Pre-Processing of ERP Data. Peter J. Molfese, Ph.D. Yale University

Transcription

1 Pre-Processing of ERP Data Peter J. Molfese, Ph.D. Yale University

2 Before Statistical Analyses, Pre-Process the ERP data

3 Planning Analyses

4 Waveform Tools

5

6

7

8 Types of Tools Filter Segmentation Visual Review Artifact Detection, Bad Channel Correction, Baseline Correction, Average Reference (Montage Operation), Averaging Move on to Statistical Analyses

9 Filtering

10 Filtering Filtering reduces total frequencies by specific amounts: 60 Hz (50 Hz in Europe) Typical: Hz 200 Series Amps: Select Low Pass (30 Hz)

11 Filtering 300 Series Amps (portables): First filtered using First Order High Pass Filter = 0.1 Then settings used in last slide. Do NOT mix data from 200 & 300 series amplifiers in any one study. If conducting a series of related studies, may want to use the same amplifiers.

12 More Filtering Filter to the level that you will analyze Filters are additive Can process a file many different ways e.g. Once at 0.1 to 30 hz and once at theta Filter all files in a project to the same level Data change every time they are filtered!

13 Segmentation Original data recorded is continuous EEG with markers included indicating stimulus events and subject response events (if programmed: e-prime). Segmentation breaks ongoing EEG into epochs Create categories for each stimulus Segment on stimulus, response, blink, startle, etc. Look for easy way to categorize trials into correct and incorrect responses

14 Segmentation - Part 1

15 Segmentation - Part 2

16 Segmentation - Part 3

17 Segmentation - Part 3

18 Segmentation - Part 4

19 Segmentation - Part 5

20 Segmentation - Part 5

21 Visual Inspection Check Eye channels: Could be bad because not contacting skin OR detecting eye blinks or eye movements. If bad across high number of trials, bad electrode and must be marked BAD for entire recording session would fail because of eye artifacts.

22 Visual Inspection Check Eye channels: Could be bad because not contacting skin OR detecting eye blinks or eye movements. If bad across high number of trials, bad electrode and must be marked BAD for entire recording session would fail because of eye artifacts.

23 Automating Analyses: Constructing a Script Script step 1 Filtering, then Segmentation You must filter before segmentation as filters flatten data at the beginning and end of each epoch

24 Automating Analyses: Constructing a Script Script step 1 Filtering, then Segmentation You must filter before segmentation as filters flatten data at the beginning and end of each epoch

25 Order of Operations Still have left: Artifact Detection Bad Channel Correction Baseline Correction Average Reference Averaging Does order make difference? controversy

26 Artifact Detection Examines segments for bad channels, eye blinks, and eye movements Can define artifacts by threshold Max-Min = 200µV is bad channel over segment Can define artifacts by slope of line Terms: fast-average, fast-amplitude Can define artifacts by modeling examples BESA

27 Artifact Detection Bad Channel Scan - shorted electrodes - bad scalp connection µv over segment Eye Blinks & Movement - between electrodes pair - differ by net type

28 Artifact Detection Eye Blinks & Movement - between electrodes pair - differ by net type

29 Eye Movements: 125 & 128 Eye Blinks: 8-126,

30 Things to Look Out For Problem: The algorithms can be too simple If eye channels detect an eye movement or blink, may mark the segment bad If an eye channel is bad, may mark all segments bad

31

32 Strategies First: Inspect Eye Channels Spherical Interpolation Check impedance levels from records Have Artifact Detection check for you. Second: Interpolate those bad channels Third: Continue processing files Baseline, Average Reference, Average

33 Bad Channel Correction

34 Artifact Regression Gratton, Coles, & Donchin (1983)

35 Automated ICA Joyce, C.A., Gorodntsky, I.F., Kutas, M. (2004) Automatic removal of eye movement and blink artifacts from EEG data using blind component separation

36 Bad Channel Replacement No User Settings - Interpolates Missing Channels Using Weighted Average

37 Average Important Strategies to keep in mind: Weighted Averaging Random Selection of Trials Early vs. Late in Experiment Equal Number of Averages / Condition Equal Number of Averages / Subject

38 Average Reference Re-referencing data Average Reference Defining a new zero point

39 Baseline Correction De-trend data from long slow oscillations in waveform

40 Baseline Correction Pre-stimulus period, usually ms Average value for pre-stimulus points in baseline is calculated Average value subtracted from poststimulus period Important to avoid artifacts in baseline period!

41 Average Reference

42 Average Reference Data originally recorded represent difference between an electrode & reference (Cz) To reduce bias of particular reference, data re-referenced to the average reference

43 Averaging Average together all trials of particular stimulus (or condition) to form single waveform (average) composed of ALL presentations of that stimulus (or condition). Averaging = noise reduction technique Study: Randomly presented 100 /ba/ syllables intermixed with 100 presentations of /ga/. Separate averages constructed for /ba/ and /ga/, each based on their separate ~100 presentations.

44 1. Average 2. Average Reference 3. Baseline Correction

45 Controversy of Order Early vs. Late Baseline Correction baseline changes the data Early vs. Late Averaging OAR must be followed by baseline fix* Early vs. Late Re-Referencing of data Average Reference changes the data Better to average before/after baseline & referencing?

46 Analysis Order

47 1. Average 2. Average Reference 3. Baseline Correction 1. Ave Ref then Average 2. Baseline Correct 1. Baseline then Ave 2. Ave Ref

48 Benefits of Automation Increases Replicability Fairly easy to teach analysis steps High inter-rater reliability Batch processing of files

49 Statistical Extractions Peak Amplitude & Latency Define a window Have tool extract information from waveform Problems: Variability in waveforms trial-to-trial Variability from electrode to electrode

50 Export From NS Two Steps Convert to Text Files Use Matlab to Combine/Rotate Data

51 Change NS files from AVG to RAW: (1) Use NS File Exporter (in Extras folder of NS) Drag and drop AVE files Caution - name change (older version bug) (2) If names altered, use Convert to Text in Net Station or Net Station File Exporter 2.1 (3) Gives floating point simple binary conversion. RM_4452f_1_6x25.ave --> RM_4452f_1_6x25.RAW

52 (2) Text Export 2.1 program Converts RAW to Text Drag in RAW files, click CONVERT to generate TEXT files.

53 3. Can ALSO covert to text using FIILE EXPORT TOOL in Net station Waveform Tools.

54 4. Output = one text file for EACH condition.

55 MCAT Three Functions 1. Rotate data matrix for input to stat programs. 2. Concatenate subject data text files. 3. Combine electrode sites into regional electrodes.

56 MCAT Rotate Data Matrix BEFORE AFTER Channel 1 Data Pt 1 Channel 2 Data Pt 1 Channel 3 Data Pt 1 Channel 1 Data Pt 1 Channel 1 Data Pt 2 Channel 1 Data Pt 3 Channel 1 Data Pt 2 Channel 2 Data Pt 2 Channel 3 Data Pt 2 Channel 2 Data Pt 1 Channel 2 Data Pt 2 Channel 2 Data Pt 3 Channel 1 Data Pt 3 Channel 2 Data Pt 3 Channel 3 Data Pt 3 Channel 3 Data Pt 1 Channel 3 Data Pt 2 Channel 3 Data Pt 3 Suitable for Spatial PCA Suitable for Temporal PCA

57 Combine Electrodes into Regional Electrodes HydroCel Geodesic Sensor Net 256 Channel Map Version Fp1 Fp F F Fz F F REF T3 C3 T4 217 CZ C COM P3 P Pz 190 LM RM T T O1 137 O2 123 Oz S/N For questions or additional assistance please refer to the EGI Sensor Net Technical Manual or contact us at: Electrical Geodesics, Inc Millrace Drive, Suite 307 Eugene, Oregon Phone: (541) Fax:(541) support@egi.com or info@egi.com

58 Combine Electrodes into Regional Electrodes HydroCel Geodesic Sensor Net 256 Channel Map Version Fp1 Fp F F Fz F F REF T3 C3 T4 217 CZ C COM P3 P Pz 190 LM RM T T O1 137 O2 123 Oz S/N For questions or additional assistance please refer to the EGI Sensor Net Technical Manual or contact us at: Electrical Geodesics, Inc Millrace Drive, Suite 307 Eugene, Oregon Phone: (541) Fax:(541) support@egi.com or info@egi.com

59 18 Regions HydroCel LOR ROR LIF RIF LPF RPF LIT RIT LT RT LP RP LTP RTP LO RO LIO RIO

60 Combine Electrodes into Regional Electrodes

61 MCAT For Temporal PCA Analyses

62 MCAT Drag & Drop text files For Temporal PCA Analyses!

63 MCAT!

64 MCAT

65 MCAT to SPSS

66 Benefits: Temporal PCA Plotting Programs DeltaGraph, Excel, BLOT Mathematica Imports

67 QUESTIONS???