改进型Adaboost算法在二次脑损伤上的应用  

Application of improved adaboost algorithm in diagnosis of secondary brain injury

作　　者：荆斌[1] 张鹏[1] 李巍[1] 唐剑飞[1] JING Bin;ZHANG Peng;LI Wei

机构地区：[1]解放军第307医院医学工程科,北京100071

出　　处：《中国医学装备》2012年第2期8-11,共4页China Medical Equipment

摘　　要：目的:设计一种有效的伤情评估与建模系统,用于衡量二次脑损伤因素的严重程度及预后期望,建立早期伤情评估和预测手段。方法:AdaBoost算法是提高预测学习系统能力的有效工具。采用Adaboost算法分别针对不同的训练集训练同一个基本分类器,然后把这些在不同训练集上得到的分类器集合起来,构成一个更强的最终的分类器算法。结合费氏二次脑损伤及哥拉斯科昏迷评价标准,建立规则库,将数据库技术与数据挖掘技术相结合,针对患者敏感数据进行数据挖掘分析,并评估伤情及进展。结果:本系统实现了二次脑损伤伤情的实时分析,并针对患者病情进行阶段性预测。较之回归分析,该系统长期(168h)临床预测准确度提高,加快了人工神经网络模型的收敛速度。结论:该系统具有诊断率高,操作便捷的优点,可为类似疾病的诊断和预测提供借鉴。Objective:To design an effective injury assessment ana moaenng system,which is used to measure the secondary brain injury factors reflecting the severity and of expectations of prognosis.Methods:Adaboost algorithm is an effective tool to improve the forecasting ability of the learning system.In this paper,Adaboost algorithm training set,respectively,for different basic training with a classifier,and then these get in the training setof different classifiers together to form a stronger final classifier algorithm.Combination oI tee ana GCS's seconaary brain injury coma evaluation criteria,establish the rule base.With the database and data mining technology,mine and analysis the patient sensitive data for assessment of injury and progress.Results:The system can achieve real-time analysis to secondary brain injury and periodic forecasts for the patient's condition.Compared to regression analysis,the system can improve the accuracy of clinical prediction in longterm(168hours)and speed up the convergence rate of artificial neural network model.Conclusion:The system has the advantages of high diagnostic rate,easy operation,a similar disease diagnosis and prediction for reference.

关 键 词：人工神经网络 诊断 二次脑损伤 ADABOOST算法 

分 类 号：R651[医药卫生—外科学]