 |
15 Sep 2006
An artificial silicon-based retina could offer a unique insight into how the mammalian eye works.
Researchers in the US have simulated the retinal circuitry of the eye by building a functional silicon-based version exploiting MOS (metal oxide semiconductor) transistors. According to the team, the device could be a forerunner for the intraocular prosthetic devices of the future (Journal of Neural Engineering 3 257).
"Our device recreates the functional neuroanatomy of the mammalian retina and its visual processing circuitry at a similar physical scale and energy efficiency," Kareem Zaghloul, lead author of the paper, told optics.org.
"A device similar to our silicon retina could be used one day as a visual prosthetic device. In the meantime, it serves as a real-time hardware model for the mammalian retina, allowing us to explore some of the unanswered questions in the physiology of the eye."
The University of Pennsylvania team, which also included Kwabena Boahen, says that the artificial retina could also be connected to other devices that simulate human systems further along the neural line. Such devices -- which are a common requirement in artificial intelligence and robotics -- often need visual processing inputs.
To build the device, the researchers morphed the "well-understood" circuits of the retina by substituting each synapse (neural junction) in the physiological model with a transistor. "The physical scale has to be small enough to achieve high-resolution stimulation of individual neurons, and the power usage has to be low enough to make it compatible with organic tissue," explained Zaghloul.
Using arrays of phototransistors, the team produced circuits for both the inner and outer retina. The output from these devices was then fed into circuits simulating the ganglion cells -- which, in nature, receive the signal from retina's photodetecting rods and cones to generate spike voltage signals. These voltage signals are then sent down the ganglion cells' axons into the optic nerve.
"Our device takes the visual world as its input, and outputs a spike-based neural code representative of this visual scene, as seen through an array of 3600 ganglion cells," said Zaghloul. The artificial device was also designed to mimic the mammalian retina's ability to adapt its output in response to changes in brightness and contrast.
The researchers acknowledge that a lot of scope remains for improvement. Some discrepancies exist between the physiological model and the device, while some degree of manual intervention was required to ensure that the device's external biases were fixed at specific voltages.
However, the researchers claim that the silicon retina operated independently "to a large extent", and believe that a similar approach could be taken to model other neural systems with MOS transistors.
 |  |  |  |  |  |  |
| © 2024 SPIE Europe |
|
