[FieldTrip] clarification on beam forming evoked fields in MEG

[Robert Oostenveld]

Hi Russ,

The units that are used by FieldTrip in data processing are not always consistent. The problem already starts at the level of reading (where we sometimes use our own FT code, and sometimes code from other projects). MEG data might be represented in T or fT, but for Eleka there are also gradiometers that have their data expressed as T/m or fT/cm or T/mm or … Then the anatomical MRi is read in, coregistered and used to construct a volume conduction model. In that volume conduction model there is again geometrical units (m, cm, mm) but also units of conductivity.

You can imagine that combining data (in algorithms) that has different units will lead to unexpected results. Hence I usually play it safe and report on the results of computations as “arbitrary units”. Mostly in statistics and interpretation it does not matter anyway.

If you use SI units throughout, then the low-level FieldTrip code will work as expected and also represent the output of all computations in SI units. For anatomical models you can use ft_convert_units to scale it to meters. For electrode and gradiometer arrays you should use the distance and amplitude option in ft_datatype_sens (see the code, it is not mentioned in the help) and for Elekta the scaling should be specified as ‘amplitude/distance’.

The unit handling is a known issue and we have a number of ideas on http://bugzilla.fieldtriptoolbox.org to improve it, but to little resources to actually get it resolved.

best regards,
Robert


PS Note that for spectral analysis it is again different, since spectral power or density can be differently normalized. Also in that aspect the FT code is not 100% consistent. But since you mention ERFs that does not apply.



On 31 Dec 2015, at 22:51, russ port <russgport at gmail.com> wrote:

> Dear Robert and Jan-Mathijs
> 
> I was hoping for someone to help me with a point of clarification concerning the fieldtrip wiki. I have followed Fieldtrip’s example of creating virtual sensor ERFs for MEG data using LMCV beamforming (e.g. http://www.fieldtriptoolbox.org/tutorial/aarhus/beamformingerf). I get great ERFs for auditory responses that I’m happy with. There is a point on contention left between my lab mates and I though. Following what was discuss previously (http://mailman.science.ru.nl/pipermail/fieldtrip/2015-September/009609.html), I think that the ERF are in a “unit-less” unit. They suggest however, that these (as mentioned in the wiki) are related to equivalent current dipoles for the chosen sources (and so the unit would be A/m [or since we used cm in the head model, A/cm]). Please can you help clear up this point of contention?
> 
> Thanks
> Russ

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20160104/e9ad390d/attachment.html>