Conductivity values for head model

Hi all!
Happy new year :slight_smile:

Currently I am working on hdEEG data to perform source localization. I am looking for the "right" volume conductivity values to solve the forward problem in OpenMEEG. Recently I have heard, seen and read about 3 different sets of values (there must be many others) :

  1. Alex Gramfort mentioned [0.3, 0.006, 0.3] during his talk at PracticalMEEG (Dec 2019) I believe that these are the defaults in MNE.
  2. Colleagues from Montreal are using [0.33,0.0125,0.33] (Hedrich et al. NIMG 2017)
  3. The Brainstorm defaults are [1,0.125,1]

Do you know how much these values impact the solution? (would you expect these three sets of values to give much different results?) Currently, would you still recommend to use the defaults ([1,0.125,1])? or anything else?

Thanks!
AnneSo

Hi Anne,

Happy New year!

For a BEM solution the ratios of the conductivities define the solution more or less.
Brainstorm uses an 1/80 ratio between skull and skin/brain. However this conductivity is highly variable and debated. A good reference for that is the Review on solving the forward problem in EEG source analysis by Hallez et al. That paper gives some good sources for the error made by changing this value.

Kind regards,
Steven

Ps: this is my first post so my apologies if it doesn't fully adhere to the normal reaction standards

We discussed this with @Alexandre after CuttingEEG. He said that the values he has no strong opinion about this. The values he presented are not really recommendations, but what Matti Hamalainen chose as the default values in MNE.
Alex thinks that this parameter should not have too much impact in MNE/dSPM source estimation...

The default values in Brainstorm ([.33 .0042 .33] = [1 0.125 1]) were defined by @John_Mosher and @Sylvain.
Can you remind us where these values came from?

Hi,
I've looked into this some time ago. My take on the problem is that:

  • Measuring conductivity of biological tissues is not simple and it is not even clear what a "recomended" measurement setup should be.
  • Even when measured precisely, conductivities are not flat nor linear regarding frequency, so the model simplifies the measured conductivies into flat constant. Avoiding inductive and capacitive effects simplifies things.
  • Actual conductivity values depend on the publication, but the most important idea is to consider the effect of a high(ish) conductive layer wrapped in a lower(ish) conductive layer, all wrapped in another high(ish) one. This high-low-high is one key aspect to later eeg/meg problems when doing source reconstruction. Sometimes this idea is mentioned as a the 1:1/80:1 model of the head.
  • There are not that many papers on this subject, but one that pretty much everybody knows about is Gabriel et al. The dielectric properties of biological tissues: II.Measurements in the frequency range 10 Hz to 20 GHz. While writing this I've found this FieldTrip page, which might be helpful.
    There is a web application that can show you some conductivity values for different tissues. It might come handy.
    http://niremf.ifac.cnr.it/tissprop/#over
    For instance I've got these results for grey matter.
    image
    Hope it helps.
1 Like

Hi Francois,

I want to change the conductivity values for the head model using openMEEG. If I understood correctly from these posts is that the default values in brainstorm ([1, 0.0125, 1]) are actually ratios, right? I want to create a model with new conductivity values (S/m units) but I'm not sure if I put them in the interface it will interpret it as a ratio rather than the actual value. Can you clear this for me please?

Thank you!

Valentina

What is important with these values is their relative ratios.
Multiply them all with a fixed factor, and you should get exactly the same results.
So if you enter values in S/m, it will work in the same way.

1 Like

Hi Francois,

Thank you for your answer. Actually I tried that comparison with Brainstorm's default values both in ratios and in S/m ([1 0.0125 1] and [0.33 0.0042 0.033]). I created a head model for each set with all the other parameters exactly the same, but the values in the gain matrices are completly different.

Is there something that I'm missing? Or it would be better just to transform my S/m values to ratios?

Thanks!

The two sets of values you tested are not exactly equivalent.

[0.33 0.0042 0.033] S/m => Divided by 0.33 => [1 0.0127 0.1]

With OpenMEEG, I thought the S/m and relative values should given the same result, as OpenMEEG always normalizes the conductivity values based on the maximum.
For the Brainstorm 3-shell spherical model, the values should be in S/m.

1 Like

Perfect! Thank you for your help