|NRL / Executive Directorate / Nanoscience / Programs / Nanoelectronics / Neural Electronic Interfaced / Neural Electronic Interfaced - Movies||NRL Resources|
Neural Electronic Interfaces
Overview | Movies | Posters | Publications
The movies below demonstrate the resolution of the stimulation chip. The images are a measure of the voltage at each unit cell of the 80x40 array. Please read the note at the bottom of this page for details on how the movies are obtained and ajusted.
Click on a Movie Below to Play
The retinal prosthesis test device is currently being electrically characterized at NRL. Each unit cell (or pixel) contains an amplifier/multiplier that modulates the degree to which adjacent neural tissue is stimulated, depending on the pixel value for that frame. The pixel value is updated every 60th of a second. An important element of the device is the 2-D microelectronic multiplexer, similar to those in conventional imaging arrays. All multiplexers have a common limitation, namely noise which can be generally characterized as two types, temporal and spatial. Temporal noise is the fluctuation in each pixel on a frame-to-frame basis, that is, given a uniform image that doesn't change, monitoring any pixel over time, noise fluctuations will be seen. Spatial noise is the fluctuation in response from pixel-to-pixel over an entire image frame. This is because the microelectronic fabrication of the transistors in each pixel is less than perfect and a nonuniform response will appear as fixed pattern or spatial noise.
In the retinal prosthesis device the temporal noise is considerably lower than the spatial noise. However, an important factor to consider is that fixed pattern noise is stationary, that is, it effectively stays the same given an additive or multiplicative factor and can be removed by the downstream neural processing that naturally adapts for nonuniformities. Or if need be, we can calibrate the response of each unit cell and perform nonuniformity correction to each pixel value before the image is input the device. We have previously worked on adaptive nonuniformity correction techniques and the existing algorithms are highly effective.
The examples shown are those images read-off the retinal prosthesis device while it was in operation and nonuniformity correction was performed as a post-processing step. In the planned experiments with human subjects, the nonuniformity correction will be performed as a pre-processing step before the video is piped into the device.
It is important to image a uniform, flat background prior to normal imaging to obtain the best nonuniformity correction results. The last movie shows the effect first imaging a uniform, flat background. This movie has less spatial noise. The camera was tilted on the side to capture this movie.