Extracellular stimulation

[gwalckie]

Dear Ted and Costas,

with the help of your discussion, I managed to build a model where a single axon is excited by some extracellular potential and play around with it.

I tried to complicate it while using the geomatry of the model given earlier in this topic (the McIntyre one, with nodes, MYSA, FLUT and STIN sections), and I'm facing several difficulties:

-while using the forall command to set the e_extracellular(x), the extracellular potential is set up according the declaration order (first to all the nodes, then to all what is called MYSA in McIntyre model and so on...). Is there a way to set them according to the spatial order rather than the declaration order?
-while doing that the extracellular potentials would be set to each section, but as these section have different length, the spatial position of the fixed extracellular potentials will not be regular. Is there a way to specify the extracellular potentials at some points (at the nodes for example) and then to ask Neuron to extrapolate linearly the value between these points?

Thanks a lot for your answers

Greg

[ted]

-while using the forall command to set the e_extracellular(x), the extracellular potential is set up according the declaration order (first to all the nodes, then to all what is called MYSA in McIntyre model and so on...). Is there a way to set them according to the spatial order rather than the declaration order?

You could create a new SectionList, to which you append sections in whatever order you
like. Then
forsec seclistname print secname()
to verify that you got the sequence you wanted, and if you did, assign the field like so
forsec for (x,0) seclistname { statement that assigns extracellular potential }

-while doing that the extracellular potentials would be set to each section, but as these section have different length, the spatial position of the fixed extracellular potentials will not be regular. Is there a way to specify the extracellular potentials at some points (at the nodes for example) and then to ask Neuron to extrapolate linearly the value between these points?

Why waste time interpolating the field at the correct locations at every time step? Why
not calculate the field at the correct locations in the first place?

[gwalckie]

Dear Ted,

I'm amazed by the rapidity of your answers, thanks a lot!
I'll try to applied your advices as soon as possible, and agree for the new thread if you think it's more appropriate.

[costas]

Well

I am new to this but I can tell you what I did to resolve the situation you describe in your first question:

Basically I set a

Code: Select all

forsec xyz (x>0)

procedure that starts at x=0 and incrementally progresses to x->1 in single segment steps. At each of these segments you can define the value of e_extracellular(x). This way the value e_extracellular(x) is set for increasing x.

The code for this would be:

Code: Select all

forsec xyz {
    	for (x, 0) {
		for nn = 0, nseg-1 {

			min1 = nn/nseg
			max1 = (nn+1)/nseg

				if (x > min1 && x < max1) {ii = nn}

				e_extracellular(x) = eex.x[ii][0]

				     	     }
			 }
	          } 

What this procedure does is: for a specific section xyz it applies the values of a predefined matrix eex to e_extracellular(x) for increasing x. In the same manner you could use the vector_play feature to define the values of e_extracellular(x).

Regarding your second question I came across the same issue. The way I do it is feed the morphology in Matlab and find out the position of each node in space. Then I construct externally a distance_vector which is a matrix with the distance between each node of the neuron and the stimulating electrode. I am pretty sure that this is not the most efficient way but it worked for me.

Costas

[ted]

Are you sure about this?

Code: Select all

forsec xyz {
    	for (x, 0) {
		for nn = 0, nseg-1 {
etc.

Also, it's not clear why eex.x[ii][0] (i.e. why two indices?).

For a one-dimensional eex, it does work but it risks error, and efficiency could be
improved. To see why, consider this test:

Code: Select all

create axon
access axon
proc test() { local count
  count = 0
  nseg = $1
  for (x, 0) {
    for nn = 0, nseg-1 {
      min1 = nn/nseg
      max1 = (nn+1)/nseg
      if (x > min1 && x < max1) {ii = nn}
//      e_extracellular(x) = eex.x[ii][0]
      printf("nn = %d -- e_extracellular(%g) = eex.x[%d]\n", nn, x, ii)
      count+=1
    }
  }
  print "nseg = ", nseg, " requires ", count, " iterations"
}

Here's what happens when test() is called with arguments 1, 3, and 5:

Code: Select all

oc>test(1)
nn = 0 -- e_extracellular(0.5) = eex.x[0]
nseg = 1  requires 1  iterations

oc>test(3)
nn = 0 -- e_extracellular(0.166667) = eex.x[0]
nn = 1 -- e_extracellular(0.166667) = eex.x[0]
nn = 2 -- e_extracellular(0.166667) = eex.x[0]
nn = 0 -- e_extracellular(0.5) = eex.x[0]
nn = 1 -- e_extracellular(0.5) = eex.x[1]
nn = 2 -- e_extracellular(0.5) = eex.x[1]
nn = 0 -- e_extracellular(0.833333) = eex.x[1]
nn = 1 -- e_extracellular(0.833333) = eex.x[1]
nn = 2 -- e_extracellular(0.833333) = eex.x[2]
nseg = 3  requires 9  iterations

oc>test(5)
 . . . many omitted lines . . .
nseg = 3  requires 25 iterations

So the amount of work increases as the square of the number of segments in a section,
and most of the assignments are incorrect. It works only because the incorrect
assignments fortunately don't overwrite the previously made correct assignments, and
the last assignment just happens to involve the right index into the eex Vector.

If eex has the same number of elements as the number of internal nodes in your model,
and the sections in the sectionlist are in the proper sequence, this should work:

Code: Select all

ii = 0
forsec seclist for (x,0) {
  e_extracellular(x) = eex.x[ii]
  ii += 1
}

[costas]

Dear Ted

Thanks a lot for the indication - you are right! As mentioned in my earlier note I was sure this was not the most efficient code for the job...

I implemented the loop you indicated and I have a question: It seems to work only if I define the operation as

e_extracellular(x) = eex.x[ii][0]

or

eex.x[ii][0] = e_extracellular(x)

AND NOT when using the (certainly more straightforward) notation you suggested:

e_extracellular(x) = eex.x[ii]

or

eex.x[ii] = e_extracellular(x)

The matrix is defined as eex = new Matrix(n_tot, 1). As in this case I am using 1-dimensional matrices it is not a problem but I am not sure why I receive the following error message when using the second latter notation:

"... not right number of subscripts in ..."

All the best, Costas

[ted]

It seems to work only if I define the operation as
e_extracellular(x) = eex.x[ii][0]

because you declared eex to be a Matrix, and NEURON's Matrix class is by definition
two dimensional--each element must be referenced by two indices.

So you're not using the Vector class's play() method. Does your code instead transfer a
new value from eex to e_extracellular by executing a hoc assignment statement at each
fadvance()? That would be much less efficient.