MEG and EEG Data Analysis Using MNE Software

Basics

MEG/EEG data analysis in MNE software uses information from structural MRI images, which have to be pre-processed using Freesurfer. You may want to start with the tutorial based on an example data set, as described in the MNE manual (Version 2.6, Version 2.7.1; Version 2.7.3; chapter 12), or look at some example scripts. Freesurfer is accompanied by extensive Freesurfer Wiki pages, containing a Getting Started and FAQ section. You will need some experience with Linux commands and scripting, which you may find on our beginners' pages.

If you've never used shell scripts before, this primer on shell scripting will get you on the way.

There is also a short description on how to prepare for MNE analysis and access the Matlab toolbox.

Look here for MNE Python tools, e.g. for time-frequency analysis and sensor-space statistics.

The parameters in the following examples are reasonable choices for standard analyses. However, these Wiki pages are not supposed to substitute the MNE manual (V2.6, V 2.7), and reading papers.

Step-by-step Guide

Note that some of these steps can be done in parallel, for example MRI preprocessing and MEG averaging.

1) Pre-process your MRI Data Using Freesurfer

2) Fix EEG electrode positions in Fiff-files

3) Create Source Space and Head Surfaces (incl. aligning coordinate systems)

4) Compute the Forward Solution and BEM

5) Compute the Noise Covariance Matrix

6) Compute the Inverse Operator

7) Averaging MEG data (incl. correcting EEG location information, Marking bad channels)

8) Compute the Source Estimates (incl. average cortical surface, grand-averaging)

9) ROI/Label analysis (incl. pre-defined labels, make-your-own)

All-in-One

List of Most Relevant MNE Commands

Related Issues

1) You may want to filter or maxfilter (Matlab script) your data before averaging

2) At the moment, MNE does not provide any statistics tools (but see MNE-Python tools, point 11). You can use sensor stats implemented in SPM (SensorSPM) for statistics in sensor space.

3) For SensorSPM (sensor stats), you should interpolate your MEG data on a standard sensory array.

4) For data exploration or visualisation, you may want to compute grand average data in signal space.

5) Applying the inverse operator to single-trial data requires some extra processing steps.

6) Simulate your own data in MNE, e.g. to check localisation accuracy for specific ROIs

7) Compute Sensitivity Maps for EEG and MEG configurations

8) Baseline Correction for source estimates

9) Converting vertex locations from MNE STC-files to MNI coordinates

10)The MNE Sample Data Set (CBU only)

11) MNE Python tools and example scripts (e.g. averaging, time-frequency analysis, non-parametric statistics)

Dan's Pages (from Martinos Center for Biomedical Imaging)

MEG Data Processing

Structural Analysis