Abstract: Objective To analyze the risk factors for recurrence of trigeminal neuralgia(TN) after percutaneous balloon compression(PBC),construct a recurrence risk prediction model,and verify its prediction effect. Methods Retrospective collection of data from January 2020 to December 2023 on 317 TN patients treated with PBC at the Fourth Affiliated Hospital of Harbin Medical University was conducted to form the modeling group.Univariate analysis and logistic regression analysis were employed to screen for risk factors for TN recurrence after PBC.Three machine learning algorithms(Logistic regression,random forest,and XGBoost) were utilized in R software to construct predictive models,and their performance was compared.The optimal algorithm for predicting TN recurrence after PBC was determined,followed by external validation of the model using data from 131 patients treated at the First Affiliated Hospital of Harbin Medical University from January 2020 to December 2023 as the validation group. Results The logistic regression analysis indicated that a disease duration > 5 years,non-typical pain,non-pear-shaped balloon,compression time > 120 seconds,and delayed disappearance of pain are independent risk factors for the TN recurrence following PBC(P<0.05).On the other hand,facial numbness serves as a protective factor against recurrence in patients undergoing this procedure (OR=0.289,95%CI:0.143~0.582).Furthermore,the random forest model exhibited superior performance compared to the other two predictive models,with areas under the ROC curve of 0.824(95%CI:0.774~0.873) for the modeling group and 0.835(95%CI:0.763~0.892) for the validation group,indicating its greater efficacy in predicting postoperative recurrence. Conclusion The model based on the random forest algorithm is the optimal predictive model for TN recurrence after PBC.It is beneficial for clinical screening of high-risk groups for recurrence of trigeminal neuralgia post-surgery.This can provide a reference for medical staff to take targeted preventive measures early.

Key words: Trigeminal Neuralgia, Percutaneous Balloon Compression, Recurrence, Machine Learning, Predictive Model

摘要： 目的 分析经皮微球囊压迫术(percutaneous balloon compression,PBC)后三叉神经痛(trigeminal neuralgia,TN)复发的危险因素,构建复发风险预测模型,并验证其预测效果。方法 回顾性收集2020年1月~2023年12月于哈尔滨医科大学附属第四医院行PBC治疗的317例TN患者为建模组,通过单因素分析和Logistic回归分析筛选PBC术后复发的危险因素,使用R软件中Logistic回归、随机森林、XGBoost 3种机器学习算法构建预测模型,并对其性能进行比较,得到最优的PBC术后TN复发的预测模型,并回顾性收集2020年1月~2023年12月于哈尔滨医科大学附属第一医院行PBC治疗的131例TN患者为验证组对模型进行外部验证。结果 Logistic回归分析结果显示,病程＞5年、疼痛分型为非典型疼痛、球囊形状非梨形、压迫时间＞120 s、有延迟性疼痛消失是PBC术后TN复发的独立危险因素(P<0.05),而面部麻木是PBC术后TN患者复发的保护因素(OR=0.289,95%CI:0.143～0.582);随机森林模型建模组和验证组的ROC曲线下面积分别为0.824(95%CI:0.774~0.873)和0.835(95%CI:0.763~0.892),其性能均优于其他2种预测模型。结论 基于随机森林算法的模型是最优的PBC术后TN复发预测模型,有利于临床筛选术后TN复发高危人群,可为医护人员早期采取针对性的预防措施提供借鉴。

关键词: 三叉神经痛, 经皮微球囊压迫术, 复发, 机器学习, 预测模型