利用小波包的脑电源定位算法仿真研究

An Electroencephalograph Source Localization Algorithm Based on Wavelet Packet Transform with Simulation

提出了一种利用小波包变换的脑电源定位新算法,可获得大脑皮层上的神经电活动分布信息。该算法首先对脑电信号进行小波包分解、子空间分量选择以及信号重构,之后将重构信号与基于真实头部磁共振图像的3层边界源头模型相结合,进行逆问题求解得到源信息,并从信噪比、源深度、脑电信号导联数3个方面对提出算法和原始算法进行了对比分析。实验结果表明,在不同的情况下,提出算法的重建源与设定源平均距离均在3mm以内,标准差在1mm左右,而原始算法的重建源与设定源平均距离最大达9.4mm,标准差达4.5mm;在同样的情况下,提出算法重建源的均值和标准差均小于原始算法,最高可分别减少6.6mm和3.5mm。提出算法的源定位精度明显强于原始算法(P<0.05),且稳定性强,具有一定的临床价值。

A new algorithm of electroencephalograph （EEG） source localization based on wavelet packet transform is proposed to obtain the neural electrical activity distribution of brain. Conducting wavelet packet decomposition for neural electrical signals, subspace component selection and signal reconstruction, an inverse problem with a MRI-based three-layer head model is solved. A comparative investigation includes signal-to-noise ratio （SNR）, dipole source depth and electrode number. The results indicate that in different circumstances, the average distances between the reconstructed sources of the proposed algorithm and the setting sources get less than 3 mm, and the standard deviations about 1 mm, while the average distances and the standard deviations of the existing algorithm achieve up to 9.4 mm and 4. 5 mm, respectively; in the same circumstances, the means and standard deviations of the proposed algorithm get less than that of the algorithm, and can be reduced by up to 6.6 mm and 3.5 mm, respectively. The proposed algorithm is endowed with better localization accuracy and enhanced robustness than the existing algorithm （P〈0.05） to show the clinical significance.