Docstoc

Preprocessing for EEG _ MEG - Wellcome Trust Centre for

Document Sample
Preprocessing for EEG _ MEG - Wellcome Trust Centre for Powered By Docstoc
					Preprocessing for EEG & MEG

   Tom Schofield & Ed Roberts
Data acquisition
Data acquisition
Using Cogent to a generate marker pulse..

       drawpict(2);
      outportb(888,2);
      tport=time;
      waituntil(tport+100);
      outportb(888,0);
      logstring( [„displayed „O‟ at time ' num2str(time) ]);
Two crucial steps
   Activity caused by your stimulus (ERP) is
    „hidden‟ within continuous EEG stream
   ERP is your „signal‟, all else in EEG is „noise‟
   Event-related activity should not be random,
    we assume all else is
   Epoching – cutting the data into chunks
    referenced to stimulus presentation
   Averaging – calculating the mean value for
    each time-point across all epochs
Extracting ERP from EEG
                     ERPs
                     emerge
                     from EEG
                     as you
                     average
                     trials
                     together
Overview

   Preprocessing steps
   Preprocessing with SPM
   What to be careful about
   What you need to know about filtering
mydata.mat
Epoching
Epoching - SPM




        Creates:   e_mydata.mat
Downsampling
              Nyquist Theory –
               minimum digital
               sampling frequency
               must be > twice the
               maximum frequency
               in analogue signal
              Select „Downsample‟
               from the „Other‟ menu
Downsample




     Creates:   de_mydata.mat
Artefact rejection
                Blinks
                Eye-movements
                Muscle activity
                EKG
                Skin potentials
                Alpha waves
Artefact rejection
                Blinks
                Eye-movements
                Muscle activity
                EKG
                Skin potentials
                Alpha waves
Artefact rejection - SPM




     Creates:   ade_mydata.mat
Artefact correction
   Rejecting „artefact‟ epochs costs you data
   Using a simple artefact detection method will
    lead to a high level of false-positive artefact
    detection
   Rejecting only trials in which artefact occurs
    might bias your data
   High levels of artefact associated with some
    populations
   Alternative methods of „Artefact Correction‟
    exist
Artefact correction - SPM
                               SPM uses a
                Weighting
                 Value          robust average
 Outliers are
  given less
                                procedure to
    weight                      weight each
                                value according
                                to how far away
                                it is from the
Points close                    median value for
 to median
weighted „1‟                    that timepoint
Artefact correction - SPM
                        Normal
                         average
                        Robust
                         Weighted
                         Average
Robust averaging - SPM




     Creates:   ade_mydata.mat
Artefact Correction
   ICA
   Linear trend detection
   Electro-oculogram
   „No-stim‟ trials to correct for overlapping
    waveforms
Artefact avoidance
   Blinking
   Avoid contact lenses
   Build „blink breaks‟ into your paradigm
   If subject is blinking too much – tell them
   EMG
   Ask subjects to relax, shift position, open mouth slightly
   Alpha waves
   Ask subject to get a decent night‟s sleep beforehand
   Have more runs of shorter length – talk to subject in between
   Jitter ISI – alpha waves can become entrained to stimulus
      Averaging
R = Noise on single trial
N = Number of trials

Noise in avg of N trials
(1/√N) x R

More trials = less noise
Double S/N need 4 trials
Quadruple need 16 trials
     Averaging




Creates: made_mydata.mat
Averaging
   Assumes that only the EEG noise varies
    from trial to trial
   But – amplitude will vary
   But – latency will vary
   Variable latency is usually a bigger
    problem than variable amplitude
Averaging: effects of variance




           Latency variation can be a
             significant problem
Latency variation solutions
   Don‟t use a peak amplitude measure
Time Locked Spectral Averaging
Other stuff you can do – all
under „Other‟ in GUI
   Merge data sessions together
   Calculate a „grand mean‟ across
    subjects
   Rereference to a different electrode
   FILTER
Filtering

   Why would you want to filter?
Potential Artefacts
   Before Averaging…
       Remove non-neural voltages
       Sweating, fidgeting
       Patients, Children
       Avoid saturating the amplifier
       Filter at 0.01Hz
Potential Artefacts
   After Averaging…

       Filter Specific frequency bands
       Remove persistent artefacts
       Smooth data
Types of Filter

1.   Low-pass – attenuate high frequencies

2.   High-pass – attenuate low frequencies

3.   Band-pass – attenuate both

4.   Notch – attenuate a narrow band
    Properties of Filters
      “Transfer function”
    1.    Effect on amplitude at each frequency
    2.    Effect on phase at each frequency

      “Half Amp. Cutoff”
    1.   Frequency at which amp is reduced by 50%
High-pass
Low-pass
Band-pass and Notch
Problems with Filters
                  Original waveform, band pass of
                   .01 – 80Hz
                  Low-pass filtered, half-amp
                   cutofff = ~40Hz
                  Low-pass filtered, half-amp
                   cutofff = ~20Hz
                  Low-pass filtered, half-amp
                   cutofff = ~10Hz
Filtering Artefacts
   “Precision in the time domain is inversely related to precision in
    the frequency domain.”
    Filtering in the Frequency Domain
                  B                     C
A




        D             E
Filtering in the Time Domain


                  Filtering in the time domain is
       X-1         analogous to smoothing
        X

         X+1      At a given point an average is
                   calculated in relation to two
                   nearest neighbours or more
Filtering in the Time Domain


                   Waveform progressively
                    filtered by averaging the
                    surrounding time points.

                   Here x = ((x-1)+x+(x+1))/3
Recipe for Preprocessing

    1. Band-pass filter e.g.0.1 – 40Hz
    2. Epoch
    3. Check/View
    4. Merge
    5. Downsample?
    6. Artefacts; Correction/Rejection
    7. Filter
    8. Average
Recommendations
1.   Prevention is better than the cure

2.   During amplification and digitization minimize
     filtering

3.   Keep offline filtering minimal, use a low-pass

4.   Avoid high-pass filtering
     Summary

1.   No substitute for good data
2.   The recipe is only a guideline
3.   Calibrate
4.   Filter sparingly
5.   Be prepared to get your hands dirty
References
   An Introduction to the Event-related
    Potential Technique, S. J. Luck
   SPM Manual

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:86
posted:4/8/2011
language:English
pages:44