mne-bids-pipeline ERF/ERP sensor-space analysis


I am using the mne-bids-pipeline for MEG/EEG processing and really like many aspects including the .html outputs. However, I would like to be able to do a few things with the condition contrasts that I think are quite standard in MEG/EEG sensor-space analysis and that I did previously in mne python. I thought these steps might come about through options specified in the evoked contrasts and in the Group Analysis but I don’t think so or maybe I didn’t understand something? I would like to do the following:

(1) have ERP/ERF plots that show the multiple conditions together on a single plot with the difference contrast as a topographic map (e.g. standard N400).
(2) perform a standard sensor-level analysis (e.g. cluster-based permutation test) for the ERP/ERF amplitudes for a given contrast over a specified time window for a group analysis (e.g. unrelated vs. related words)

From my .html output (and looking at examples, including ERP CORE - MNE-BIDS-Pipeline which has very standard ERP effects), the pipeline is currently producing ERP/ERF plots for the conditions separately and performs some stats on the contrast but for decoding only.

Are there any plans to implement (1) and (2) above or did I miss something? Thank you.

Best wishes,

  • MNE version: 1.2.1
  • operating system: Centos 7


I believe there’s a bug in the pipeline and some of the outputs regarding contrasts are actually missing from the reports. For example for the grand mean N400 report for ERP CORE:

There should also be sections for the “related” condition and for the contrast between “related” and “unrelated”, but they’re currently missing

@larsoner I assume this bug was introduced during the recent refactorings…

Regarding cluster permutation tests on the sensor level, like you said this is currently not implemented. Are you not happy with the time-by-time decoding we added (which I personally believe is much more elegant)? Is there something missing?

Thank you

Hello @richard and thanks for your reply.

Good to know that you think there is a bug in output of the contrasts. Would be great if this is fixed.

Please can you explain why you believe the time-by-decoding is more elegant? I’m certainly appreciating that feature but it’s a new way of analysis for me and I would also like to do the more traditional sensor level analysis on amplitudes. Personally, it’s important that I can compare my new analysis with the mne-bids-pipeline with some previous results. It would also be helpful to see how results for the time-by-time decoding relate to old-fashioned amplitude analysis. Best, Lucy