ICube seminar - Some analysis and modeling problems in Neurosciences

Past event
Scientifique

A conference by Dr. Christophe Pouzat, CNRS Researcher of the MAP5 the "Applied Maths Laboratory" of the Paris-Descartes University

26 November 2019
10h30
ROOM 211 - Forum of the Faculty of Medicine, Strasbourg

In this seminar, four issues will be discussed

  • spike sorting,
  • spike train analysis,
  • neuronal networks modeling and simulations—for models on which mathematicians can work,
  • the analysis of fluorescence data used to measure free calcium concentration

The first issue recently came back in the spotlight with the development of MEA (Multi-Electrode Arrays). It requires "classical" clustering techniques together with some additional ones for dealing with superposition—co-occurrences of action potentials (spikes) from several neurons in a short time interval.

The second issue follows once a "satisfying" answer has been found to the first. It requires what probabilists call point processes or counting processes, but still stumbles over a pervasive feature of (extracellular) neuronal data: only a tiny fraction of the neurons of a given network is actually observed. This is what motivates the third issue: develop models allowing to say something about the whole network from the observation of a sample of its members.

The fourth issue could be seen as a "trivial" non-linear regression problem, but it clearly shows the improvements in model parameters estimation that can be gained from dealing properly with the measuring apparatus properties—the data recorded by a CCD chip follow a Poisson distribution.

Finally, "reproducible research"—more precisely, "reproducible data analysis"—is a methodological theme common to all these issues that will be present in the background of the discussion.

References

Invited by

Dr. Christophe Pouzat is invited by Laura-Adela Harsan, IMIS Team, ICube.

Read last issue