pOPulation coding of moTIon in norMal and pAthological retinas – OPTIMA

Thanks to a new technique designed by the candidate to record a large population of ganglion cells, we will estimate the ability of the different cell types to code for motion, in normal, then degenerated and restored retinas. Collaborating with the NeuroMathComp project team (INRIA), we will quantify the interaction between the outputs of each type of ganglion cell (statistical analysis), and their ability to code motion alone or together (decoding of the motion trajectory from the spike trains). This analysis will be repeated on a rat model of photoreceptor degeneracy, after its optogenetic reactivation by expressing halorhodopsin in the degenerated photoreceptors.

We have shown that it is possible to reconstruct precisely the trajectory of an object moving randomly in the visual field only from the activity of the population of rétinal cells we have recorded. We are thus able to read and understand the « language » of the retina, and to decode the the visual information it contains with a great precision. We have shown that this is only made possible by the large number of neurons recorded simultaneously.
We will now use the tools designed and the data we have to understand how the different subtypes contribute to the transmission of visual information, by decoding each subtype separately.

We also want to determine how the reorganisation of the retinal network following photoprecetor degeneracy affects visual information processing. The same analysis of population coding will be repeated on a rat model of photoreceptor degeneracy, after its optogenetic reactivation by expressing halorhodopsin in the degenerated photoreceptors. This will provide, for the first time, a benchmark to assess the retinal capacity to process motion after a rehabilitation strategy, and thus quantify its efficiency. From a more fundamental point of view, we will determine how the different cell types construct together a representation of the moving stimulus.

Results have been presented to two meetings (Cosyne 2012, Areadne 2012), as well as during seminars outside the Institute (Ecole Normale Superieure, Paris, March 2012 ; FMI, Basel planned in July 2012).
We are preparing a paper about the results obtained in the decoding part of the project. A paper about the methods used for the statistical modeling part has been submitted.