[Simnibs-discuss] Differences in Efield depending on Processing

[Axel Thielscher]

Joel,


when you have small electrodes as is usually the case for 
multi-electrode montages, a simple solution is to lower the conductivity 
of the modelled electrode materials. This will not affect the field 
distribution in the brain, but reduce the amount of shunting when you 
place many electrodes on the head.

This solution will be fine for small electrodes, where it is not 
important that the modelled electrode properties correspond well to the 
ones used in the real experiment.


Best regards,

Axel


On 15-11-2022 20:33, Joel Upston wrote:
> Axel,
>
> Thanks for the clarification. In the inital post I modeled the second 
> pathway with only the electrodes that were used in the pipeline, but 
> your reasoning makes sense about the conductive paste being used 
> essentially covers a majority of the head.
>
> I reran using all modeled electrodes, and the electrodes not used are 
> modeled with 0 mA current. I have attached the images where the right 
> is the non-leadfield model and the left is the leadfield model. Where 
> now the fields look very similar between the two different pipelines. 
> It looks like it does lower the field by about ~10% (max given with 
> all electrodes 0.47 V/m and only used electrodes 0.521 V/m). This 
> would be consistent with what I have found with other stimulations. 
> The benefit is that although it does seem to change the magnitude of 
> the electric field, the spatial distribution does not seem to change.
>
> So, if we were to target a specific value efield magnitude for a ROI, 
> any optimization problem using the leadfield might be incorrect due to 
> the lower magnitude of the leadfield processing. I wonder how to 
> correct for this during any optimization, or maybe not necessary since 
> it could be that the scale is similar no matter what stimulation is 
> modeled. So maybe you could just correct in post optimal solution 
> found. Do you think it is worth looking at the spherical model with a 
> possible additional tissue representing the conductive paste compared 
> to the one without the additional conductive paste tissue, to get a 
> sense of the scale estimate?
>
> Thanks,
>
> Joel
> ------------------------------------------------------------------------
> *From:* Axel Thielscher <axelt at drcmr.dk>
> *Sent:* Wednesday, November 2, 2022 10:21 AM
> *To:* Joel Upston <jupston at unm.edu>; discuss at simnibs.org 
> <discuss at simnibs.org>
> *Subject:* Re: [Simnibs-discuss] Differences in Efield depending on 
> Processing
>
> 	
> You don't often get email from axelt at drcmr.dk. Learn why this is 
> important <https://aka.ms/LearnAboutSenderIdentification>
> 	
>
> *  [EXTERNAL]*
>
> **
>
> Hi Joel,
>
>
> a question for better understanding: When you set up the 2nd pathway, 
> do you then model ALL electrodes of the EEG10-20 system, or only those 
> which you actually need for the montage?
>
>
> I am asking because adding many well-conducting electrodes to the 
> model will slightly increase the amount of current shunted through the 
> scalp(+electrodes). Think of it as covering many parts of the scalp 
> with well-conductive paste. We never quantified the size of this 
> effect, but it will cause the leadfield e-fields in the brain to be 
> consistently lower.
>
>
> Best regards,
>
> Axel
>
>
> On 19-10-2022 21:07, Joel Upston wrote:
>> Hi there,
>>
>> I have noticed a systematic difference between the two different 
>> lines of simulating the same currents, electrodes and underlying 
>> dataset. One is through processing through the leadfield and using 
>> only the electrodes and currents as given and the other is through 
>> only modeling those electrodes and currents without the leadfield. 
>> Now the 2^nd  pathway (no leadfield) is giving higher E_mag than the 
>> leadfield version (~7-12%), the spatial maps look similar but just 
>> the intensity differences. I think it has to do with the scaling that 
>> is done after the solve by scaling the estimated current calibration 
>> error, which if I understand correctly isn't done in the leadfield 
>> processing.  If this is the case I am trying to establish which would 
>> be more accurate to use in your view if I need a specific target 
>> value of (0.3 V/m for example). I have attached an image of a case 
>> where this is shown with the leadfield processed is on the left and 
>> the non-leadfield is on the right. I have tried this on about 5 
>> different datasets and the pattern is consistent.
>>
>> Thanks,
>>
>> Joel
>>
>> _______________________________________________
>> Simnibs-discuss mailing list
>> Simnibs-discuss at drcmr.dk  <mailto:Simnibs-discuss at drcmr.dk>
>> https://mailman.drcmr.dk/mailman/listinfo/simnibs-discuss  <https://mailman.drcmr.dk/mailman/listinfo/simnibs-discuss>
> -- 
> Dr. Axel Thielscher
> Professor of Neurophysics and Neuroimaging
> Danish Research Center for Magnetic Resonance
> Copenhagen University Hospital Hvidovre
> DK-2650 Hvidovre, Denmark
> www.drcmr.dk  <http://www.drcmr.dk>
> &
> Department of Health Technology
> Technical University of Denmark
> DK-2800 Kgs. Lyngby
> http://www.healthtech.dtu.dk/  <http://www.healthtech.dtu.dk/>

-- 
Dr. Axel Thielscher
Professor of Neurophysics and Neuroimaging
Danish Research Center for Magnetic Resonance
Copenhagen University Hospital Hvidovre
DK-2650 Hvidovre, Denmark
www.drcmr.dk
&
Department of Health Technology
Technical University of Denmark
DK-2800 Kgs. Lyngby
http://www.healthtech.dtu.dk/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.drcmr.dk/pipermail/simnibs-discuss/attachments/20221116/64fd22c0/attachment-0001.html>