From raw MEG/EEG to publication: how to perform MEG/EEG group analysis with free academic software

Free academic toolboxes have gained increasing prominence in MEG/EEG analysis as a means to disseminate cutting edge methods, share best practices between different research groups and pool resources for developing essential tools for the MEG/EEG community. In the recent years large and vibrant research communities have emerged around several of these toolboxes. Teaching events are regularly held around the world where the basics of each toolbox are explained by its respective developers and experienced power users. However, most teaching examples only show analysis of a single ‘typical best’ subject whereas most real MEG/EEG studies involve analysis of group data. It is then left to the researchers in the field to figure out for themselves how to make the transition and obtain significant group results. The special research topic aims to address this gap by publishing detailed descriptions of complete group analyses of datasets available online. The level of detail of the description should be such that the reviewers and the readers will be able to fully reproduce the analysis and results and port the analysis to their own data.

We invite research groups who have original analysis pipelines either based on their own toolboxes or on using free academic software in a non-trivial way to contribute to the research topic.

The developers of six major free academic toolboxes for MEG/EEG analysis: Brainstorm, EEGLAB, FieldTrip, MNE, NUTMEG, and SPM have committed to contributing to the research topic. Their submissions will all be based on the same group MEG/EEG dataset. This dataset containing evoked responses to face stimuli was acquired by Richard Henson and Daniel Wakeman. The raw data are available at https://openfmri.org/dataset/ds000117/

Detailed instructions for each toolbox will be made available online including analysis scripts and figures of results.

The submissions should comply with the following requirements:

1. The analyses should be based on a group of subjects with a recommended size of at least 10 (using data sets with fewer subjects should be well justified).


2. The quality of the analysis and the results should be sufficient for publication as a peer-reviewed research paper.


3. The submissions will be evaluated by at least two reviewers one of whom will be a data analysis expert and the other a novice user with minimal data analysis expertise. Authors should ensure that their descriptions are accessible to novice users at the level of Ph.D. students with basic background in MEG/EEG and signal processing.


4. Papers describing previously published analyses are acceptable and encouraged as long as they comply with all the requirements herein.


5. The data analyzed should be downloadable from an online repository guaranteed to persist at least until the end of 2027. Repositories recommended by Frontiers and conforming to this requirement are zenodo.org and figshare.com. The authors’ personal or university sites will not be acceptable. The conditions of access to the data should not restrict replication of the described analysis.


6. Analysis scripts allowing for replication of all described analyses should be submitted as supplementary material.


7. The analysis should use free academic software and only rely on commercial software widely available in academic settings (e.g. MATLAB, Windows).


8. Any dependencies on software not included in the submission should be clearly documented and the versions of this software should be clearly identified. These exact versions should be downloadable from online repositories guaranteed to persist until the end of 2027.


9. If the analysis depends on any manual steps (e.g. artifact rejection) the criteria for manual decisions should be described in detail and the outcomes of the authors’ own manual processing (e.g. indices of rejected trials) should be submitted in the supplementary material to allow for subsequent fully automatic replication of the analysis.


10. If the paper contains novel results not previously endorsed by peer review, these should not be emphasized and in particular not reported in the title and the abstract. The reason for this requirement is that the papers will be evaluated on technical correctness of the analysis and clarity of the description rather than scientific content and the reviewers will generally not be qualified to judge the validity of the results.


11. The analyses should be reproducible in reasonable time (at most a few hours) on reasonable hardware (typical data analysis PC) to allow for replication by the reviewers.


12. To avoid repeated descriptions of the same processing steps in several papers authors are allowed to use the output of an analysis described in a different research topic submission as input to their analysis. Links to draft versions of analyses performed with major free academic toolboxes will be available at http://neuroimage.usc.edu/brainstorm/Biomag2016. Authors using this option should clearly identify what files from the other analysis they are using and make sure prior to submitting the final version of their paper that their analysis works with the final version of its dependencies.


13. Authors should clearly identify in the cover letter what operating system and software platform their analysis uses to facilitate reviewer selection.


14. To ease porting of the described analyses to the readers’ own data we recommend that raw data sets included in the submissions are organized according to the draft version of the MEG extension to the Brain Imaging Data Structure (BIDS) standard. A link to the current version is available at http://bids.neuroimaging.io/.

We hope that this Research Topic will lead to creation of invaluable resource for the whole MEG/EEG community and contribute to establishment of good practice and promoting consistent and reproducible analysis approaches.