[Eeglablist] Formal challenge to SCALE and the small patch model-interpretation of ICA

[Makoto Miyakoshi]

Hi EEGLAB mailing list,

I would like to share a formal challenge to the motivation and
justification of the SCALE algorithm. Here is the preprint that
is under review.
https://urldefense.com/v3/__https://www.medrxiv.org/content/10.64898/2026.01.23.26344529v1*libraryItemId=18612094__;Iw!!Mih3wA!FZLmY-qaHhk5CbZK0UGLmLuCk_42n6AFkdmZ1zUQA_M_vMjatNF8O_d8rFtx27fdLSiV7qPhxU-jXp23-dY5AlkWR3I$ 

For those unfamiliar with SCALE, the original reference is:
Simultaneous Conductivity and Location Estimation (SCALE)
https://urldefense.com/v3/__https://doi.org/10.1016/j.neuroimage.2015.08.032__;!!Mih3wA!FZLmY-qaHhk5CbZK0UGLmLuCk_42n6AFkdmZ1zUQA_M_vMjatNF8O_d8rFtx27fdLSiV7qPhxU-jXp23-dY5Xa-bZ60$ 

As I understand it, the motivation of SCALE is as follows. In distributed
source estimation methods (e.g., LORETA, MNE, BESA; SCALE uses SCS, an
in-house developed algorithm), solutions may be iteratively computed while
parametrically varying skull conductivity, which is considered the most
unreliable parameter. Among these solutions, one selects the conductivity
value that yields the smallest cortical source distribution under
regularizations (sparse, compact, and smooth, hence SCS), under the
assumption that “smaller cortical patches are more biophysically plausible.”

The target of my criticism is very clear: these videos.
https://urldefense.com/v3/__https://www.youtube.com/watch?v=7TP-xrVokXk__;!!Mih3wA!FZLmY-qaHhk5CbZK0UGLmLuCk_42n6AFkdmZ1zUQA_M_vMjatNF8O_d8rFtx27fdLSiV7qPhxU-jXp23-dY5JxTEo50$ 
https://urldefense.com/v3/__https://www.youtube.com/watch?v=qFkUCTBan6c__;!!Mih3wA!FZLmY-qaHhk5CbZK0UGLmLuCk_42n6AFkdmZ1zUQA_M_vMjatNF8O_d8rFtx27fdLSiV7qPhxU-jXp23-dY5wSzZh3Y$ 

According to Electric Fields of the Brain (EFB, 2006), scalp EEG
measurements (and their ICA decompositions as well, naturally) are
dominated by contributions from much larger cortical patches (which may be
spatiotemporally dynamic as well). Activity from very small cortical
patches is unlikely to exceed the sensitivity limit of EEG amplifiers (on
the order of a few microvolts). Thus, such demonstration of scalp
projection would be reasonable only if shown on nanovolt color scales, not
microvolt. For this reason, I claim these visualizations are misleading (I
was misled).

To be clear, my intention is not to dismiss ICA. On the contrary, my goal
is to make physiological interpretation of ICA usable (which means, for
example, to provide a real explanation for depth bias of IC-dipoles).
Replacing the small patch model with the large patch model was unavoidable
to make physiological interpretation of ICA compatible with a standard
physiology of EEG.

Makoto