摘要：

目的: 基于影像组学筛查原发性三叉神经痛相关危险因素。方法: 选择2017年1月至2020年12月广东省第二人民医院诊断与治疗的48例原发性三叉神经痛患者，测量平均动脉压，行头部MRI检查构建神经血管压迫模型。单因素和多因素Logistic回归分析筛查原发性三叉神经痛相关危险因素，根据危险因素绘制受试者工作特征（ROC）曲线，评价其预测效能。结果: 共48例患者平均动脉压为56.89~120.44 mm Hg，平均为（94.32 ± 11.34）mm Hg；42例（87.50%）病灶位于单侧、6例（12.50%）病灶位于双侧，计54例次神经血管压迫模型。根据是否引发三叉神经痛分为患侧组（40例次）和健侧组（14例次），患侧组神经血管压迫面积（Z=-2.823，P=0.005）和神经血管压力（Z=-0.365，P=0.006）均大于健侧组。Logistic回归分析显示，神经血管压力大是原发性三叉神经痛的危险因素（OR=1.001，95%CI：1.0003~1.0022；P=0.011）。ROC曲线显示，神经血管压迫面积预测原发性三叉神经痛的曲线下面积为0.747（95%CI：0.605~0.890，P=0.006），灵敏度为42.50%、特异度为100%，截断值为25.34 mm²；神经血管压力的曲线下面积为0.755（95%CI：0.616~0.895，P=0.005），灵敏度为67.50%、特异度为78.60%，截断值为1672.99 mm Hg·mm²；且二者预测效能相当（Z=-0.250，P=0.805）。结论: 神经血管压迫面积和神经血管压力对原发性三叉神经痛具有重要诊断价值。

关键词: 三叉神经痛, 磁共振成像, 影像组学（非MeSH词）, 危险因素, Logistic模型, ROC曲线

Abstract:

Objective: To screen the risk factors related to primary trigeminal neuralgia (PTN) based on radiomics. Methods: A total of 48 patients with PTN admitted to Guangdong Second Provincial General Hospital from January 2017 to December 2020 were selected. The mean arterial pressure (MAP) of the patients was measured, and the neurovascular compression model was constructed by head MRI examination. Univariate and multivariate Logistic regression analyses were used to screen for risk factors associated with PTN, and the predictive efficacy was evaluated by receiver operating characteristic (ROC) curve according to the risk factors. Results: Among 48 patients, 42 patients (87.50%) had unilateral lesions and 6 patients (12.50%) had bilateral lesions. The MAP was 56.89-120.44 mm Hg, with an average of (94.32 ± 11.34) mm Hg. The neurovascular compression model of 54 cases was divided into the affected side (n = 40) and the healthy side (n = 14) according to whether the disease occurred. The neurovascular compression area (Z =-2.823, P = 0.005) and neurovascular pressure (Z =-0.365, P = 0.006) on the affected side were greater than those on the healthy side. Logistic regression analyses showed that high neurovascular pressure (OR = 1.001, 95%CI: 1.0003-1.0022; P = 0.011) was showed the area under the curve (AUC) of neurovascular compression area for predicting PTN was 0.747 (95%CI: 0.605-0.890, P = 0.006), the sensitivity was 42.50%, the specificity value was 25.34 mm². The AUC of neurovascular pressure was 0.755 (95%CI: 0.616-0.895, P = 0.005), the sensitivity was 67.50%, the specificity was 78.60%, and the cut-off value was 1672.99 mm Hg·mm². Conclusions: The neurovascular compression area and neurovascular pressure are important in the diagnosis of PTN.

Key words: Trigeminal neuralgia, Magnetic resonance imaging, Radiomics (not in MeSH), Risk factors, Logistic models, ROC curve