MACCS ERP Laboratory ERP Training

Transcription

1 MACCS ERP Laboratory ERP Training 2008 Session 1 Set-up and general lab issues 1. General Please keep the lab tidy at all times. Room booking: MACCS has an online booking system (see Testing Labs link top left); at the moment students can view but not add bookings. I propose that once students are testing and the lab is busy, that preferred days are allocated to each person to make bookings easier. This will be worked out when/if it becomes necessary. Lab use is limited by 1) time taken to test AND clean up and 2) the number of caps we have. Access: please see Katie Webb (Rm 417) with your student card, to have a lab key issued to you. Problems: See Christopher or Gen, especially if urgent. Record issues / problems and consumables that need ordering in the lab folder (also notify Carmen of consumable needs). 2. Set-up Instructions to participants EEG recording, electrode placement Refer to the ERP lab test guide for a detailed explanation of the set-up procedure (last page this can be detached so it can be easily referred to during your first few set-ups). 3. Cap cleaning Please pay particular attention to cap cleaning as it is extremely important that residual gel does not remain in the electrodes over time the saline can encourage corrosion.

2 4. Sessions 2, 3, 4, 5 Scheduling of sessions, who is testing who. me with day / time for room bookings and access. Training certificate 5. Scheduling of session 6 We will arrange a time (via ) once everyone has done their practice sessions.

3 Session 6 Data processing 1. What is an ERP? The ERP is obtained by time-locking the recording of the EEG to the onset of specific events. Usually the continuous EEG trace is marked according to the event of interest (usually the onset of a stimulus), and epoched by taking a prestimulus baseline period and post-stimulus period (with the marker indicating time point zero). To visualize ERPs the experimenter must use signal processing techniques to eliminate non-event activity and increase the signal relative to the noise. Typically this involves averaging the traces of multiple presentations of the same or similar events together, which tends to decrease the influence of random activity (i.e., the background or non-event related EEG) while maintaining the consistent event-related activity. = signal + noise In this figure, only one epoch is taken and so the signal is small while the noise is relatively large, resulting in a messy-looking trace (centre).

4 = signal + noise In this figure, more epochs are taken but the number is still not optimal as the relative amount of noise is still quite high. = signal + noise The above figure demonstrates how noise decreases, and the signal increases, as more repetitions (epochs) are included in the average, and the smoother ERP emerges from the ongoing EEG. ** This is of course just an example in reality at least 30, and preferably 50 or even 100 trials (depending on the stimulus / experiment) are needed in each average to get a good clean and reliable ERP. 2. Data processing The ERP test manual also contains typical processing steps. Carmen and/or your supervisor will work with you to customize (if necessary) the processing required for your particular data. In the meantime, feel free to come in to the lab and practice with the files in the practice folder (C:\ERPtrain-practice\practiceEEG.cnt). 3. Data analysis As above, analysis of each person s individual data will vary and should be conducted in consultation with your supervisor.

5 Data processing 1. Scan through the.cnt file and block ( ) artifact. 2. EOG (blink) correction: Transforms Ocular Artifact Rejection Select LDR+CNT and specify your output filename by clicking on CNT >>. Also make sure the correct channel (VEOG) is specified; Click on Channel to change if necessary. Other settings must be as shown in the diagram. Click OK. The Maxima are reviewed first; accept (Y) any sweep that contains a clear blink, anywhere. Blinks are reviewed second; accept (Y) any sweep that contains a clear blink at the start of the sweep. Reject (N) all others. This is a good blink. Note the sharp rise and gradual fall. At least 20 (and preferably 30 or 40) clean blinks are required to construct a good blink model for the correction. 3. Re-referencing: Transforms GFP/Reference Under Average Reference Channels, select Specified and then select the channels to be included. All is not recommended; even if using a common average reference the EOG channels are typically not used. Add GFP Channel can be unselected. Note that if referencing to one channel, that channel will become a flat line (since the reference channel is subtracted from all other channels, including itself). Check Include Reference Channel to include the mastoid (the original reference). 4. Filtering: Transforms Filter The set-up file currently bandpass filters outside 0.1 and 100 Hz. Filtering off-line to a lowpass of 30 Hz would be recommended for most studies. Therefore select Bandpass, Zero Phase Shift (makes two passes, one forwards and one backwards), and 0.1 (or 1) to 30 Hz. The higher the roll-off settings (db per

6 Octave), the greater the attenuation of frequencies outside this range; changing the settings will demonstrate this by changing the slope on the graph at the top of the dialog box. 12 db/oct is sufficient (greater than 20 db/oct will have trouble with noise spikes). All Channels should be selected. Specify the output filename. Note that at this point, the.cnt file becomes.eeg. 5. Epoching: Transforms Epoch file Mode: Port/Internal Interval: specify period before and after stimulus onset (0). Specify the Output filename It is also possible to select trials based on stimulus and response parameters (and other criteria) at this stage, e.g. all stimulus codes could be selected and the responses discarded. 6. Baseline correction: Transforms Baseline correct Select Pre Stim Interval, leave Channels as All, specify the output filename. Note if there are any channels labeled as Bad or Skipped, these can be omitted from correction if desired. 7. Artifact rejection: Transforms Artifact Reject Time Domain: specify amplitude range. +/- 150 µv is a good range. Check papers in the literature. Default settings (Keep already rejected blocks, Don t Reject on Bad or Skipped Channels) should be left, and Channels specified to NOT include the VEOG channel (which still contains blinks that may exceed the rejection criteria). 8. Smoothing: Transforms Smooth data Smoothing, while not strictly necessary, will even out the waveforms by averaging each data point with the ones immediately beside it, as seen in the figure below. This example used 3 consecutive points, and made 5 passes. Again, All Channels are selected and output filename must be specified.

7 Original Smoothed 9. Average: Transforms Average The default Time Domain is selected; click on Sort Criteria and enter the parameters to be included in the average e.g. select Type and enter the stimulus code corresponding to the trials that are to be averaged together. Note that this will have to be done for each type individually (e.g. congruent, incongruent and neutral trials). The above steps should be implemented to process at least one participant s data manually, however a batch file can be constructed to perform steps 3 onwards automatically. Scanning the.cnt file for noisy patches, and blink correction, should always be done manually, although the blink correction can be performed in batch mode especially if the recording is clean. The current version of Scan Edit contains a history window which records the command for all processes performed, and these can be pasted into the batch editor to very easily construct a batch file that contains all the correct parameters. This batch file can then be updated with each subsequent participant s code. Therefore, at the end of each step, copy the command from the history window into the tcl batch editor. Note that the CreateSort command will have to be used prior to the Average command in the batch file. Grand averages (Transforms Group/Individual Average, while no data files are open) can be constructed once at least a few participants have been tested.