摘要
目的动脉瘤性蛛网膜下腔出血(aSAH)发病后的神经功能缺失与aSAH出血量之间存在显著相关性。临床多使用基于CT影像评估的改良Fisher分级来评估aSAH的出血量。本研究拟使用机器学习技术建立可能影响aSAH出血量的相关危险因素预测模型,通过模型帮助临床医生更好地诊断和预防疾病。方法本研究共纳入155例aSAH患者,以回顾分析的方法收集相关临床数据,统计可能导致aSAH出血量增加的相关危险因素,其中包括人口信息学因素、既往病史、院前血压水平和动脉瘤影像结构的相关特点等。剔除数据缺失的部分病历后,共有150例有效数据纳入机器学习模型进行分析,得出AUC,并用特征重要性排名分析影响动aSAH出血量的相关危险因素。结果研究团队使用一种机器学习方法-极度梯度提升树(XGBOOST),在aSAH样本中选择了14种可能影响出血量的危险因素作为自变量,aSAH患者出血量的评判采用改良Fisher分级为作为因变量,用机器学习预测后得出受试者工作曲线曲线下面积为0.84。结论利用XGBOOST中特征重要性排序得出,年龄是最强的危险因素,说明年龄对动脉瘤患者出血量的影响可能较大,动脉瘤体颈比、动脉瘤最大横径、患者院前血压中的舒张压、动脉瘤瘤颈长度、患者院前血压中的收缩压和脉压等因素被认为是重要特征。患者院前血压中的舒张压与改良Fisher分级有一定的正相关关系。使用机器学习有助于临床医生识别aSAH高危患者,具备疾病的诊疗与预防价值。
Objective There is a significant correlation between neurological deficits after the onset of aneurysmal subarachnoid haemorrhage(aSAH)and the amount of aSAH bleeding.The modified Fisher grading based on CT is often used to evaluate the extent of aSAH bleeding.The purpose of this study is to use machine learning technology to establish a prediction model of risk factors that affect the amount of aSAH bleeding,and help clinicians diagnose and treat the disease.Methods This study included 155 patients with aSAH to investigate the relevant risk factors that may contribute to increased bleeding in aSAH,including demographic factors,medical history,pre-hospital blood pressure levels,and characteristics of the aneurysm.After excluding patients with missing data,a total of 150 data were included in the machine learning model analysis to obtain the area under the curve(AUC)and analyze the ranking of feature importance in influencing the amount of aSAH.Results We used an extreme gradient boosting tree(XGBOOST),a machine learning method,to compare 14 influencing factors in samples of aSAH.By using the modified Fisher grading as the dependent variable for aSAH patients and predicting with machine learning,we obtained an AUC of 0.84.Conclusion Based on the feature importance ranking in XGBOOST,age is identified as the strongest risk factor,indicating that age may have a significant impact on aSAH.Factors such as aneurysm size/neck ratio,maximum aneurysm diameter,diastolic pressure,aneurysm neck length,systolic pressure,and pulse pressure are considered important features.There was a positive correlation between diastolic blood pressure and modified Fisher grading.The use of machine learning aids clinicians in identifying high-risk patients with aSAH and strengthens medical care.
作者
陈磊
赵迪
薛贵生
张伟航
焦建伟
赵照
贺峰
Chen Lei;Zhao Di;Xue Guisheng;Zhang Weihang;Jiao Jianwei;Zhao Zhao;He Feng(Department of Radiology,the Second Hospital,of Hebei Medical University,Shijiazhuang 050000,China)
出处
《脑与神经疾病杂志》
CAS
2024年第11期696-701,共6页
Journal of Brain and Nervous Diseases
基金
河北省医学科学研究计划课题(20240986)。
