Bruno Cessac - Inria Biovision team - Keywords: retina, visual system, neuronal networks

Bruno Cessac - Inria Biovision team - Keywords: retina, visual system, neuronal networks

Contribution title: Multi scale modeling of the retina

The retina, located at the back of the eye, converts the light coming from a visual scene into sequences of impulses (spikes) conveyed to the brain by the optic nerve, and interpreted by the visual cortex. This process involves multi scale biophysical dynamics: molecular level, cellular level (neurons), network of neurons. The Inria Biovision team is interested in the modeling of the retina in normal and pathological conditions (sight impairments) using methods from physics (non linear and statistical physics) and mathematics (dynamical systems and bifurcations theory, probabilities). In this talk I will give brief overview of our activity focusing on subjects having strong potential overlaps with other UCA teams.

* Mesoscopic modeling of the retina. Retina transformations occurring during development, under pathologies, or pharmacological manipulations, start at the molecular level, impacting eventually the whole retina structure and functioning. The global dynamics involves a tremendous number of variables and parameters, which makes it difficult to model and analyze. We develop models of this multi-scale evolution using the physicists approach. Our goal is to reduce the global «microscopic » dynamics to a reduced « mesoscopic » dynamics, where a few key variables and parameters resume the evolution, yet staying close to biophysics and experiments. I will briefly present a few examples retinal development or pharmacologically induced pathologies.

* Statistical analysis of retina responses. Recent technological advances allow to record simultaneously the spiking activity of thousands of neurons in the retina. This is a step toward understanding how the retina encodes a visual stimulus in a parallel stream of spikes. The statistical analysis of these data requires however elaborated methods. I will present examples of such methods constructed from non equilibrium statistical physics and information geometry with applications to real data analysis.