ERP (Event Related Potentials) Mode

The ERP module was designed for providing a dynamic, fast and flexible way to visualize and analyze averages computed from an EEG, EMG, EKG, MEG signals or ICA and CSP components.  A virtually unlimited number of averages can be computed and maintained in memory (depending on your computer hardware and operating system) from a single EEG (or MEG) file and it is possible to quickly switch among them. It is also possible to compare two averages, even statistically (t-test on a sample by sample basis). This can be done because every time an average is generated, the entities that are computed are:

• The mean signals relative to each sensor (unipolar data), sample by sample;
• The standard deviation relative to each sensor (unipolar data), sample by sample;
• The number of trials used to compute the mean and SD of each sensor (unipolar data), sample by sample. This information is necessary for the sample by sample t-test when partial artifact removal is activated (described later);
• The mean signals relative to each trace (bipolar data), sample by sample;
• The standard deviation relative to each trace (bipolar data), sample by sample;
• The number of trials used to compute the mean and SD of each trace (bipolar data), sample by sample. This information is necessary for the sample by sample t-test when partial artifact removal is activated (described later);

The ERP module implements two main working modalities: Average mode and Single Trial mode. In the Average mode (which is the most frequently used modality), you will work with fully computed averages, while in the Single Trial mode you can build them deciding whether to accept or reject each trial singularly. In this modality you can also see how your averages are “growing”: every time a single trial is accepted or rejected you can immediately see the changes on the superimposed average.

Another (a third) modality, the Moving Range, was implemented for some Brain-Computer Interface (BCI) protocols, is still experimental and has not been widely tested. It allows to select a fixed number of trials for building the averages to see how the “sub-averages” change when different trial subsets are selected from the whole set. For example one can choose to compute the averages using 15 trials and then (if 100 trials are available) move the average from the first 15 trials to the next 15 and so on to see if the averages change. This was used to help to determine if some “fatigue” effects are observed and to investigate about how many trials are necessary to evoke a “reasonable” response: in some BCI protocols it is not important to have a well established, clear and close to the “state of the art” average, but just to be able to detect, as soon as possible, if two averages are different, thus using a minimum number of trials.

The main features of the ERP module are:

●    Computation and visualization of averages (and their standard deviations) according to many different strategies, including 2-D and 3-D mapping and cartooning mode (screenshots below):

●    Spectral analysis, through a flexible FFT engine, which includes power, amplitude, coherence and Standard Deviation computation, t-test statistics, and much more (screenshots below):

●    Statistical analysis, to compare different averages and visualize, in a highly intuitive way, when and where two ERPs are statistically different (screenshots below):

Created with the Personal Edition of HelpNDoc: Easily create iPhone documentation