Differential EEG
kpaulc
kpaulc at removeearthlink.net
Wed Mar 3 23:06:44 EST 2004
This is an incomplete "draft" that I'd
saved before Windows came to a
full-"stop", and I had to reboot.
I'll follow-up if I can.
K. P. Collins
"Doktor DynaSoar" <targeting at OMCL.mil> wrote in message
news:q5pc40d2nbkmmopth3r235rqt0hv6jmkbg at 4ax.com...
> On Wed, 03 Mar 2004 19:57:49 GMT, "kpaulc" <kpaulc at earthlink.net>
> wrote:
>
> } "k p Collins" <kpaulc@[----------]earthlink.net> wrote in message
> } news:zMS0c.29106$hm4.12683 at newsread3.news.atl.earthlink.net...
> } > "NMF" <nm_fournier at ns.sympatico.ca> wrote in message
> } > news:3ft0c.2257$qA2.152895 at news20.bellglobal.com...
> } > > [...]
> } > [...]
> }
> } > in condition 1, activation at locus A de-
> } > creases, but activation at locus B can go up or
> } > down, and in condition 2, activation at locus A
> } > increases, but activation at locus B can go up
> } > or down.
> } >
> } > So nothing can be said about locus A, without
> } > knowing what goes on at =every= "locus B".
> } >
> } > It does an good to look at any "locus A" with-
> } > knowing what's going on at =every= "locus B".
> } > [...]
> }
> } All existing EEG usage runs headlong into this
> } 'difficulty', but the 'difficulty' is 'easily' overcome
> } by doing "Simultaneous-Differential EEG" [SDEEG;
> } which I'm discussing, for the first time, here].
>
> You go on to describe coherence calculation using average reference.
>
> Mounier R. [Real-time statistical analysis of electroencephalographic
> signals: proper and mutual power density; coherence] French.
> Agressologie. 1969 Jun 16;10:Suppl:519-24.
>
> Had you included phase in the frequency dependent measure, it would
> have been complex coherence.
>
> Grindel' OM, Gershman SG, Boldyreva GN, Vakar EM, Malina ZA.
> [Intercentral relations in the cortex of the human cerebral
> hemispheres according to findings of the spectrum of coherence and
> phase spectrum of the EEG] Russian.
> Zh Vyssh Nerv Deiat Im I P Pavlova. 1973 Jul-Aug;23(4):771-81
>
> Coherence requires re-referencing to average reference as a matter of
> course. However, re-referencing can be and is done to any arbitrary
> electrode or collection of them. Recording montage does not dictate
> analysis montage. In addition, many techniques, such as current source
> density, independent component analysis and Laplacian smoothing, do a
> fine job of extrapolating source placement at arbitrary distances
> between actual electrodes (though they are limited in number of
> sources so specified to some function of the original number of
> recording electrodes).
>
> I've used high dentity recording with re-referencing (Electrical
> Geodesics). 128 channels doesn't give you significantly better answers
> than 64 in terms of source localization, and no better than 32 in
> terms of localization of effects. What you gain in spatial resolution
> you lose in statistical power due to multiple comparisons.
>
> Even with 32 channels, two of them being VEOG and HEOG, you have 30
> active channels, and that means 435 unique comparisons per test.
>
> } Aside: Im 'smiling' as I write this because it'll be
> } just one more thing that gets 'borrowed' by the
> } 'profit seekers', and one more thing that, when
> } I 'protest' its being 'borrowed', will result in
> } folks 'thinking' that I'm 'being untoward', when
> } all I'm doing is what I can to Earn a place for
> } NDT's understanding.
>
> Although computing power to calculate coherence was not commonly
> available before such machines as the PDP-1, the concept was already
> well understood. Even Berger had noted similarities and differences in
> modulation of oscillatory activity recorded simultaneously at
> different locations. That was about 75 years ago.
I stand on what I've posted.
I'd use at least 3 simultaneous sets.
It's called "Triangulation".
The difference between SDEEG and what
you've discussed is that the 'differentials'
are all cross-correlated with respect to
the same 'time' [you like 'time' :-].
When you use only one set, there's not
enough information in-there to triangulate.
In triangulation, you have to have inform-
ation that is "three"-to-"one".
What you've described is "one"-to-"one",
with some guessing with respect to the
'fog' of unknown-summations.
When you use "three"-to-"one", each of the
three sets =simultaneously= records a 'picture'
of the one thing. So each of the three sets
independently contains 'differentials' with re-
spect to the one thing.
The independence and completeness of each
set is necessary - so that the 'fog' of unknown
summations can be cleared. [The completeness
of each set captures the unknown summations
in three different ways, yielding the 'differentials'
that cen be cross-correlated to disclose all of
the unknown summations. When you use only
one set of electrodes, the unknown summations
are captured, but only in one way, so, instead
of a collection of three cross-correlated 'ramps',
you get one collection of 'ramps', which is in-
sufficient to eliminate the 'fog' of unknown
summations. [We're not talking RADAR,
here. There's no 'time'-'constancy', so the data
are 'mushy'. The only way to un-'mush' the
data is to impose 'constancy' through the use
of multiple =independent= [but 'equivalent', in
that they each record the same unknown sum-
mations simultaneously] sets.
[What's the difference between what you're
discussing and what I'm discussing?
One can cross-correlate between x elec-
trodes in one set of electrodes, no?
Yup.
But one gets only one set of 'ramps' [ampli-
tude and frequency-summation gradients].
One can permute, using a varied subset of
electrodes, but it's all just the =same= set of
'ramps' ['differentials'].
What's the difference in using three independ-
ent sets of electrodes?
It gives three independent sets of 'ramps'.
Why can't the data recorded from a single
set of electrodes give as many sets of 'ramps'
as there are permutations of subsets of elec-
trodes?
Be-cause, no matter how the electrodes are
permuted, the locations of the electrodes on
the scalp remain fixed - so it's all just the one
set of 'ramps' - there's nothing analogous to
what's available to for cross-correlation in the
SDEEG technique - SDEEG's three sets of
'ramps', each of which 'sees' the same record-
ed activity from a different perspective, not
a single perspective.
Start small, with only three electrodes in each
of the three sets.
It's the three sets of =independent=, but
complete, 'differentials' that allow the
'differentials' to triangulate the =one= thing.
The one thing is the 3-D-continuity.
The only thing I'm not sure of is, as I dis-
cussed in my prior post, whether the sens-
itivity of the recording electrodes is sufficient
to 'see' the entirety of the brain's 3-D depth
[and that's an Engineering Problem].
[I'm having =HUGE= difficulty accessing
bionet.neuroscience [=any= NGs], so
every msg might be my 'last'. Man-O-man!
I dislike 'working' this way.]
K. P. Collins
More information about the Neur-sci
mailing list