[转载]EEGLAB使用9问

  EEGLAB是个免费的EEG分析工具包，由于它功能强大，更新迅速，是广大科研工作者的不错选择。这个是网址：http://sccn.ucsd.edu/eeglab。下面的很多回答来自网上，我用斜体标出。

 

1. 参考电极的问题
一个常听到的问题就是如何将自己的数据变成其他电极的参考，以双耳参考为例，如果左右耳的电极序号为31，32，我们可以通过一句简单的matlab代码实现这一功能：

EEG.data=EEG.data-mean(EEG.data([31 32],:));

当然，任何参考都有自己的弊端，最理想的参考应该类似物理学里面的“无穷远参考”。下面这段英文的问题和回复分析得很有道理，我这里就不赘述。他里面提到一个“无穷远参考”，其实，早在2001年，电子科技大学尧德中教授就提出了一个Reference Electrode Standardization Technique，简称REST

参考文献在这里：

D.Yao. A method to standardize a reference of scalp EEG recordings to a point at infinity. Physiol. Meas., 22(4),693-711,2001.

软件下载地址：http://www.neuro.uestc.edu.cn/rest/index.asp

推荐采用尧老师教研室的REST技术来统一参考电极的问题

 

I have some doubts regarding the referencing of the channels. I have come across many papers which have appreciated average reference while others have really criticized it. Some papers are taking linked earlobes as reference while some are using only ipsilateral ear lobe. Certain papers have referred to Laplacian or Hjorth transformatio regarding the referencing of the datasets.
I'm working on saccade related potentials. I have tried linked earlobes and average referencing for my studies but the data seems to be very much contaminated with the saccade.
So I'll be really glad if some one can suggest me which would be the best way of referencing in my dataset.

Basically, the best “practical” reference would cover the whole body evenly, or at least the whole head evenly, in order to approach the “reference-at-infinity” used in physics. The more channels, the better. The only reason to exclude any channel is if it is known to have noise on it.

这篇论文值得推荐

Jungh?fer M, Elbert T, Tucker DM & Braun C. (1999) The polar average reference effect: a bias in estimating the head surface integral in EEG recording.  Clinical Neurophysiology 110(6): 1149-1155

 

由于EEGLAB核心技术是独立成分分析(ICA)，我先把关于ICA的若干问题罗列如下：

2. 两任务的时候我是该分别ICA还是把两任务合到一起ICA？
ICA should be run using the concatenated data from both conditions. The reason is that otherwise it is going to be harder to compare the ICA components in the two conditions or remove the ICA artifacts in the two conditions (if you run ICA on each dataset separately, someone may always argue that the difference you observe between the two conditions arises because you have selected or removed different ICA components in each condition).

3.运行ICA总是out of memory怎么办？
no artifactual component activities will only be subtracted from the 
channel you selected. There is a runicalowmem() function which we 
recently modified so it allows runing ICA with little memory. It has 
to be run from the command line though. I think it is in the latest 
release of EEGLAB
4. 可用来分析gamma频段(高频)的节律吗？
ICA is a linear decomposition that passes through all the frequencies that are in the data. I believe there is much to be learned still about how ICA parses gamma EEG (and higher) activity
5. 我现在采了一批数据，是148导的，我想计算ICA，请问需要多少个采样点？
ICA learns a (channels,channels) unmixing matrix, so the number of data frames (time points) needed to separate as many components increases as the square of the number of channels. The faq is a bit out of date -- in our work with 72-256 channel data in the lab, we have found that as the channel density becomes high, good ICA solutions typically require a considerable multiple of the channel number squared (up to 30 or more for 256-channel data).
For 148-channel data I would like to collect 30*128^2 ~ 650k (~40 min of data at 256 Hz) ... though it could well be that smaller data sets could also give useful solutions.

 

其他问题

6. 两任务的时频分析是分析了再减还是减了再分析？
Time-frequency analysis is just another way of looking at signals,
especially non-stationary signals. Since, the study of non stationary
signals in time domain is difficult one needs to apply
wavelet/STFT/Time-Frequency analysis to study the time and frequency at
both time. So, any features can be studied only after converting signal to
time-frequency domain alone and one can do this in many ways.

 

7. 如何在EEGLAB中导入数据

    数据的导入很简单，File==>import data==> from ASCII/float ... 就可以了，但注意是[电极数*记录点数]，而不是相反；同时要记得填采样率 Data sampling rate

8. 如何在EEGLAB中导入事件标志

    时间标准复杂一些，但包括3列：type latency duration

比如我的一个txt文本就如同下面：

hand 38266 0
hand 51067 0
hand 63867 0
hand 74108 0
hand 81788 0
hand 92028 0
hand 97148 0
hand 108193 0

然后在column里填type latency duration，以及时间换算里填1/250 (我的采样为250)就可以了！

9.如何导入电极位置信息
I want to know how to construct my own chanlocs. I only know the X,Y and lables of the channels. how to get the theta, radius, sph_theta, sph_phi, sph_radius, sph_theta_besa, sph_phi_besa values.

To do so, from the command line:
pop_chanedit([])
then enter manually your channel labels and x and y positions, then 
press the button "xyz -> polar & sph". You may then save the result in 
a text file. You may also use the menu "Edit > Channel locations" from 
EEGLAB although you will need to have a dataset loaded.