[FoRK] Annihilation from within...

Eugen Leitl < eugen at leitl.org > on > Thu Dec 7 01:20:24 PST 2006

On Thu, Dec 07, 2006 at 09:25:50AM +0100, Dave Long wrote:

> - external measurement is noisy and low-res.  Scalp measurements 

Yes. Your limits is physics. Moreover, a BCI is a bidirectional
interface, and by noninvasive means you can only get output,
and the only input is high-power, and merely disruptive (TMS).
I do expect nice things with lots of massive DSP from EEG or
even MEG -- it might be even enough to produce a realtime 6DOF
vector, or a low-bandwidth ASCII channel. If anyone has
access to http://ieeexplore.ieee.org/Xplore/login.jsp?url=/iel5/10382/33012/01547485.pdf?arnumber=1547485
I'd rather like to see that paper.

> naturally so, fMRI because there are intrinsic tradeoffs between scan 
> rate, resolution, and field strength[0] (even if one thinks in russian, 
> "Привет Товарищ" takes a couple of minutes, as current 

I haven't heard anyone calling someone else comrade in Russia in a long time.
Unless you're a member of the Communist party, maybe.

> practice has bandwidths of spb, not bps)
> - internal measurement, as pointed out, requires brain surgery.  

Invasive methods are risky, expensive, have side effects (unless
someone figures out how to do transdermal I/O and get in inductive power
in a small package -- the recent resonant approach to wireless power
would come handy here) so they're only currently limited to research
purposes during surgery, or for highly experimental prototypes
for locked-in patients.

> Wireless implants currently seem like the best way to go, in terms of 
> avoiding tissue response and hygiene issues.  Luckily for this field, 

Don't forget that you'll need elastic electrodes which don't degrade
in situ and/or do bad things to the surrounding tissue. Given that
the brain floats and flexes, and has the consistency of a jelly (the
bouncy-rubbery things in jars you see are fixated, the real thing oozes
out once you do an incision into dura mater).

> the head has had the best track record in terms of retaining abutments, 
> but as far as I am aware, these currently fall into two distinct 
> categories:  (a) structural support, and (b) temporary for experiments.
> - except among rats, for whom the distinction between "temporary" and 
> "lifetime" is pretty slim.
> -Dave
> :: :: ::
> [0] what's up with the low-T MRI stuff?  Have they figured out a way 

You can completely forget MRI, and even MEG as long as both
need cryogenics and you're cm removed from the cortex. 

> around the tradeoffs, or just a way to practically achieve some 
> resolution with long enough scans?

Long enough scans and realtime are mutually exclusive. Rapid
processes are for galvanic, magnetic and NIR pickups.
> [1] once you have the implant in there, what signal were you thinking 
> to relay?

If you can get the raw activity output from a multichannel (N>>1000) array
that doesn't degrade and drift in the tissue your task is half done
already. The rest is "only" massive DSP.

Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org
ICBM: 48.07100, 11.36820            http://www.ativel.com
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE

More information about the FoRK mailing list