that the duration of disease, pain type, balloon shape, compression time, and delayed disappearance of pain were influencing factors for TN recurrence after PBC, while facial numbness was a protective factor. All three predictive models exhibit high accuracy. In the modeling group, the AUC values for the logistic regression, random forest, and XGBoost models are 0.810, 0.824, and 0.816, respectively. Furthermore, the random forest model outperforms the other two models in terms of accuracy, sensitivity, and specificity. Additionally, external validation also demonstrates that the random forest model has good predictive value for TN after PBC (AUC = 0.835).

The random forest model showed excellent performance in predicting TN recurrence after PBC, providing a powerful reference for clinical prevention.

Figures