Retinal Implant

Our lab is part of the Biomimetic MicroElectronic Systems Engineering Research Center (http://bmes-erc.usc.edu). One of the center goals is to develop a retinal prosthesis for blind patients. Our lab is working to understand the retinal ganglion cells code of natural images and to develop techniques to mimic this code by electrical-current injection. Therefore, our long-term goal is to develop a signal processing algorithm that will convert the natural light stimuli a healthy patient sees into the equivalent electrical stimuli a blind patient equipped with a retinal implant would need, to acquire the same percept. To develop this algorithm, we have been building mathematical models to account for the responses of visual cells to natural image (http://vpl.usc.edu/research/circuits.html).
In parallel, we are developing a device to record & stimulate simultaneously the retinal ganglion cells. The stimulator will be unique in that it can output arbitrary waveform shapes and control multiple electrodes independently. The stimulator will stimulate retinal tissue in-vitro to determine the optimal waveforms to excite retinal ganglion cell sub-populations. A preliminary computational study of those optimal waveforms have revealed that current time courses that minimize power consumption and charge delivery are different from those currently used in prosthetic devices.

Nadav1

Figure 1: Waveforms for extracellular model for power (solid line) and charge (deshed line) optimization for 19 time durations.


The next step in building our biomimetic Retinal Spike Generator algorithm will be to learn how to replicate the natural retinal responses with electrical stimuli. The main difficulty for such algorithm is that a current may activate presynaptic inteneurons associated with the retinal ganglion cells. Hence, we propose to find the parameters of a Volterra type model that will replicate the natural responses.The starting point will be the Volterra model developed with visual stimulation. We’ll then collect the responses of ganglion cells to the electrical stimulation, compare this responses to the desired Volterra model, optimizing it through a learning algorithm.

Nadav2

Figure 2: Flow diagram illustrating the operation of the retinal spike generator.


In addition we are looking into the morphological changes induced in the retina by electrical stimulation and electrode implantation. This is done through histology of dye stained retinal tissue (Brain Res. 2009 1255:89-97)