Logistic回归与人工神经网络应用于重度颅脑损伤继发认知功能障碍的预测评判
|
摘要
目的通过搜集重度颅脑损伤(traumatic brain injury,TBI)患者的临床样本资料,使用Logistic回归及人工神经网络,分别构建其继发认知功能障碍(Cognitive impairment,CI)的临床模型,并验证评判效能。方法连续性、回顾性收集2017年6—12月本院神经外科及急诊病房收治的重度颅脑外伤患者179例为研究对象;前瞻性、连续性搜集2018年1—8月本院神经外科收治的重度颅脑外伤患者23例为验证对象。基于研究对象的临床资料,分别构建Logistic回归及人工神经网络评判重度TBI继发CI的临床预测模型;并将验证对象的临床资料代入前述两种评判模型,基于其预后结果,绘制ROC曲线以验证评判效能。结果研究对象179例,其中36例患者继发CI;验证对象23例,其中7例患者继发CI。Logistic回归模型结果显示年龄、吸烟史、气管切开、监护室住院时间是重度TBI患者住院期间继发CI的独立影响因素。基于相同潜在影响因素,人工神经网络模型显示继发CI的自变量依照重要性分别为年龄、手术耗时>4 h、既往吸烟史、监护室住院时间、NSE含量。通过ROC曲线参数对比两种评判体系的总体效果,Logistic回归和人工神经网络模型的曲线下面积分别为0.866及0.897,总体预测效果均较好;彼此之间预测能力对比,人工神经网络模型的拟合效果相对更佳。结论 Logistic回归及人工神经网络应用于预测重度TBI继发CI均有良好的评判效能,相比之下人工神经网络模型能更加真实的反映出变量之间的真实关系。
Objective By collecting clinical sample data from patients with traumatic brain injury( TBI),using logistic regression and artificial neural networks,to construct a clinical model of secondary cognitive impairment( CI) to verify the effectiveness of the judgment.Methods Continuously,retrospectively collected the179 cases of patients with traumatic brain injury admitted to the neurosurgical and emergency wards of the first affiliated hospital of Bengbu Medical College from June 1 st,2016 to September 1 st,2017,as the study subjects.Prospectively and continuously collected 23 patients with severe craniocerebral trauma admitted to the department of neurosurgery from October 01,2018 to December 1,2018 as the confirmation subjects.Based on the clinical data of the study subjects,logistic regression and artificial neural network were used to evaluate the clinical prediction model of severe TBI secondary CI.And the clinical data of the verification object were substituted into the above two evaluation models. Based on its prognostic results,the ROC curve was plotted to verify the judgment performance.Results 36 study patients with secondary CI,7 validated subjects with secondary CI.The results of the logistic regression model showed that the age,smoking history,tracheotomy,and intensive care unit time were independent factors( including risk factors) for secondary CI during hospitalization for patients with severe TBI.Based on the same potential influencing factors,the artificial neural network model showed that the independent variables of secondary CI were age,duration of surgery over 4 h,history of previous smoking,length of hospital stay,NSE content according to their importance.Compare the overall effect of the two evaluation systems by ROC curve parameters.The area under the curve of logistic regression and artificial neural network model were 0. 866 and 0.897,respectively.Comparison of predictive power between each other,the fitting effect of the artificial neural network model was relatively better. Conclusions Logistic regression and artificial neural network have good evaluation performance for predicting severe TBI secondary CI.In contrast,the artificial neural network model could reflect the true relationship between variables more realistically.
Objective By collecting clinical sample data from patients with traumatic brain injury( TBI),using logistic regression and artificial neural networks,to construct a clinical model of secondary cognitive impairment( CI) to verify the effectiveness of the judgment.Methods Continuously,retrospectively collected the179 cases of patients with traumatic brain injury admitted to the neurosurgical and emergency wards of the first affiliated hospital of Bengbu Medical College from June 1 st,2016 to September 1 st,2017,as the study subjects.Prospectively and continuously collected 23 patients with severe craniocerebral trauma admitted to the department of neurosurgery from October 01,2018 to December 1,2018 as the confirmation subjects.Based on the clinical data of the study subjects,logistic regression and artificial neural network were used to evaluate the clinical prediction model of severe TBI secondary CI.And the clinical data of the verification object were substituted into the above two evaluation models. Based on its prognostic results,the ROC curve was plotted to verify the judgment performance.Results 36 study patients with secondary CI,7 validated subjects with secondary CI.The results of the logistic regression model showed that the age,smoking history,tracheotomy,and intensive care unit time were independent factors( including risk factors) for secondary CI during hospitalization for patients with severe TBI.Based on the same potential influencing factors,the artificial neural network model showed that the independent variables of secondary CI were age,duration of surgery over 4 h,history of previous smoking,length of hospital stay,NSE content according to their importance.Compare the overall effect of the two evaluation systems by ROC curve parameters.The area under the curve of logistic regression and artificial neural network model were 0. 866 and 0.897,respectively.Comparison of predictive power between each other,the fitting effect of the artificial neural network model was relatively better. Conclusions Logistic regression and artificial neural network have good evaluation performance for predicting severe TBI secondary CI.In contrast,the artificial neural network model could reflect the true relationship between variables more realistically.
引文
[1]张溢华,邱俊,王昊,等.338083例颅脑损伤流行病学特点分析[J].创伤外科杂志,2016,18(6):328-330.
[2]邬树凯,陈峻严,高宏志,等.重型创伤性颅脑损伤后认知功能障碍特征及其影响因素的初步研究[J].中国医师进修杂志,2011,34(29):26-29.
[3] Pieruccini-Faria F,Muir-Hunter SW,Montero-Odasso M. Do depressive symptoms affect balance in older adults with mild cognitive impairment? Results from the“gait and brain study”[J].Exp Gerontol,2018,108:106-111.
[4] Gauthier S,Reisberg B,Zaudig M,et al.Mild cognitive impairment[J].Lancet,2006,367(9518):1262.
[5]贺宪民.医学多元资料分析中的人工神经网络方法研究[D].第二军医大学,2001.
[6]包龙,徐峰,丁礼,等.AIS-ISS评分对创伤性TBI患者的预后价值[C].//中华医学会急诊医学分会第十八次全国急诊医学学术年会论文集.苏州大学附属第一医院,2015:42.
[7]中华医学会神经病学分会,痴呆与认知障碍学组.血管性认知障碍诊治指南[J].中华神经科杂志,2011,44(2):142-150.
[8]刘华,李兵,阮海林,等.GCS、ISS、RTS对TBI合并多发伤患者预后评估的价值[J].山东医药,2015,55(21):4-7.
[9]汪正光,张牧城,黄嵘,等.全面无反应性量表评分对TBI患者90天预后的评估价值[J].中国全科医学,2015,18(26):3208-3210.
[10]郑兴珍,朱简,于强.APACHE-Ⅱ评分在临床应用的现状及进展[J].医学综述,2011,17(21):3297-3299.
[11] Ramesh VG,T hirumaran KP,Chockaiah RM. A new scale for prognostication in head in jury[J]. J Clin Neurosci,2008,15(10):1110-1113.
[12] Andersen J,Kot N,Ennis N,et al. Traumatic brain injury and cognitive impairment in men who are homeless[J]. Disabil Rehabil,2014,36(26):2210-2215.
[13]高长城.人工神经网络模型-神经元矩阵[J].电子技术与软件工程,2014,(17):48.
[14]毛华,陈涛,刘强,等.数据挖掘技术在建立颅脑损伤患者预后预测模型中的应用[J].临床外科杂志,2013,21(9):694-696.
[15]姚顺,徐国政.弥漫性轴索损伤的神经网络研究进展[J].中国临床神经外科杂志,2015,(1):52-54.
[2]邬树凯,陈峻严,高宏志,等.重型创伤性颅脑损伤后认知功能障碍特征及其影响因素的初步研究[J].中国医师进修杂志,2011,34(29):26-29.
[3] Pieruccini-Faria F,Muir-Hunter SW,Montero-Odasso M. Do depressive symptoms affect balance in older adults with mild cognitive impairment? Results from the“gait and brain study”[J].Exp Gerontol,2018,108:106-111.
[4] Gauthier S,Reisberg B,Zaudig M,et al.Mild cognitive impairment[J].Lancet,2006,367(9518):1262.
[5]贺宪民.医学多元资料分析中的人工神经网络方法研究[D].第二军医大学,2001.
[6]包龙,徐峰,丁礼,等.AIS-ISS评分对创伤性TBI患者的预后价值[C].//中华医学会急诊医学分会第十八次全国急诊医学学术年会论文集.苏州大学附属第一医院,2015:42.
[7]中华医学会神经病学分会,痴呆与认知障碍学组.血管性认知障碍诊治指南[J].中华神经科杂志,2011,44(2):142-150.
[8]刘华,李兵,阮海林,等.GCS、ISS、RTS对TBI合并多发伤患者预后评估的价值[J].山东医药,2015,55(21):4-7.
[9]汪正光,张牧城,黄嵘,等.全面无反应性量表评分对TBI患者90天预后的评估价值[J].中国全科医学,2015,18(26):3208-3210.
[10]郑兴珍,朱简,于强.APACHE-Ⅱ评分在临床应用的现状及进展[J].医学综述,2011,17(21):3297-3299.
[11] Ramesh VG,T hirumaran KP,Chockaiah RM. A new scale for prognostication in head in jury[J]. J Clin Neurosci,2008,15(10):1110-1113.
[12] Andersen J,Kot N,Ennis N,et al. Traumatic brain injury and cognitive impairment in men who are homeless[J]. Disabil Rehabil,2014,36(26):2210-2215.
[13]高长城.人工神经网络模型-神经元矩阵[J].电子技术与软件工程,2014,(17):48.
[14]毛华,陈涛,刘强,等.数据挖掘技术在建立颅脑损伤患者预后预测模型中的应用[J].临床外科杂志,2013,21(9):694-696.
[15]姚顺,徐国政.弥漫性轴索损伤的神经网络研究进展[J].中国临床神经外科杂志,2015,(1):52-54.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |