[FieldTrip] About source reconstruction outputs

[ivano triggiani]

Let's see...
If you use pwelch per x,y, and z, you are basically calculating the power
spectrum (which is related to the squared magnitude of the Fourier
transform), so I doubt it's correct to square them again.
I know somebody uses the average, which is a possibility, but it can
underestimate the total spectrum.
The correct solution (again, it can look weird) should technically be the
sum of the 3 PSDs, which are already squared, so Ptot(f)=Px(f)+Py(f)+Pz(f)
is the vector sum and should be the correct solution.

Everyone feels free to correct me if I am wrong.

Ciao,

Ivano
---

*A. Ivano Triggiani, Ph.D.*


________________________________________________



On Wed, Dec 17, 2025 at 6:50 AM Fabio Strappazzon <fab.strappaz at gmail.com>
wrote:

> Ciao Ivano,
> thank you for your anwer.
>
> 1) As expected, thanks.
>
> 2) Your solution would be fine but the problem is that for each dipole I
> have a 3 x pnts array (orientation x time points) and I do not know how
> FieldTrip would take care of that. But even trying to using classical
> MATLAB function like pwelch (as I'm curretly doing) I'm not super sure how
> should I proceed: should I calculate a separate PSD for orientation? And
> then what: summing across orientation? Taking the average? Taking the
> squared root of sum?
>
> Thank you for your help
>
> Il giorno mar 16 dic 2025 alle ore 22:44 ivano triggiani <
> ivanotriggiani at gmail.com> ha scritto:
>
>> Caro Fabio,
>>
>> My humble opinion:
>>
>> 1. Since source reconstruction is a well-known ill-posed problem, every
>> method relies on different constraints. For instance, MNE minimizes the
>> overall L2 norm, which often results in smooth solutions, easily distorted
>> if noise is significant.
>>
>> To stabilize the solution, regularization methods use, for example,
>> estimation of the noise covariance and a regularization parameter. This
>> scaling process changes the output, which is no longer strictly measured in
>> physical units (like current density) but is reported in a.u. or normalized
>> quantity. sLORETA and eLORETA operate under a similar concept of weighted
>> or standardized normalization (I am simplifying!).
>>
>> 2.  I think so, but you must calculate the PSD on the estimated source
>> time series. For distributed solutions the data is contained in the
>> source.avg.mom field, not source.avg.pow, if I remember correctly.
>> You need to restructure the source.avg.mom field so that FieldTrip treats
>> each source location as a separate "channel" before running ft_freqanalysis.
>> Something like:
>>
>> cfg = [];
>> cfg.method = 'mtmfft'; % or whatever
>> cfg.output = 'pow';
>> cfg.foi = 0.5:0.5:40; % freqs
>> freq_source = ft_freqanalysis(cfg, source_data);
>>
>> The resulting PSD may look "weird" compared to sensor-level data because
>> the spatial filter (source method) affects the spectral profile of the
>> noise and signal.
>> Ciao,
>>
>> Ivano
>>
>> ---
>>
>> *A. Ivano Triggiani, Ph.D.*
>>
>>
>> ________________________________________________
>>
>>
>>
>> On Tue, Dec 16, 2025 at 11:55 AM Fabio Strappazzon via fieldtrip <
>> fieldtrip at science.ru.nl> wrote:
>>
>>> Dear all,
>>>
>>> I'm currently using wMNE as inverse operator.
>>> I would like to know two things:
>>>
>>> 1) What are the units of measure of source reconstructed activity? I see
>>> that many papers report them as arbitrary units (a.u.) but I would like to
>>> know more why is that the case.
>>>
>>> 2) Would it be possible to still apply the FieldTrip functions for
>>> calculating PSDs (ft_freqanalysis) on outputs from source recontructed data
>>> with distributed apporaches (like MNE) or there are some specific functions
>>> for those kind of data? If so, on which outputs are calculated: on the
>>> source.avg.mom or the source.avg.pow?
>>>
>>> Many thanks,
>>>
>>> --
>>> *Strappazzon Fabio, *
>>> *PhD Student at ITAB (Institute for Advanced Biomedical Technologies) -
>>> Chieti*
>>> _______________________________________________
>>> fieldtrip mailing list
>>> https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmailman.science.ru.nl%2Fmailman%2Flistinfo%2Ffieldtrip&data=05%7C02%7Cfieldtrip%40science.ru.nl%7C52a7bcbc18b54dcae6f208de3d7a4bb3%7C084578d9400d4a5aa7c7e76ca47af400%7C1%7C0%7C639015792622990143%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=2fq6Szm2Oat02XdJzGNxL73cF%2BanVgKi7BvTkqk9B1k%3D&reserved=0
>>> https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdoi.org%2F10.1371%2Fjournal.pcbi.1002202&data=05%7C02%7Cfieldtrip%40science.ru.nl%7C52a7bcbc18b54dcae6f208de3d7a4bb3%7C084578d9400d4a5aa7c7e76ca47af400%7C1%7C0%7C639015792623033236%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=Mp%2BRgWQAFfF5rgGuopN%2F04RTmIrMKJeE2qrbCcO7dpM%3D&reserved=0
>>>
>>
>
> --
> *Strappazzon Fabio, *
> *PhD Student at ITAB (Institute for Advanced Biomedical Technologies) -
> Chieti*
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20251217/d2b5ff58/attachment.htm>