基于多模态CT建立急性缺血性卒中预后不良的列线图预测模型

Establishment of a nomogram prediction model for poor prognosis of acute ischemic stroke based on multimodal CT

目的:本研究旨在建立基于多模态CT的列线图模型以预测急性缺血卒中预后不良发生的风险。方法:回顾性收集193例急性缺血性卒中患者的一般资料及多模态CT指标,根据3月后的改良Rankin量表(mRS)评分分为预后良好和预后不良两组。通过多因素逐步Logistic回归分析以确定急性缺血性卒中预后不良发生的独立危险因素,构建列线图模型。使用校准曲线评价模型的拟合优度,采用受试者工作特征(ROC)曲线评价模型的预测效能,并采用决策曲线分析(DCA)评价模型的临床应用价值。结果:多因素逐步Logistic回归分析结果显示基线美国国立卫生研究院卒中量表(NIHSS)评分、中性粒细胞计数与淋巴细胞计数比值(NLR)、梗死核心体积及水摄取率(NWU)是急性缺血性卒中预后不良的独立危险因素。列线图模型显示出良好的校准和区分能力,ROC曲线下面积(AUC)为0.956(95%CI:0.927~0.985)。DCA表明其具有临床实用价值。结论:基于多模态CT构建的列线图对急性缺血卒中预后不良的发生具有良好的预测效能,能为临床个性化诊疗决策的制定提供帮助。

Objective:To establish a nomogram model based on multimodal computed tomography(CT)to predict the risk of poor prognosis in acute ischemic stroke.Method:The general data and multimodal CT indexes of 193 patients with acute ischemic stroke were collected retrospectively.According to the modified Rankin scale(mRS)score 3 months later,the patients were divided into two groups:good prognosis and poor prognosis.Multivariate stepwise Logistic regression was used to determine the independent risk factors for poor prognosis of acute ischemic stroke.The goodness of fit of the model was evaluated by calibration curve,the predictive efficiency of the model was evaluated by receiver operating characteristic(ROC)curve,and the clinical application value of the model was evaluated by decision curve analysis(DCA).Results:Multivariate Logistic regression analysis showed that baseline National Institutes of Health Stroke scale(NIHSS)score,neutrophil/lymphocyte count ratio(NLR),core infarction volume and net water uptake(NWU)were independent risk factors for poor prognosis of acute ischemic stroke.The nomogram model showed good calibration and discrimination ability,and the area under the ROC curve(AUC)was0.956(95%CI:0.927~0.985).DCA showed that it had clinical practical value.Conclusion:The nomogram model based on multimodal CT has good clinical prediction efficiency for poor prognosis in acute ischemic stroke,and can provide help for clinical personalized diagnosis and treatment decision-making.