The retina has many parallel information processing channels. As a result, the ganglion cells which pass information to the brain consist of dozens of different types [c.f. Baden et al. 2016 Nature]. All the ganglion cells of a single type constitute a visual information channel; and each of these channels carries a different message about the visual world to the brain (see Figure). Ensuring that bionic retinal implants fully exploit this diversity of information channels has been a long-term goal of bionic vision research.

In this project, retinal ganglion cell (RGC) coding diversity is quantified by recording simultaneously from large populations of RGCs in the mouse retina using a microelectrode array (MEA). The spiking responses to a standardized set of visual stimuli are parameterized and projected into a high-dimensional space to be clustered according to RGC type. Based on this population clustering, a visual characterization toolbox is being developed to categorize individual RGCs. This toolbox will be the enabling tool for testing RGC type-specific activation through the presentation of specially designed electrical stimulation patterns.

Results

• Preliminary experiments have been conducted and we are currently comparing these extracellular spike train response to the calcium signal-based RGC categories published in [Baden et al. 2016 Nature