摘要
目的研究提取心电图(ECG)高频分量(100 Hz以上)中有价值的特征信息,改善不同阶段心肌梗死(myocardial infarction,MI)ECG的分类精度。方法采用PTB诊断数据库中的数据,包括:健康状态ECG、早期MI ECG和急性期MI ECG。通过多变量回归模型对宽带(wide band,WB)正交导联ECG信号建模提取特征,并进行基于支持矢量机的分类,同时展开与WB标准导联ECG分类的比较研究。结果MI分类精度随ECG带宽增加而提高,且从频宽为0.05 Hz^250 Hz的正交导联ECG中提取的分类特征最为高效。结论引入ECG中的高频分量能使MI分类精度得到改善,并取得理想的分类效果。
Objective To investigate the valuable features extracted from higher frequency components（ above100 Hz） in electrocardiogram（ ECG） signals so as to improve the classification accuracy of ECG’s in different myocardial infarction（ MI） stages. Methods The data for analysis was collected from PTB clinical diagnostic database including ECG in health control,MI ECG in early stage and MI ECG in acute stage. Multivariable autoregressive modeling technique was employed to model the wide band（ WB） orthogonal lead ECG signals,and the model coefficients were used as ECG features for the classification performed by support vector machine. The comparison with WB standard lead ECG signals was carried out in the same way. Results MI classification accuracy increased with the bandwidth. The features would be the best efficient representation of ECGs,which were extracted from WB orthogonal lead ECG signals with a frequency range of 0. 05 Hz ~ 250 Hz. Conclusion MI classification accuracy can be improved by introducing higher frequency components and good classification performance can be obtained.
出处
《航天医学与医学工程》
CAS
CSCD
北大核心
2015年第5期326-331,共6页
Space Medicine & Medical Engineering
基金
浙江省自然科学基金资助项目(LY14E050016)
国家自然科学基金资助项目(61074143)
关键词
心肌梗死
宽带心电图
特征提取
分类
myocardial infarction
wide band electrocardiogram
feature extraction
classification
