摘要
提出基于主元分析的超声多普勒栓子信号检测方法,先从超声多普勒信号中提取出主元成分并获得相应的特征参数, 再结合时域的特征参数对栓子、干扰噪声和正常血流信号进行分类。通过对300例仿真超声多普勒信号和168例临床采集的脑动脉超声多普勒血流信号进行分析,结果表明:主元分析方法对仿真信号和临床信号的最小误判率分别为2％和5％,其分类的准确率相比常规的时域检测方法有了较大提高。可见,基于主元分析的超声多普勒栓子检测方法具有较高的检测率,克服了常规的时域和频域检测方法中由于干扰噪声的影响和频谱分析中时频分辨率的矛盾导致的灵敏度不高的局限,有望用于栓子的临床自动检测。
The Principal Component Analysis （PCA） algorithm is proposed to detect the embolic Doppler ultrasound signal. This method was applied to extract the main component of the Doppler ultrasound signal and calculate characteristic parameters. Together with the time domain method, the PCA method was used to distinguish emboli, noises and blood flow signals. Comparison experiments with 300 simulated Doppler ultrasound signals and 168 clinical cerebrovascular Doppler ultrasound signals showed that the accuracy of the emboli detection was greatly improved with the PCA approach. The minimum error rates of this detection were 2% and 5%, respectively for the simulated and clinical signals, It is concluded that the PCA method achieves a better detection performance and overcomes the limitations of common signal analysis methods in time and/or frequency domain due to the interference noise and the tradeoff between the time and frequency resolution in the spectrogram analysis, Thus the PCA method may be applied to the automated emboli detection system in the clinical environment.
出处
《声学学报》
EI
CSCD
北大核心
2006年第3期228-232,共5页
Acta Acustica
基金
上海市曙光计划资助(2003-01)
关键词
超声多普勒信号
主元分析方法
信号检测
时域检测
特征参数
干扰噪声
时频分辨率
频谱分析
自动检测
脑动脉
Acoustic noise
Biomedical engineering
Blood
Doppler effect
Errors
Frequency domain analysis
Principal component analysis
Signal detection
Time domain analysis
Ultrasonic measurement
