基于临床-病理-影像指标构建预测模型评估脑膜瘤术后脑水肿加重的风险

Value of a predictive model based on clinical-pathological-imaging indicators in assessing the aggravation of cerebral edema after meningioma surgery

摘要:

目的 基于临床-病理-影像指标构建预测模型对脑膜瘤患者术后脑水肿加重的风险进行评估。方法 回顾性分析2020年1月—2024年1月开封市中心医院100例接受脑膜瘤手术治疗患者的临床、病理及影像学资料，根据其术后脑水肿是否加重分为加重组（n=26）和未加重组（n=74）。多因素Logistic回归分析脑膜瘤患者术后脑水肿加重的影响因素，并以此构建Logistic回归预测模型，绘制受试者工作特征（ROC）曲线，采用ROC曲线下面积（AUC）评估临床-病理-影像指标构建模型对脑膜瘤患者术后脑水肿加重的预测价值。采用Bootstrap法对其进行内部验证。结果 100例脑膜瘤患者中，有26.00%的患者出现术后脑水肿加重。多因素Logistic回归分析结果显示，术前瘤周水肿（OR=1.687，95%CI：1.443~1.931）、肿瘤大小（OR=1.790，95%CI：1.653~1.926）、肿瘤部位（OR=1.840，95%CI：1.733~1.948）、CT增强形态（OR=1.811，95%CI：1.702~1.920）均为脑膜瘤患者术后脑水肿加重的影响因素（P＜0.05）。基于上述因素构建临床-病理-影像Logistic回归预测模型：Logistic（P）=-4.326+0.523（术前瘤周水肿）+0.582（肿瘤大小）+0.610（肿瘤部位）+0.594（CT增强形态），其对脑膜瘤患者术后脑水肿加重预测的敏感度、特异度及AUC分别为81.21%、90.58%及0.914（P<0.001）。对模型鲁棒性进行检测，风险模型敏感度为81.21%、特异度为90.58%、阳性预测值为85.30%、阴性预测值为88.10%、准确率为87.50%、精准率为85.30%、召回率为81.21%、综合评价指标（F1-Score）为83.20%，F1-Score变异系数CV=5.2%（<10%）。结论 基于术前瘤周水肿、肿瘤大小、肿瘤部位、CT增强形态构建的临床-病理-影像联合模型对预测脑膜瘤患者术后脑水肿加重的敏感度、特异度及AUC水平均较高，预测效能好，临床应用价值较高。

Abstract:

Objective To investigate the value of a predictive model based on clinical-pathological-imaging indicators in assessing the aggravation of cerebral edema after meningioma surgery.Methods A retrospective analysis was performed for the clinical， pathological， and imaging data of 100 patients with meningioma who received meningioma surgery in Kaifeng Central Hospital from January 2020 to January 2024， and according to the presence or absence of the aggravation of cerebral edema after surgery， the patients were divided into aggravation group with 26 patients and non-aggravation group with 74 patients. A logistic regression analysis was used to investigate the influencing factors for the aggravation of cerebral edema after surgery in patients with meningioma， and a logistic regression predictive model was established based on the above risk factors. The receiver operating characteristic （ROC） curve was plotted， and the area under the ROC curve （AUC） was used to assess the value of the model based on clinical-pathological-imaging indicators in predicting the aggravation of cerebral edema after surgery in patients with meningioma. The Bootstrap method was used for internal validation.Results Among the 100 patients with meningioma， 26.00% （26/100） had aggravation of cerebral edema after surgery. The multivariate logistic regression analysis showed that preoperative peritumoral edema （odds ratio ［OR］=1.687， 95% confidence interval ［CI］：1.443-1.931）， tumor size （OR=1.790，95%CI：1.653~1.926）， tumor location （OR=1.840，95%CI：1.733~1.948）， and CT-enhanced morphology （OR=1.811，95%CI：1.702~1.920） were risk factors for aggravation of cerebral edema after surgery in patients with meningioma. The logistic regression clinical-pathological-imaging predictive model based on the above factors was established as Logistic （P） = -4.326 + 0.523 （preoperative peritumoral edema） + 0.582 （tumor size） + 0.610 （tumor location） + 0.594 （CT-enhanced morphology）， with a sensitivity of 81.21%， a specificity of 90.58%， and an AUC of 0.914 in predicting the aggravation of cerebral edema after surgery in patients with meningioma. Robustness testing was performed for the risk prediction model with a sensitivity of 81.21%， a specificity of 90.58%， a positive predictive value of 85.30%， a negative predictive value of 88.10%， an accuracy rate of 87.50%， a precision rate of 85.30%， a recall rate of 81.21%， and a comprehensive evaluation index （F1-Score） of 83.20%， with a coefficient of variation of 5.2% （<10%） for F1-Score.Conclusions The clinical-pathological-imaging model constructed based on preoperative peritumoral edema， tumor size， tumor location， and CT-enhanced morphology has high sensitivity， specificity， and AUC in predicting the aggravation of brain edema after surgery in meningioma patients， with a good predictive performance and a high clinical application value.