1) When recording pay particular attention to getting good data on the eye channels (EGI channels: 8, 26, 125, 126, 127, 128). The eye channels should be bipolar; that is, the left/right VEOG should consist of channels 26/127, 8/126 and HEOG 128/125. A bipolar eog signal is computed from the difference of the two channels which reduces the amount of EEG signal content since  EEG signals tend to be similar at both electrodes and thus cancel out.  If two channels are not input to doeogcoeffs for VEOG/HEOG a warning message is displayed stating that the EOG channel is not bipolar.  Use a gain of 10K on the EGI system if the experiment is expected to include lots of eye movement. The recording range is increased to +-500 microVolts for the 10K gain so that more eye movement is recorded without clipping. Flagclip should be run with a range of +-450 for data recorded at 10K gain.

2) Preliminary eegscoring for EOG corrections: DO NOT start general eegscoring before EOG correcting.  Unless there are particularly noisy channels it may be best to do EOG corrections right after flagclip (and filt60, if necessary) has been run.

i) Use eegscore with eye channels fixed at bottom.

DO NOT mark bad any eye related signal on any channel

 Moderate noise should average out in computing the correction coefficients but if there are very bad spots on any channel score them out

ii) Make a decision on which eye channels are good enough to use for EOG correction

For EGI: need  125 and 128 for HEOGmust choose between pairs 8/126 or 26/127 for VEOG

want good eye signal with little channel drift or strong noise not related to eye movement;

3) Compute EOG correction coefficients with the Matlab program doeogcoeffs.  EOG correction coefficients can be computed for a single data file or averaged over several data files recorded under similar conditions. There should be a unique configuration file for each .DAT file or group of files for which EOG correction coefficients are computed since the coefficients are linked to the configuration file and will be used later by the doeogcorr program to apply the corrections.
The configuration file looks like:

   VEOG1: 26
   VEOG2: 127
   HEOG1: 128
   HEOG2: 125
   CHANNELS: 1:128

and can be created with a text editor prior to running doeogcoeffs or interactively while it is running.

i) use single file: doeogcoeffs('.',' ','DAL014R1_NO60.DAT'); where DAL014R1_NO60.DAT has been flagclipped and perhaps put through filt60. The second argument, ' ', indicates that a configuration file hasn't yet been created so the program will prompt for eye channels to use and the list of channels upon which to apply the EOG corrections. This list should probably include all data channels (e.g. 1:128) but leave out event channels. Channels marked all bad by eegscore will have a missing data value for the EOG correction coefficient and thus will not be corrected. Give the configuration file a descriptive name like 'DAL014R1cfg' since it   will later be used as an input argument for program doeogcorr.

ii) use multiple files from same session: doeogcoeffs('.',' ','datFileList'); where datFileList  is an ascii file with one .DAT file name per line.  Program will prompt as in (i) for the necessary configuration information. For a list of three files in datFileList name the configuration file (at the program prompt) something like 'DAL014R1-3cfg'.

iii) use all .DAT files from same session: doeogcoeffs('.');

iv) doeogcoeffs saves the EOG correction coefficients in a text file named using the configuration  file name with _coeffs added (e.g. DAL014R1-3cfg_coeffs). This file can be opened with a text editor to see if any of the coefficients look particularly bad.  They should have a magnitude generally less than one and be larger for the electrodes in the front of the head. Laterally symmetric electrodes such as F7(34)/F8(122), T3(46)/T4(109), T5(58)/T6(97) should have approximately equal VEOG coefficients and equal magnitude opposite sign HEOG coefficients. The mean, standard deviation and count are listed which is useful if multiple data files have been used to estimate the coefficients.

 May 2001 added plot of eog coefficients for 128 channel EGI data and 30 channel SAI data. Below are examples from an eye movement test:

 LNT20KEOSCRIPT2.DAT (VEOG)

            
            and HEOG
            

 

4) do EOG corrections  with the Matlab program doeogcorr .

i) for single file: doeogcorr('.','DAL014R1cfg','DAL014R1_NO60.DAT'); the EOG corrections are applied to DAL014R1_NO60.DAT and the corrected data are written to a new file DAL014R1_NO60_EOGC.DAT. The mask file DAL014R1.MASK is left unchanged.

ii) for multiple files from the same session: doeogcorr('.','DAL010R1-3cfg','datFileList');  where datFileList contains three lines with the appropriate file names. Three new corrected data files are created as described in (i).  All mask files are left unchanged.

iii) to mark bad times/channels when the EOG correction is above a certain threshold use doeogcorr('.','DAL010R1-cfg','datFileList',corrThreshold,timeWidth). If the magnitude of the EOG correction exceeds corrThreshold (microVolts) the data is marked bad in the Mask file. In addition all data points within timeWidth (seconds) are also marked bad so that the entire eye artifact can be masked out - not just the part of it that exceeds corrThreshold. Some experimentation with corrThreshold and timeWidth may be necessary to obtain optimum results.  A new mask file DAL014R1_EOGC.MASK is created. The original .MASK and .DAT files are left unchanged. 

Doeogcoeffs/doeogcorr perform the EOG correction described in Gasser et al. EUR. Arch. Psychiatry  Clin. Neurosci. (1992) 241: 352-356.   A 7.5 Hz low-pass filter is used on the EOG channels as described in the paper.  Coefficients are averaged over all trials contained in the DAT files specified in datFileList.