[1]冯金周,刘发健,江华.颅脑损伤患者临床死亡预测:一项基于机器学习的主成分分析-逻辑回归模型[J].临床神经外科杂志,2019,16(2):99-103.[doi:10.3969/j.issn.1672-7770.2019.02.002]
 FENG Jin-zhou,LIU Fa-jian,JIANG Hua..Establishment of clinical death prediction in patients with craniocerebral injury: A PCA-Logistic regression model based on machine learning[J].Journal of Clinical Neurosurgery,2019,16(2):99-103.[doi:10.3969/j.issn.1672-7770.2019.02.002]
点击复制

颅脑损伤患者临床死亡预测:一项基于机器学习的主成分分析-逻辑回归模型()
分享到:

《临床神经外科杂志》[ISSN:1672-7770/CN:32-1727/R]

卷:
16
期数:
2019年第2期
页码:
99-103
栏目:
转化医学专题
出版日期:
2019-04-17

文章信息/Info

Title:
Establishment of clinical death prediction in patients with craniocerebral injury: A PCA-Logistic regression model based on machine learning
作者:
冯金周刘发健江华
610101 成都,四川省医学科学院·四川省人民医院东院神经外科(冯金周,刘发健);创伤与代谢多学科实验室(冯金周,江华)
Author(s):
FENG Jin-zhou LIU Fa-jian JIANG Hua.
Department of Neurosurgery, Sichuan Provincial People's Hospital, Sichuan Academy of Medical Sciences, Chengdu 610101, China
关键词:
重度颅脑损伤模式识别主成分分析逻辑回归机器学习
Keywords:
traumatic brain injurypattern recognitionprincipal component analysislogistic regressionmachine learning
分类号:
R651.1+5
DOI:
10.3969/j.issn.1672-7770.2019.02.002
文献标志码:
A
摘要:
【摘要】目的 探索主成分分析(principal component analysis,PCA)-逻辑回归模型在颅脑损伤患者临床死亡预测建模中的应用,以及影响临床预后的病理生理模式和重要风险因素。方法收集2011—2017年四川省人民医院创伤中心数据库符合研究标准的108例颅脑损伤患者的临床资料,建立PCA-逻辑回归模型,应用ROC评估模型预测效果,验证死亡结局模型的预测效能。结果PCA-逻辑回归模型分析结果显示,影响患者死亡结局的分别为第1、第8、第11及12主成分。计算出的指标系数对应的临床指标与措施具有较大影响力的因素,分别为开放性颅脑损伤、凝血改变、气管切开、脑干伤、血肿量、感染性并发症、糖皮质激素、肠内营养时间及舒张压。经ROC曲线评估PCA-逻辑回归模型,死亡结局模型具有较高的预测效能(灵敏度:92.3%,特异度:93.7%,AUC:0.983)。结论PCA-逻辑回归分析方法可以有效地挖掘颅脑损伤患者伤后的临床变量,建立其临床死亡预测模型。严重颅脑损伤后出现的血流灌注不足可能是影响患者生存的重要病理生理模式。
Abstract:
Abstract: ObjectiveTo explore the application of PCA-logical regression model in predicting clinical death in patients with brain injury and to find the pathophysiological patterns and important risk factors that affect clinical prognosis. MethodsThe linical data of 108 brain injury patients who met the research criteria in the database of the trauma center of Sichuan Provincial People's Hospital from 2011 to 2017 were collected. The PCA-logical regression model was established, and the prediction effect of the death outcome model was verified by using the ROC model. ResultsThe PCA logistic regression model analysis found that the impact of patients' death outcomes were the first, eighth, eleventh and twelfth main components, respectively. The calculated index coefficient corresponding to the indications and measures had a greater influence on the open cranial brain damage, coagulation changes, tracheotomy, brain stem injury, hematoma volume, infectious complications, glucocorticoids, EN use time, and diastolic blood pressure. The PCA logistic regression model was evaluated by ROC curve. The PCAR model were able to identify the risk factors and forecast the clinical outcomes(Mortality, sensitivity 92.3%, specificity 93.7%, AUC 0.983). ConclusionsPCA logic regression analysis can effectively explore the clinical variables of patients with craniocerebral injury and establish a prognostic model of clinical death. The insufficiency of blood flow perfusion after severe craniocerebral injury may be an important pathophysiological model that affects the survival of patients.

相似文献/References:

[1]李玉伟,雷燕妮,陈必耀,等.闭合性重度颅脑损伤患者开颅术后颅内感染相关因素分析及对策[J].临床神经外科杂志,2015,(04):253.
[2]黄培赞,赵应群,宋永福,等.TCD对重度颅脑损伤去骨瓣减压术围手术期血流动力学监测的意义[J].临床神经外科杂志,2017,14(06):466.
 HUANG Pei-zan,ZHAO Ying-qun,SONG Yong-fu,et al.Significance of TCD perioperative hemodynamic monitoring on severe brain injury patients who accepted decompressive craniectomy[J].Journal of Clinical Neurosurgery,2017,14(2):466.

更新日期/Last Update: 2019-04-17