support Laplacian referencing for EEG

[drammock]

Describe the new feature or enhancement

Add support for Laplacian (sometimes called local average) referencing for EEG.

Here is one reference: https://doi.org/10.1016/j.ijpsycho.2015.04.023 though it's part of a special issue on current source density analysis so there are likely other papers worth consulting before implementing.

Describe your proposed implementation

as a new option mne.set_eeg_reference(ref_channels='local') The function would probably also need a new param for the distance threshold.

Describe possible alternatives

could make a dedicated function instead. set_eeg_reference is already overloaded to handle REST referencing (with extra param forward) so I don't see the harm in extending it to Laplacian as well.

Additional comments

Requested on the forum https://mne.discourse.group/t/laplacian-or-local-average-reference-in-mne/3415

cross-ref to #8481 where Laplacian referencing was mentioned for sEEG, although I think the implementation for that would likely be fairly different, given the shaft-based vs scalp-based electrodes.

[paulokanda]

More references:

Yao, D., Qin, Y., Hu, S., Dong, L., Vega, M. L. B., & Sosa, P. A. V. (2019). Which reference should we use for EEG and ERP practice?. Brain topography, 32(4), 530-549.

Abbass, H. A. (2014, November). Calibrating independent component analysis with Laplacian reference for real-time EEG artifact removal. In International Conference on Neural Information Processing (pp. 68-75). Springer, Cham.

Kayser, J., & Tenke, C. E. (2015). Issues and considerations for using the scalp surface Laplacian in EEG/ERP research: A tutorial review. International Journal of Psychophysiology, 97(3), 189-209.

Carvalhaes, C., & de Barros, J. A. (2015). The surface Laplacian technique in EEG: Theory and methods. International Journal of Psychophysiology, 97(3), 174-188.

Trujillo, L. T., Stanfield, C. T., & Vela, R. D. (2017). The effect of electroencephalogram (EEG) reference choice on information-theoretic measures of the complexity and integration of EEG signals. Frontiers in neuroscience, 11, 425.

Bringas Vega, M. L., Nunez, P., Riera, J., Zhang, R., & Valdes-Sosa, P. A. (2019). Through a glass, darkly: The influence of the EEG reference on inference about brain function and disorders. Frontiers in neuroscience, 13, 1341.

Levin, A. R., Varcin, K. J., O’Leary, H. M., Tager-Flusberg, H., & Nelson, C. A. (2017). EEG power at 3 months in infants at high familial risk for autism. Journal of Neurodevelopmental Disorders, 9(1), 1-13.

[mmagnuski]

@paulokanda @drammock Sometimes the terms "Laplacian" and "Current Source Density" are used interchangeably (see for example: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4610715/). Are we sure that in this context CSD (which can be done in mne) is not good enough?

[drammock]

ha! I didn't know that CSD and surface laplacian were the same thing. @paulokanda it looks like you can use mne.preprocessing.compute_current_source_density(raw). I'll close this, but if I'm mistaken and this doesn't meet your needs please respond here and we can re-open the issue.

[DominiqueMakowski]

If mne.preprocessing.compute_current_source_density(raw) is equivalent to what is described in the literature as rereference using a Laplacian (LAP) “reference-free” transformation (e.g., https://www.frontiersin.org/articles/10.3389/fnins.2017.00425/full), then it would be useful for visibility to have this as an explicit option in mne.set_eeg_reference() (and then explain in the docstring that it basically just runs the CSD transfo)

Just to clarify, does that mean that we don't need to first run raw.set_eeg_reference() before running mne.preprocessing.compute_current_source_density(raw)? Or should we still re-reference using for instance an average ref?

[agramfort]

I would say try to see if it has any impact to average reference first. Just by looking at the outputs.

On Fri, Jun 3, 2022 at 5:06 AM Dominique Makowski @.***> wrote:

If mne.preprocessing.compute_current_source_density(raw) is equivalent to what is described in the literature as rereference using a Laplacian (LAP) “reference-free” transformation (e.g., https://www.frontiersin.org/articles/10.3389/fnins.2017.00425/full), then it would be useful for visibility to have this as an explicit option in mne.set_eeg_reference() (and then explain in the docstring that it basically just runs the CSD transfo)

Just to clarify, does that mean that we don't need to first run raw.set_eeg_reference() before running mne.preprocessing.compute_current_source_density(raw)? Or should we still re-reference using for instance an average ref?

— Reply to this email directly, view it on GitHub https://github.com/mne-tools/mne-python/issues/9659#issuecomment-1145540198, or unsubscribe https://github.com/notifications/unsubscribe-auth/AABHKHGAWLPHF4IE7PQ3HJDVNFZEFANCNFSM5B67OXLQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[larsoner]

IIRC the average ref is done within the Laplacian/CSD code internally, so it shouldn't matter whether or not you apply one first. But feel free to check in code or by example

then it would be useful for visibility to have this as an explicit option in mne.set_eeg_reference()

Since it changes the channel type in MNE it does a bit more than this function/method typically does. I'd rather leave it as a separate function, and just mention it somewhere in the Notes and/or See Also. Would you be up for updating the docstring of set_eeg_reference @DominiqueMakowski to reference the CSD function in some way that makes it more discoverable?