[FieldTrip] Fwd: Beamformer confusion
matt.craddock at uni-leipzig.de
Fri Jun 27 12:51:11 CEST 2014
From the plot, it looks like the volume conductor is singlesphere but
the sourcemodel is derived from the standard boundary element model, and
the electrodes are then not aligned to either of them. First up, I'd
suggest using the standard BEM as your volume conductor and see if that
helps. The electrode positions may be the next thing to fix - they might
re-align automatically properly, in which case you won't need to do
anything. If they have standard 1020/1005 co-ordinates, they should be
aligned fine already, but if not you could try replacing the positions
in your dataset with the standard positions (i.e. if you have an Fp1,
take the Fp1 co-ordinates from the templates in the
On 27/06/2014 06:41, Eelke Spaak wrote:
> PS: Forwarding your figure to the list.
> ---------- Forwarded message ----------
> From: Eelke Spaak <eelke.spaak at donders.ru.nl>
> Date: 27 June 2014 07:40
> Subject: Re: [FieldTrip] Beamformer confusion
> To: Email discussion list for the FieldTrip project <fieldtrip at donders.ru.nl>
> Hi Tyler,
> First, please send your messages to the list, so that others may
> benefit (now or in the future through the archives).
> Then, regarding your plot, indeed I think there might be something
> wrong (though I am not sure). You are looking for a proper alignment
> between the sensors (I have colored them red to be able to better
> see), the source model (grid points for beamforming) and the volume
> conductor model (translucent mesh; sphere in your case). I have never
> used a single sphere model myself, so I cannot really comment on
> whether the sphere aligns properly. However, I would suspect not,
> since there are grid points located outside the sphere. Also, the
> sensors might not be properly located, some are intersecting the body
> of grid points.
> But I do seem to recall that sensors are automatically projected to
> the volume conductor surface in the case of EEG. Maybe someone who has
> EEG source modelling experience can comment on this (and your figure)?
> I usually only work with MEG, so I don't have the same intuition for
> your kind of plot as with an analogous plot for MEG.
> The following tutorial might be very helpful for you, by the way:
> On 27 June 2014 03:06, Tyler Grummett <tyler.grummett at flinders.edu.au> wrote:
>> Hello Eelke,
>> Thank you for replying to my email.
>> I have attached an image with the requested sourcemodel positions and sensor positions.
>> I dont really know how to interpret the figures to be honest.
>> I have attached a .fig figure so that you can have a play with it.
>> Tyler Grummett ( BBSc, BSc(Hons I))
>> PhD Candidate
>> Brain Signals Laboratory
>> Flinders University
>> Rm 5A301
>> Ext 66124
>> From: fieldtrip-bounces at science.ru.nl <fieldtrip-bounces at science.ru.nl> on behalf of Eelke Spaak <eelke.spaak at donders.ru.nl>
>> Sent: Thursday, 26 June 2014 11:17 PM
>> To: FieldTrip discussion list
>> Subject: Re: [FieldTrip] Beamformer confusion
>> Hi Tyler,
>> You write that you end up with the same result "whatever data you
>> process", so I suspect the raw EEG/MEG data is not to blame.
>> Rather, I suspect something is wrong with your leadfield or one of its
>> ingredients: the volume conductor model, source model, or
>> electrode/gradiometer definitions. It's probably a good idea to check
>> whether these three ingredients all line up. Could you plot them in
>> one set of axes, as described at the end of this section:
>> On 26 June 2014 10:00, Tyler Grummett <tyler.grummett at flinders.edu.au> wrote:
>>> Hello Fieldtrip,
>>> I have been using the LCMV beamformer in fieldtrip for some time now, so I
>>> have experience with it.
>>> I am facing an issue where whatever data I process I get an image identical
>>> to attachment 'sourceplot.png'. We have tried visual SSR, memory tasks,
>>> tactile tasks etc.
>>> Also, whatever I do with the data, the same issue occurs (filtering etc).
>>> Attachment 'original_data.png' shows the data before going into the
>>> beamformer, and 'virtual_data.png' shows the data after it has run through
>>> We have tried filtering in every possible combination ie lowpass = 40/30/20,
>>> highpass = 1/10/20, we even did a highpass of 100 and it
>>> was still identical.
>>> We dont think this is a fault of the beamformer, but we cant work out how to
>>> get rid of the issue. It is overpowering the other data.
>>> We have also had a look at some spectra of the brain region with the high
>>> power and a brain region that doesnt have the high power and it appears
>>> as though there is higher power over all frequencies.
>>> Before asking:
>>> -There isnt any muscle, the data was recorded from a paralysed person.
>>> -We have tried it on CAR'd data and data that hasnt been CAR'd
>>> We are all out of ideas.
>>> Kind regards,
Dr. Matt Craddock
More information about the fieldtrip