X-Message-Number: 26965
From: 
Date: Thu, 8 Sep 2005 02:44:56 EDT
Subject: Uploading technology (1.iii.2) Dendrite processing. 

Uploading technology (1.iii.2) Dendrite processing.

Dendrites  was seen as simple conductors working along the cable theory 

scheme or at best,  the Hodgkin-Huxley model. Two add on have come into play 
these 
past years:  First, there was the realisation that far more currents are 

working here with  important effects on the firing pathern of neurons. The 
second 
is that  differential equations, must not be dealt with special, continuous  
solutions,  they must be worked out with general, discontinuous ones. On  the 

practical ground, that translate into signal boosts by potential gated  channels
and collective synapse spine effects.

At first, all of  that was seen as a way to amplify a distant signal so that 
it would be not  negligible when coming to the soma. In fact, in the general 
case, the  conductivity and amplification is such that a distant signal comes 
with the same  power as a nearby one. On a simulation ground, that may be 
approximated at the  first level as a lossless transmission.

This active transmission  open yet a pandora box, there are many 

possibilities for modulation and control.  The dendrite is no more a passive 
cable device. 
It is a full processor of its  own.

The first control is with tension gated channel and gap  junctions. Gap 

junctions can impose a regional depolarisation or  hyperpolarisation able to 
work 
on a part of the dendritic tree or even a full  neuron group. this can hinder 
or facilitate the postsynaptic potential  propagation. This is an area 
selective effect.

Potential gated  channels, using Na+ can modulate the conductivity up to zero 
loss and beyond,  that is signal amplification depending on their density. 
They can work too as  stepping stone for Ca++ channels with harder to get  
thresholds.

Ca++ ion can act on the neurofibrils holding the  membrane and for example 
expand or squeeze the dendrite branch diameter. This  may hinder of facilitate 
the transmission at a branching point, giving more or  less weight to synapses 
on the distal part.

Back propagating action  potentials may help to depolarise the membrane and 
trigger a postsynaptic  potential, particularly on spines. As seen in the soma 
part, this back  propagating AP can induce some resonance effects.

The general  solutions of differential equations include the possibility of 
coherent effects  between synapses. There are two possibilities:
A/ The potential gated  channels of a synapse on a spine could fire a 

postsynaptic potential even  without presynaptic signal, particularly if there 
is a 
back propagating AP and  some gap junction effects. This is a contribution to 
the lossless  transmission/amplification.
B/ A dendrite branch could harbor a set of  synapses all linked to the same 
axon and so firing at the same time. This  synchronized activity would trigger 
other synapses not sufficiently excited by  themselve to fire independently. 
If a dendrite tree would then be broken into a  number of functional blocks, 
each under the rulling of a synaptic pool working  toghether. This is an 

alternative to the Hebbian rule of reinforcement for  neuron firing in 
synchrony, 

here the reinforcing effect would be spatial and  under the control of a set of
synapses of the same neuron. Synchronised firing  and dendrite domains must be 
taken with a pinch of salt yet.

There  is fast sync. for synapse firing with the sole power of potential 

gated and fast  chemical channels. This is synchronisation at the some 
millisecond 
scale.  Another firing simultaneity may be defined for G-protein channels, 

with  simultaneity at the second duration scale. The common dendritic domain may
be  different here. A third scale is given by second messenger gated 

channels, here  synapses synchronisation may be at the minute or more. This is a
collective  behaviour extending may be at regional scale, for example a 
microcolumn, that  is, up to 100,000 neurons.

Allof that is skewed towards excitatory  functions, inhibitory ones are less 
glorious, they are neverthless a key  component of all neuronal function. 

GABA-A synapse are hyperpolarising near the  outside potential, using for that 
Cl- 
ions. They work as short-cut throughout  the membrane. They cancel everything 
upstream and are nearly without effect on  down stream synapses. When on the 
axon hillock, they can cancel even a  progressing action potential. Put at the 
root of a dendrite domain, they can  control it. They are yes/no gate for a 
single spine, a dendrite domain or a full  neuron.

GABA-B inhibitory synapses produce a larger  hyperpolarisation and work as 
excitatory ones, simply the other way. They are  threshold enhancer and work 
both, upstream and downstream. To  summarize:

The input parameters are so: The post-synaptic  potentials, the back 
propagating action potential and the gap junction  potential.
 
The processing elements are: The Na+ and Ca++ channel densities and  

repartitions. The branching impedance, the spine density working as amplifier  
and the 
coherent spine density. The yes/no GABA-A synapses and thrshold  controler 
GABA-B. A dendrite tree is broken into a number of functional domains,  the 

domains limits evolve in time and for different currents, particularly for  
fast, 
g-protein and second messenger channel classes. May be ten to fifteen  current 
must be worked out simultaneously.
 
The output parameters are: The computed fast currents at the domain base  
trunk, the back propagating potentials, the dendrite junction diameter  
evolution, the changing channel dentity and localisation.
 
This conclude the analysis of the neuron information processing for the  
uploading purpose. The next step is to turn this wish list into a number of  
functions with their requested I/O and memory,  so that this system can be  
implemented on an ASIC neuromorphic chip or an existing FPGA.

Yvan  Bozzonetti.




 Content-Type: text/html; charset="US-ASCII"

[ AUTOMATICALLY SKIPPING HTML ENCODING! ] 

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=26965