This is the weblink for the models and algorithms associated with the
paper


Mechler F, Victor JD (2012) Dipole characterization of single neurons
from their extracellular action potentials. J Comput Neurosci 

Authors: Ferenc Mechler & Jonathan Victor
e-mail:	 fmechler@med.cornell.edu

The DipoleLocalizationKit archives in its 3 parts: (i) Matlab code of
the neuron localization algorithm; (ii) the lead fields (and lead
potentials) of 4 Thomas tetrodes precomputed in a FEM model and saved
in Matlab [struct] format; and (iii) real extracellular action
potential (EAP) data from 3 single neurons in visual cortex, plus the
results of the localization algorithm applied to them; all in Matlab
[struct] format. An enclosed ReadMe document explains the use of these
tools and data.

We localize a single neuron from the spatial sample of its EAP
amplitudes recorded with a multisite probe (with 6 or more independent
measurement sites or channels, e.g., a silicon polytrode, a stepped
tetrode, etc.) This is an inverse problem and we solve it by fitting a
model to the EAPs that consists of a volume conductor model of the
neural tissue (known), a realistic model of the probe (known), and a
single dipole current source of the model neuron (unknown). The dipole
is free to change position, size, and orientation (a total of 6
parameters) at each moment during the action potential.

This algorithm numerically solves the dipole optimization problem on a
discrete grid in two, separable, stages. The first stage, a linear
dipole moment optimization (location is fixed), is solved everywhere
on the grid using the lead fields (LFs). The LFs are required input to
this algorithm and assumed to be pre-calculated. (Software to compute
LFs is not provided here. LFs can be calculated by, e.g., the finite
element method; for details of LF computations, see Mechler & Victor
(2012).) The second stage, a nonlinear global optimization of the
source location, is solved using Tikhonov regularization.