摘要
目的探究基于平扫CT的Logistic回归模型和朴素贝叶斯(NB)模型预测自发性脑出血早期血肿扩大(HE)的价值。方法回顾性分析208例自发性脑出血患者的临床、初次CT扫描和24 h内随访CT资料,以随访CT显示血肿体积比例较前增加>33%或体积差>6 ml为HE,将患者分为HE组(86例)和非HE组(122例)。比较2组间临床资料和影像学特征的差异,采用Logistic回归和NB方法分别建立预测HE模型,并评估其预测性能。结果多因素方差分析表明男性患者(OR=3.814)、糖尿病病史(OR=0.442)、CT卫星征(OR=0.083)与漩涡征(OR=0.232)及伴脑室内出血(OR=0.442)是HE的独立预测因素。Logistic回归模型曲线下面积(AUC)为0.823,NB模型训练集和测试集的AUC分别为0.768和0.847。结论基于平扫CT的Logistic回归模型和NB模型有助于预测自发性脑出血早期HE,NB模型预测HE效能优于Logistic回归模型。
Objective To analyze the predictive effect of Logistic regression model and naive Bayesian(NB)model based on plain CT for predicting hematoma enlargement(HE)in the early stage of spontaneous intracerebral hemorrhage.Methods Data of 208 patients with spontaneous intracerebral hemorrhage who underwent initial CT scan within 6 h and follow-up CT scan within 24 h from symptom onset were retrospectively analyzed.HE were identified when the volume of hematoma increased by more than 33%or the volume increased by more than 6 ml from initial CT to follow-up CT scanning.The patients were divided into HE group(n=86)and non-HE group(n=122).Clinical data and CT image features were compared between two groups.A NB model and a Logistic regression model were established to predict HE,and their prediction performances were evaluated.Results Multivariate analysis showed that male(OR=3.814),diabetic(OR=0.442),satellite sign(OR=0.083),swirl sign(OR=0.232)and intraventricular hemorrhage(OR=0.442)were all independent predictors of HE.Area under the curve(AUC)of the Logistic regression model for predicting HE was 0.823,of NB model in the training set and testing set was 0.768 and 0.847,respectively.Conclusion Logistic regression model and NB model based on plain CT could predict HE in the early stage of spontaneous intracerebral hemorrhage.NB model was superior to the Logistic regression model for predicting HE.
作者
宋祖华
周治明
郭大静
唐茁月
李欣
SONG Zuhua;ZHOU Zhiming;GUO Dajing;TANG Zhuoyue;LI Xin(Department of Radiology,the Second Affiliated Hospital of Chongqing Medical University,Chongqing 400010,China;Department of Radiology,Chongqing General Hospital,Chongqing 400013,China)
出处
《中国医学影像技术》
CSCD
北大核心
2021年第1期30-34,共5页
Chinese Journal of Medical Imaging Technology
基金
重庆市科卫联合医学科研重点项目(2019ZDXM010)
重庆市基础与前沿研究计划项目(cstc2016jcyjA0294)
重庆市医学科研计划重点项目(20141016)。
关键词
脑出血
机器学习
体层摄影术
X线计算机
cerebral hemorrhage
machine learning
tomography,X-ray computed