[FieldTrip] Precisions on DICS beamforming on EEG data and normalisation

Marc Lalancette marc.lalancette at sickkids.ca
Tue Mar 14 17:46:50 CET 2017

Hi Michel-Pierre,

Regarding question 2, I'm mostly familiar with LCMV, and I can't remember exactly how DICS works, but I would guess normalization approaches have the same properties in both.  (Please someone correct me on this if I'm wrong.)  One great reference for LCMV beamformer in general, and normalization in particular, is the book by Sekihara and Nagarajan.  For a scalar beamformer, yes normalizing the leadfield ("array-gain") will correct depth bias, but I find these absolute values harder to interpret.  Dividing instead by projected noise ("unit-noise-gain") also corrects depth bias, and has better spatial resolution.  For a vector beamformer, things get a bit more complicated as the "array-gain" and "unit-noise-gain" vector formulae in that book are not rotationally invariant and I would not recommend using them.  (See my recent post: https://mailman.science.ru.nl/pipermail/fieldtrip/2017-March/011390.html)  Fieldtrip does not by default use these normalizations, but I also haven't seen an analysis of (or had time to investigate much) how its vector beamformer normalization approach fares in terms of bias and resolution compared to others.  Maybe it exists somewhere?  Sorry if it's not a very practical answer...


Marc Lalancette
Lab Research Project Manager, Research MEG
The Hospital for Sick Children, Diagnostic Imaging, Room S742
555 University Avenue, Toronto, ON, M5G 1X8
416-813-7654 x201535

-----Original Message-----
Date: Mon, 13 Mar 2017 12:47:25 +0000
From: MP Coll <mpcoll at mac.com>
To: fieldtrip at science.ru.nl
Subject: [FieldTrip] Precisions on DICS beamforming on EEG data and
        using   normalisation
Message-ID: <28ea0ca0-fed2-3cba-aa11-a79cc8c7c1a3 at mac.com>
Content-Type: text/plain; charset=utf-8; format=flowed

Dear Fieldtrip Community,

My name is Michel-Pierre Coll and I am a postdoctoral researcher at King's college London.

A reviewer recently suggested we perform DICS beamforming to source localise EEG effects in the frequency domain during an action observation/execution paradigm. I was able to perform these analyses using the very good tutorial on the Fiedltrip website. However, I have some questions regarding these analyses. I have searched the literature and the mailing list but somehow I can't find clear answers to these basic questions.

1) Is it appropriate to perform DICS beamforming using EEG with 60 channels (standard montage) ? If not, what would be the appropriate number of channels ? Can you suggest a reference discussing this issue ?

2) When not using a contrast to perform the beamforming, is the normalisation of the lead field an adequate procedure to correct for the depth/centre bias ? The Fieldtrip tutorial suggest it is but other posts on the mailing list suggest that it is not.

3) How does one choose an optimal time window for DICS beamforming when the duration of the effect is quite long (e.g. several seconds of alpha changes in response to a visual stimulus) ? Is it correct to use a longer time-window (e.g. 2 seconds) that is representative of the duration of the effect ?

I would greatly appreciate any hints on these questions or if you could point me towards relevant texts discussing these issues.

MP Coll


fieldtrip mailing list
fieldtrip at donders.ru.nl

End of fieldtrip Digest, Vol 76, Issue 14


This e-mail may contain confidential, personal and/or health information(information which may be subject to legal restrictions on use, retention and/or disclosure) for the sole use of the intended recipient. Any review or distribution by anyone other than the person for whom it was originally intended is strictly prohibited. If you have received this e-mail in error, please contact the sender and delete all copies.

More information about the fieldtrip mailing list