|
|
Christophe PouzatLaboratoire de Physiologie CérébraleCNRS UMR 8118 UFR biomédicale de l'Université René Descartes (Paris V) 45, rue des Saints Pères 75006 Paris France tel: +33 1 42 86 38 28 fax: +33 1 42 86 38 30 christophe.pouzat@univ-paris5.fr sip:christophe.pouzat@ekiga.net |
| Cette page en Français |
Sébastien Joucla and I are developing statistical and computational methods for analyzing extracellular data and now calcium imaging data recorded from two preparations: rats / mice brain slices and the insect olfactory system. |
|||||||||
|
Like
many others we
are
interested in the way populations of neurons represent
information, for instance how the projection neurons of the insect
antennal lobe represent different odors. We are particularly trying to find out if, and
if yes how, correlations in
neuronal activity (e.g., synchronization) have any
role in
this representation.
In order to study this kind of question we obviously have to record
from several neurons at once.
Then we roughly have the choice between two techniques, optical imaging
and extra-cellular
recordings. We chose first the latter because it is a cheap method and, more to the point, it potentially enjoys a fine time
resolution (in the
millisecond range or bellow). I wrote here "potentially" because the raw data we collect need to be
rather heavily processed before we can start looking for correlations,
we have to perform Spike
Sorting.
Once this is done spike train analysis can be carried out. Our initial
biologically motivated project led us to develop several software tools
that can be found in the box on the left side of this page. Our work on calcium imaging data analysis is still much "in progress". You can get get an idea of what we do by checking these unfinished notes which can be viewed as an R tutorial showing how to reproduce the modeling / theoretical part of Tank et al (1995) and of Neher and Augustine (1992). The R script file required to reproduce the analysis of this web page is available. You should not use the material presented to analyze directly actual data. A proper noise model as well as a way to handle calibrated parameters are a prerequisite. We sketched how to do that on a poster presented at the last (November 2008) SFN meeting. A paper describing this work will be submitted in a couple of months at which time both manuscript and codes will be available from this site.
Collaborators:
Former Lab Members:
|
