摘要
背景:目前对心律失常的诊断大多是由医生人工完成,费时费力,诊断结果依赖于医生的个人业务水平和责任心。心律失常的自动识别对于心脏病患者的救护和早期治疗具有非常重要的意义。目的:实现临床心律失常的自动识别和诊断。方法:首先从心电图中动态提取完整心律失常心拍形态,并采用离散余弦变换和反变换压缩数据;然后设计用于心律失常识别的BP神经网络,并用DNA算法优化该BP网络;最后用MIT/BIH心电数据库中心电图数据对DNA-BP网络进行检验。结果与结论:对于5种心拍类型,包括正常、左束支阻滞、右束支阻滞、心室跳脱心搏及Paced心搏,利用DNA-BP网络进行分类,实验达到了很好的识别效果,平均识别正确率达到99%。
BACKGROUND：Arrhythmia is commonly diagnosed by doctors,and the diagnosis depends on doctor experience and responsibility. Arrhythmia recognition is great important for rescuing and early treatment of cardiopathy sufferers. OBJECTIVE：To investigate efficient method of auto-recognizing and diagnosis of arrhythmia. METHODS：The whole morphology of arrhythmic heartbeat was extracted from electrocardiograph （ECG）,and discrete cosine transform and inverse discrete cosine transform were used to compress the data. A BP neural network was designed for arrhythmic heartbeat recognition,and the initial weights and thresholds of network were optimized by DNA algorithm. Finally,the MIT/BIH ECG database was used to test the DNA-BP neural network. RESULTS AND CONCLUSION：For the five types heart beat,including normal,left bundle branch block,right bundle branch block,ventricular escape heartbeat,Paced heartbeat,the experiment results demonstrate efficient by using DNA-BP neural network,with an average recognition rate of 99%.
出处
《中国组织工程研究与临床康复》
CAS
CSCD
北大核心
2010年第39期7353-7357,共5页
Journal of Clinical Rehabilitative Tissue Engineering Research
基金
国家自然科学基金(60841004):基于基函数超完备集的动物视觉图像重构研究
国家自然科学基金(60971110):初级视觉皮层中视像整体特征的稀疏表象模型的研究~~
