Hauk O, Keil A, Elbert T, Müller M M
Medical Research Council, Cognition and Brain Sciences Unit, 15 Chaucer Road, Cambridge CB2 2EF, UK.
J Neurosci Methods. 2002 Jan 30;113(2):111-22. doi: 10.1016/s0165-0270(01)00484-8.
We describe a methodology to apply current source density (CSD) and minimum norm (MN) estimation as pre-processing tools for time-series analysis of single trial EEG data. The performance of these methods is compared for the case of wavelet time-frequency analysis of simulated gamma-band activity. A reasonable comparison of CSD and MN on the single trial level requires regularization such that the corresponding transformed data sets have similar signal-to-noise ratios (SNRs). For region-of-interest approaches, it should be possible to optimize the SNR for single estimates rather than for the whole distributed solution. An effective implementation of the MN method is described. Simulated data sets were created by modulating the strengths of a radial and a tangential test dipole with wavelets in the frequency range of the gamma band, superimposed with simulated spatially uncorrelated noise. The MN and CSD transformed data sets as well as the average reference (AR) representation were subjected to wavelet frequency-domain analysis, and power spectra were mapped for relevant frequency bands. For both CSD and MN, the influence of noise can be sufficiently suppressed by regularization to yield meaningful information, but only MN represents both radial and tangential dipole sources appropriately as single peaks. Therefore, when relating wavelet power spectrum topographies to their neuronal generators, MN should be preferred.
我们描述了一种将电流源密度（CSD）和最小范数（MN）估计作为单试次脑电图（EEG）数据时间序列分析预处理工具的方法。针对模拟伽马波段活动的小波时频分析情况，对这些方法的性能进行了比较。在单试次水平上对CSD和MN进行合理比较需要正则化，以使相应的变换数据集具有相似的信噪比（SNR）。对于感兴趣区域方法，应该能够针对单个估计而非整个分布式解来优化SNR。描述了MN方法的有效实现。通过用伽马波段频率范围内的小波调制径向和切向测试偶极子的强度来创建模拟数据集，并叠加模拟的空间不相关噪声。对MN和CSD变换后的数据集以及平均参考（AR）表示进行小波频域分析，并绘制相关频带的功率谱。对于CSD和MN，通过正则化都可以充分抑制噪声的影响以产生有意义的信息，但只有MN能将径向和切向偶极子源都恰当地表示为单峰。因此，当将小波功率谱地形图与其神经元发生器相关联时，应首选MN。
