基于CT平扫放射学及临床特征对脑出血早期血肿扩大的预测价值  

Predictive value of early hematoma expansion in cerebral hemorrhage based on CT plain radiological and clinical features

作　　者：伍发 杨钰林 伍婷婷 蒋锐 王鹏 于红梅 杨金刚 李果 盛金平 WU Fa;YANG Yu-lin;WU Ting-ting;JIANG Rui;WANG Peng;YU Hong-mei;YANG Jin-gang;LI Guo;SHENG Jin-ping(Western Theater General Hospital,Chengdu 610083,China;Department of Ultrasound,Chengdu No.5 People's Hospital,Chengdu 611100,China)

机构地区：[1]西部战区总医院,四川成都610083 [2]成都市第5人民医院超声科,四川成都611100

出　　处：《中国临床医学影像杂志》2024年第12期837-841,共5页Journal of China Clinic Medical Imaging

基　　金：西部战区总医院助推基金(2019ZT09);西部战区总医院院管课题(2021-XZYG-C04、2021-XZYG-C05、2024-YGLC-B07);四川省医学会(恒瑞)科研基金专项科研课题(2021HR75)。

摘　　要：目的:通过基于CT平扫放射学及临床特征构建脑出血(sICH)血肿扩大(HE)的预测模型,并对该模型进行验证。方法:选取符合要求的244例sICH患者,按照血肿是否扩大分为HE组(65例)和未扩大(NHE)组(179例),对两组患者的CT征象及临床特征进行Lasso-Logistic回归分析,构建放射学及临床特征联合预测模型,将选出的独立危险因素构建列线图模型(Nomogram)并验证;另选取85例患者进行内部验证其预测结果。结果:首次CT检查时间、抗凝药物治疗史、低密度征、旋涡征、岛征、卫星征是预测HE的独立危险因素;联合预测模型的曲线下面积(AUC)为0.774,约登指数为0.4211,灵敏度为53.85%,特异度为88.27%,准确度为81.15%;内部验证模型的AUC为0.890,约登指数为0.6287,灵敏度为89.66%,特异度为73.21%,准确度为81.18%。结论:放射学及临床特征联合模型对预测HE有一定的价值,可为临床制定有效的治疗措施提供科学依据。Objective:To construct a prediction model of hematoma expansion(HE)in cerebral haemorrhage(sICH)based on CT scanning radiological and clinical features,and to validate the model.Methods:Two hundred and forty-four patients with cerebral haemorrhage who met the requirements were selected and divided into 65 cases in the HE group and 179 cases in the non-expansion(NHE)group according to whether the haematoma was enlarged or not.Lasso-Logistic regression analyse was performed on the CT signs and clinical characteristics of the patients in the two groups to construct the joint prediction model of radiological and clinical characteristics,and a Nomogram model was constructed with the selected independent risk factors and validated,furthermore,the predictive results were also validated by an internal study of 85 patients.Results:The time of the first CT examination,history of anticoagulant therapy,hypodensity sign,vortex sign,island sign,satellite sign were the independent risk factors for the prediction of HE,the joint prediction model had an area under the curve(AUC)value of 0.774 and a Yoden index of 0.4211,with a sensitivity of 53.85%,a specificity of 88.27%,and an accuracy of 81.15%,the internally validated prediction model had an AUC value of 0.890,a Jordon index of 0.6287,a sensitivity of 89.66%,a specificity of 73.21%,and an accuracy of 81.18%.Conclusion:The joint model of radiological and clinical features is valuable in predicting HE,which can provide scientific basis for the clinical development of effective therapeutic measures.

关 键 词：脑出血 血肿 体层摄影术 X线计算机 

分 类 号：R743.34[医药卫生—神经病学与精神病学] R814.42[医药卫生—临床医学]