[Eeglablist] sLORETA question

Mon Nov 29 16:26:55 PST 2021

Dear John,

I'm so glad that you gave us a detailed comment. You seem a real expert, I
appreciate your input.

I remember that my colleagues in different labs showed me some kind of
LORETA images but about 50% of the cortex was 'on fire' irrespective of the
difference between white and gray matters. When I generated my first
eLORETA visualizations, which still contained 3 or 4 serious bugs including
90 degrees rotation to the right, the cortical distribution map was more
localized as if it were my favorite SPM96-style surface rendering. Now your
comment perfectly explains it--my colleagues must have used sLORETA!

Since only a few months ago I have started to use eLORETA. I'm pretty new
to this tool, so I'd appreciate it if you can tell me a couple of things.

1. In eLORETA visualization, what is color coded? From one of the LORETA
website (https://urldefense.proofpoint.com/v2/url?u=https-3A__www.uzh.ch_keyinst_NewLORETA_Misuse_Misuse.htm&d=DwIFaQ&c=-35OiAkTchMrZOngvJPOeA&r=kB5f6DjXkuOQpM1bq5OFA9kKiQyNm1p6x6e36h3EglE&m=YIIDJxTYvYfxQ5aVf47QkRtEOvugKtoZFM6RxeGQ36kbPfYtEoyxhkJwUAiBnKIA&s=cUOSw2oa_SjP4cm-mRkjhD7fl1bAdHjhGnnMb8-Rq1g&e= ), the
output is referred to as 'current density'. If I add a color bar next to
the color-rendered cortical plot, what is the appropriate title? Is it
correct to say 'Current Density' with no unit?

2. It's a bit technical question. Moreover, I may not know the exact
language to ask this question. So please bear with me.
I have been reading 'Electric Field of the Brain' (Nunez and Srinivasan,
2006) with great interest. I found that the area of the dipole sheet on the
cortex exponentially reduces distance-dependent attenuation of the electric
field. I understood it as one of the most basic forms of transducer array
effect with no phase array. I summarized what I learned in this slide.
https://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline#Physiologically_invalid_deep_dipoles.3F_.28For_130.2C000_page_views.2C_07.2F26.2F2021_Update.29
Now, when I calculate eLORETA, it takes electrode power can covariance. You
kindly explained how eLORETA uses the covariance matrix, I'm so glad to
understand it. Now my question is, does eLORETA forward model address this
transducer array effect?
For example, according to Cooper et al. (1965), expanding the source area
from a needle point to 1 square inches reduced the distance-dependent
attenuation about 10,000 times. Does eLORETA also reflect this relation
between source size and scalp potential?
If linear superposition principle also applies to the electric fields (?)
represented in eLORETA, it should be the case, but I'd like to confirm it
with someone knowledgable.

%%%%%%%%%%%%%%%%%%%%
Dear Ugo,

Thank you for your offer. I appreciate your cooperation.
I have a functional snippet to generate a figure. I'd like to give it to
you for your opinion.
Before that, I'd like to upload a movie to Youtube to demonstrate what it
does, but for that I need to obtain my colleague's permission because the
data used is his clinical data.
Please wait.

%%%%%%%%%%%%%%%%%%%%%
Dear Neri,

> 4. Interesting question - eLORETA currently provides t-statistics and
effect sizes with the pixel-based permutations, which have been useful.
(This was comparing the condition averages, though, not time points.)

Ah ok, it does it already huh. If you can point me to the source of code,
I'd appreciate it.

> A couple of concerns on the scalp approach: could there be a significant
difference at a voxel in 3D that projects to multiple electrodes (maybe at
a distance) and therefore doesn't reach significance at any scalp location,
and is therefore missed? Could nonsignificant voxels look significant if
they projected to an electrode which also received signals from a
significant voxel? Those two concerns would make me lean towards current
density stats.

Yes, signal cancellation generally occurs in the process of mixing, so
comparing the 'sources' and their mixing observed at scalp locations do not
exactly match.
Think about an extreme case--there are only two sources in a brain, the
same sine wave but with opposite phase. If they cancel each other during
the mixing process, you observe no signal at some scalp locations.
So yes, comparing source activations directly across conditions makes more
sense, but for that you need many iterations of eLORETA to generate
surrogate statistics which is not particularly beautiful.
Maybe there are already smart solutions available somewhere. I'd appreciate
it if you can tell me info about these solutions.

> 5. I would really like to extract a time series for voxels around a
specified ROI coordinate to analyse further (although maybe this is already
available in the dipfit function? I have not explored that yet).

The problems is that the obtained voxel values are very bulky. If you are
talking about only 500 ms-long data (such as MMN) sampled at 250Hz, you
only need 125 eLORETA data so it may not be too difficult. But if you are
looking for 1-hour recording data converted to continuous eLORETA, you may
need a complicated solution such as memory mapping?

Makoto

On Sat, Nov 20, 2021 at 9:07 AM Richards, John <RICHARDS at mailbox.sc.edu>
wrote:

> Someone familiar with fieldtrip might be able to answer the sLORETA vs
> eLORETA questions.  I've cc'ed a couple of people familiar with FT who know
> about these tools.
>
> My understanding:
>
> MNE works by using a lambda to change the inverse calculation--which
> calculated by itself is singular (nvoxels cf nelectrodes).  The lambda
> "typically" is based on something about the data, e.g., often the
> covariance matrix of a noise section is used, or the covariance of the
> prestimulus data.  The choice of the lambda affects the computations.
>
> The MNE has an issue with "deep" sources, it tends to emphasize shallow
> sources.  Someone noticed this and created a modification of the MNE
> formula to divide the source output by the std of the sources.  A current
> version of this is dSPM, which uses the noise covariance as the divisor.
> The sLORETA takes the MNE and explicitly estimates the STD of the source
> output and the final output is a number "similar" to a z score, eg. Is not
> current but a ratio of the sources to the std(sources).  sLORETA (and dSPM)
> alleviate the surface issue with MNE; since the deeper sources were
> smaller, the STD was smaller, and so deeper sources and shallow sources
> were equated.
>
> eLORETA is relatively unrelated to MNE, or its derivatives sLORETA, dSPM.
> eLORETA uses the inverse computation and a lambda.  However, the lambda and
> the inverse computation are found through an iterative process that
> minimizes something about the inverse formula.   This part I am not sure
> of.   One characteristic of the eLORETA is that MNE, dSPM, and sLORETA all
> depend on a lambda value from the data, whereas the eLORETA uses only the
> source model elements (electrodes, source model, head model, lead field).
> We use this to great advantage by being able to calculate the eLORETA
> filters with only the structural MRI, electrodes and source materials
> (source model, conductivity, head model, leadfield); and can apply it to
> any data relevant for that MRI.  Also these inverse correction methods can
> be used with any type of head model / leadfield (e.g., spherical, realistic
> BEM, FEM). The accuracy of the solution depends partially on the realism of
> the head model and partially on the inverse correction method.
>
> We have compared sLORETA and eLORETA informally in our work.  My
> conclusion is that the sLORETA tends to smear sources over a wider area and
> eLORETA tends to have more localized sources.  The correct lambda to use in
> sLORETA is also a big issue, since the size of the lambda is inversely
> related to the magnitude of the source output.  Re " If someone can
> enlighten us as to the difference between sLoreta and other Loreta source
> reconstruction methods, please do so." There is a fairly extensive
> literature comparing these inverse correction tools (LORETA, dSPM, MNE,
> sLORETA, eLORETA....)
>
> Re " sLoreta is an obscure Loreta decomposition, which nobody uses except
> the NeuroGuide Neurofeedback software. As far as I know, the only way to
> use sLoreta is to use the NeuroGuide software."  Its not really obscure,
> and a lot of source analysis papers have used sLORETA. These methods are in
> nearly all "source analysis" programs; MNE, Brainstorm, BESA, Fieldtrip,
> SPM, ....    Some of the procedures are matlab based and EEGLab/ERPLab
> files can be exported into their format.  Some programs like Fieldtrip have
> a proc to convert back-and-forth from EEGLab/ERPLab to Fieldtrip formats,
> though its touchy.
>
> John
>
> ***********************************************
> John E. Richards
> Carolina Distinguished Professor
> Department of Psychology
> University of South Carolina
> Columbia, SC  29208
> Dept Phone: 803 777 2079
> Fax: 803 777 9558
> Email: richards-john at sc.edu
>
> https://urldefense.proofpoint.com/v2/url?u=https-3A__jerlab.sc.edu&d=DwIFAw&c=-35OiAkTchMrZOngvJPOeA&r=pyiMpJA6aQ3IKcfd-jIW1kWlr8b1b2ssGmoavJHHJ7Q&m=HEa4HL_Gd2oumxtLr1UXBZVVpXYVDbTW-tX2Z-aVy4wX7A-neqFZpYo9EMbpdPdH&s=yEBiV2cBTrOKzvX8YIbt_X5TIK0IXFzLN8b479XN0gQ&e=
> *************************************************
>
> -----Original Message-----
> From: eeglablist <eeglablist-bounces at sccn.ucsd.edu> On Behalf Of Makoto
> Miyakoshi via eeglablist
> Sent: Saturday, November 20, 2021 11:27 AM
> To: eeglablist at sccn.ucsd.edu
> Subject: Re: [Eeglablist] sLORETA question
>
> Dear Matthew, Michael, Ivano, Neri, Vahid, and others,
>
> Thank you for your responses.
> As Michael wrote, the eLORETA implementation is taken from Fieldtrip. Arno
> developed the first EEGLAB wrapper to use eLORETA function a couple of
> years ago, which I stripped and modified several times.
> And Arno is right, it is eLORETA (published in 2005) and not sLORETA
> (published in 2002). For detail, see this official website
>
>
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.uzh.ch_keyinst_eLORETA_index.html&d=DwIFaQ&c=-35OiAkTchMrZOngvJPOeA&r=kB5f6DjXkuOQpM1bq5OFA9kKiQyNm1p6x6e36h3EglE&m=xphCaEnijvfiM1Vr6hU_VWnsK3CnX4b04UAWtiKp-WCYSEAWDOZesvN6kPlL7goo&s=dRD6EAVD8Vodcjanki5ZJXDtLrK21tSaP6GTOdU-Qs0&e=
>
> Let me quote the author's description about the difference between eLORETA
> and sLORETA.
>
>
> *...One particular member of this family is sLORETA (standardized low
> resolution brain electromagnetic tomography; Pascual-Marqui, Methods Find.
> Exp. Clin. Pharmacol. 2002,
> 24D:5-12;
> https://urldefense.proofpoint.com/v2/url?u=https-3A__protect2.fireeye.com_v1_url-3Fk-3De917986c-2Db68ca0b8-2De917d6ad-2D8680d8d86cc5-2D810f9b2b256fa9d5-26q-3D1-26e-3D87585b85-2D3b01-2D4f95-2D93ad-2D8914236ccbdb-26u-3Dhttp-253A-252F-252Fwww.unizh.ch-252Fkeyinst-252FNewLORETA-252FsLORETA-252FsLORETA-2DMath01.pdf&d=DwIFaQ&c=-35OiAkTchMrZOngvJPOeA&r=kB5f6DjXkuOQpM1bq5OFA9kKiQyNm1p6x6e36h3EglE&m=YIIDJxTYvYfxQ5aVf47QkRtEOvugKtoZFM6RxeGQ36kbPfYtEoyxhkJwUAiBnKIA&s=3yU4KuAi1QgPFDmqKgOh-To92utX5CpvyCJGk1eqiYk&e= 
> <
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.unizh.ch_keyinst_NewLORETA_sLORETA_sLORETA-2DMath01.pdf&d=DwIFaQ&c=-35OiAkTchMrZOngvJPOeA&r=kB5f6DjXkuOQpM1bq5OFA9kKiQyNm1p6x6e36h3EglE&m=xphCaEnijvfiM1Vr6hU_VWnsK3CnX4b04UAWtiKp-WCYSEAWDOZesvN6kPlL7goo&s=33s51y3NeCBlXToUvPDvo1kRbLweAyZIIFevrmH93tI&e=
> >). It is shown here that sLORETA has no localization bias in the presence
> of measurement and biological noise. Another member of this family, denoted
> as eLORETA (exact low resolution brain electromagnetic tomography;
> Pascual-Marqui 2005:
>
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.research-2Dprojects.unizh.ch_p6990.htm&d=DwIFaQ&c=-35OiAkTchMrZOngvJPOeA&r=kB5f6DjXkuOQpM1bq5OFA9kKiQyNm1p6x6e36h3EglE&m=xphCaEnijvfiM1Vr6hU_VWnsK3CnX4b04UAWtiKp-WCYSEAWDOZesvN6kPlL7goo&s=ERAVRh0Xm1kOHLhCFwaVZPzeSPmfkPLNKBn2bxvJuvU&e=
> <
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.research-2Dprojects.unizh.ch_p6990.htm&d=DwIFaQ&c=-35OiAkTchMrZOngvJPOeA&r=kB5f6DjXkuOQpM1bq5OFA9kKiQyNm1p6x6e36h3EglE&m=xphCaEnijvfiM1Vr6hU_VWnsK3CnX4b04UAWtiKp-WCYSEAWDOZesvN6kPlL7goo&s=ERAVRh0Xm1kOHLhCFwaVZPzeSPmfkPLNKBn2bxvJuvU&e=
> >), is a genuine inverse solution (not merely a linear imaging method) with
> exact, zero error localization in the presence of measurement and
> structured biological
> noise.*
>
> I don't know what he means by 'a genuine inverse solution (not merely a
> linear imaging method)' and I can't explain the difference between sLORETA
> and eLORETA. If you can explain it in a plain language, please help me
> understand it.
>
> I want to know what level of solution you need. Please answer the
> following questions.
>
>    1. Do you use continuous data (resting etc) or event-related epoched
>    data (any ERP paradigm)?
>    2. How comfortable it is for you to use these functions in Matlab
>    command line? In other words, do you need GUI?
>    3. I ask the users to input parameters for (1) latency to plot and (2)
>    color scale limit (optionally, color scheme itself?). What else do you
> want
>    to specify when generating the LORETA image?
>    4. It comes with a subtraction function i.e. eLORETA result 1 - eLORETA
>    result 2. This works, but the result is without stats. I know how to
> run a
>    stats on this result using permutation test, but it is horribly
>    inefficient. It's much easier if you determine statistically significant
>    differences on the scalp topos, and generate a corresponding eLORETA
>    difference image (a LORETA version of kernel trick) Does it make sense?
> Do
>    you think LORETA stats on current density is still necessary?
>    5. Any other general request before I start the finalization?
>
> Please let me know what you want to see. I'd like to wrap it up into a
> convenient tool for you.
>
> Makoto
>
> On Thu, Nov 18, 2021 at 8:18 PM Neri Baker via eeglablist <
> eeglablist at sccn.ucsd.edu> wrote:
>
> > Hi Arno,
> >
> > Thanks for this info. I have been using eLORETA, but I was interested
> > in whether similar functionality was available in EEGLAB (partly to
> > reduce data transfer back and forth between packages, and partly to
> > see the source code to understand exactly what the various parameters
> do).
> >
> > Thank you for the suggestion of LCMV beamforming - it looks very
> > interesting.
> >
> > Many thanks,
> > Neri
> >
> > -----Original Message-----
> > From: eeglablist <eeglablist-bounces at sccn.ucsd.edu> On Behalf Of
> > Delorme, Arnaud via eeglablist
> > Sent: Friday, 19 November 2021 12:19 PM
> > Cc: eeglablist at sccn.ucsd.edu
> > Subject: Re: [Eeglablist] sLORETA question
> >
> > sLoreta is an obscure Loreta decomposition, which nobody uses except
> > the NeuroGuide Neurofeedback software. As far as I know, the only way
> > to use sLoreta is to use the NeuroGuide software. It has not been
> > demonstrated to be superior (or inferior) to eLoreta (see the seminal
> > article
> > https://urldefense.proofpoint.com/v2/url?u=https-3A__www.frontiersin.o
> > rg_articles_10.3389_fnbeh.2014.00066_full&d=DwIFAg&c=-35OiAkTchMrZOngv
> > JPOeA&r=kB5f6DjXkuOQpM1bq5OFA9kKiQyNm1p6x6e36h3EglE&m=Zxef-biWZfrFE9qF
> > TeNsdtYkrHGyh-ZQ8zYm6OxqK8wZhat8hxJJrtPPTFIbHWD1&s=aB-mQ2PLolwFcEqW8bs
> > sg_UhLPlQLXH6xXjEJNMt8Do&e= ). It seems relatively equivalent in
> > numerical tests. If someone can enlighten us as to the difference
> > between sLoreta and other Loreta source reconstruction methods, please
> > do so.
> >
> > Because sLoreta is a rare and not widely accepted form of Loreta, I
> > would recommend instead Pascual Marqui's original eLoreta implementation.
> >
> > Also try LCMV beam forming (also available in DIPFIT) which provides
> > less smooth solutions, and is recommended for region of interest
> > connectivity analysis by brain connectivity analysis researchers such as
> Stefan Haufe.
> >
> > Arno
> >
> > > On Nov 17, 2021, at 11:15 AM, Neri Baker via eeglablist <
> > eeglablist at sccn.ucsd.edu> wrote:
> > >
> > > Hi Makoto
> > >
> > > This sounds really useful. I am also interested in trying it out.
> > >
> > > Kind regards,
> > > Neri
> > > ________________________________
> > > From: eeglablist <eeglablist-bounces at sccn.ucsd.edu> on behalf of
> > > ivano triggiani via eeglablist <eeglablist at sccn.ucsd.edu>
> > > Sent: Thursday, November 18, 2021 7:43:30 AM
> > > To: Makoto Miyakoshi <mmiyakoshi at ucsd.edu>; eeglablist at sccn.ucsd.edu
> > > <eeglablist at sccn.ucsd.edu>
> > > Subject: Re: [Eeglablist] sLORETA question
> > >
> > > Dear Makoto,
> > >
> > > I would be interested as well.
> > >
> > > Thank you,
> > >
> > > Ivano
> > >
> > > On Wed, Nov 17, 2021, 3:23 PM Makoto Miyakoshi via eeglablist <
> > > eeglablist at sccn.ucsd.edu> wrote:
> > >
> > >> Dear Matthew,
> > >>
> > >> Dipfit supports eLORETA but it is more like a proof of concept. I
> > >> recommend you check it out first.
> > >> I have stripped the function and wrote my own wrapper. If you are
> > >> interested, I can upload it online so that you can try it out. It
> > >> is not an EEGLAB plugin so does not come with a nice GUI. My code
> > >> also allows voxel-level subtraction to show current density
> > >> differences between two conditions. Let me know if you want to try it
> out.
> > >>
> > >> Makoto
> > >>
> > >> On Tue, Nov 16, 2021 at 12:28 PM Gunn, Matthew P via eeglablist <
> > >> eeglablist at sccn.ucsd.edu> wrote:
> > >>
> > >>> Hello,
> > >>>
> > >>> Does anyone know if there are any programs like sLORETA in EEGLAB
> > >>> or an APP that MATlab has that could do this process. I know of
> > >>> erpsource; but, didn't know if this tool existed or if a team is
> > >>> currently working on
> > >> one.
> > >>>
> > >>> Thank you for your time,
> > >>>
> > >>> Matt G.
> > >>> _______________________________________________
> > >>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > >>> To unsubscribe, send an empty email to
> > >>> eeglablist-unsubscribe at sccn.ucsd.edu
> > >>> For digest mode, send an email with the subject "set digest mime"
> > >>> to eeglablist-request at sccn.ucsd.edu
> > >>>
> > >> _______________________________________________
> > >> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > >> To unsubscribe, send an empty email to
> > >> eeglablist-unsubscribe at sccn.ucsd.edu
> > >> For digest mode, send an email with the subject "set digest mime"
> > >> to eeglablist-request at sccn.ucsd.edu
> > >>
> > > _______________________________________________
> > > Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > > To unsubscribe, send an empty email to
> > > eeglablist-unsubscribe at sccn.ucsd.edu
> > > For digest mode, send an email with the subject "set digest mime" to
> > > eeglablist-request at sccn.ucsd.edu
> > > _______________________________________________
> > > Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > > To unsubscribe, send an empty email to
> > > eeglablist-unsubscribe at sccn.ucsd.edu
> > > For digest mode, send an email with the subject "set digest mime" to
> > > eeglablist-request at sccn.ucsd.edu
> >
> > _______________________________________________
> > Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > To unsubscribe, send an empty email to
> > eeglablist-unsubscribe at sccn.ucsd.edu
> > For digest mode, send an email with the subject "set digest mime" to
> > eeglablist-request at sccn.ucsd.edu
> > _______________________________________________
> > Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > To unsubscribe, send an empty email to
> > eeglablist-unsubscribe at sccn.ucsd.edu
> > For digest mode, send an email with the subject "set digest mime" to
> > eeglablist-request at sccn.ucsd.edu
> >
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>