VISUAL
CORTEX STIMULATOR PROTOTYPE BASED
ON
MIXED-SIGNAL TECHNOLOGY DEVICES
J.-F.
Harvey, M. Roy, M. Sawan
Department
of Electrical and Computer Engineering
École Polytechnique de Montréal
ABSTRACT
This paper describes the design of a complete prototype for a miniaturized visual stimulator (MVS). The prototype is a research tool used to validate processing algorithms and stimulation strategies as well as give an insight as what blind people would see with the MVS. The prototype is realized using off the shelf mixed signals electronics components for low cost and short prototyping delays. The MVS consists of two main parts: the first is the implant which consists of a silicon die mounted on the back of an electrode array. The silicon die contains all the electronic circuits necessary to receive the command words, detect and correct transmission errors, decode the commands and control the stimulation process. The second part is the external controller that must acquire real life scenes, adapt the resolution if it is not compatible, enhance the image, extract the stimulation information, generate the command words and send them to the implant via a RF coupled technique along with the energy required to power up the receiver. The MVS prototype (MVSP) must integrate the majority of the MVS functions in order to validate the design but instead of stimulating cells, the implant prototype drive a light emitting diode (LED) matrix where each LED represent a phosphene.
INTRODUCTION
Since electrical stimulation techniques are applied in many circumstances to enhance or restore organs functions, few research teams are working on the recuperation of a limited but functional vision to totally blind persons. A functional vision means that the person will be able to do, without assistance, most of the tasks being part of the every day life. It will be limited since no system in the near future will be able to replace the natural vision system with the same accuracy. The required resolution and data processing capabilities are simply too large. To accomplish an adequate visual stimulation, two main steps are necessary. The first is to acquire a real life scene and convert it to stimulation information in the form of stimulation command words. The second is to apply the electrical stimulation according to the command words.
The history of human visual stimulation began in the 60’s when Foerster was the first to expose that when a specific part of the human brain was stimulated with an electrical current, a fixed light spot appeared in the visual field of the patient. This part of the brain was later identified as the visual cortex and the light spots are called phosphenes. In 1968, Brindley was the first researcher to publish medical experiments results related to visual stimulation. In these experiments, the phosphenes were generated with different voltage sources and electrode spacing through an array of 81 platinum electrodes. In all cases, the electrical stimulation was applied on the surface of the visual cortex. As research progressed, notable discoveries were made and can be summarized as follow: current based intracortical stimulation leads to a significant current reduction, more stable phosphenes and a phosphene intensity that is proportional to the current. Other approaches for visual recuperation includes retina stimulation, proposed by Rouch and optic nerve stimulation. Next section of this paper describes the MVS prototype. It is followed with preliminary results and the conclusion. The purpose of the prototype is not stimulate cells but to give an insight of what blind people would see with the MVS. This is accomplished by a LED matrix where a LED simulate a phosphene in the visual field.
DESCRIPTION OF THE PROTOTYPE
The MVS consists of two main parts. The
first is the implant which receives the stimulation command words and stimulate
according to the parameters. The second part of this system is the external
controller that acquires the images and generate the command words. Figure 1
shows an overview of the system. The MVS prototype must integrate the majority
of the MVS functions in order to validate the stimulation strategies. The MVSP has the same architecture and
provides the same functionality as the projected MVS. The differences between
them is that the prototype is made from discrete components instead of an
integrated circuit and the stimulation currents are sent to a LED matrix
instead of an electrode array. The proposed MVSP also consists of two distinct
systems: the implant prototype and the external controller prototype. Figure 1: Overview of the MVS
Figure 2 shows the block diagram of
the MVS implant prototype. The serial data stream coming from the external
controller is received by the Manchester decoder. The outputs of the decoder
which are the system clock and input data are then applied to the main
controller (FPGA). The controller is responsible to detect frames of data,
called command word (CW).
Figure 2: Overview of the MVS implant prototype.
These command words contains instructions to be executed by the implant. There are two kinds of CW, configuration command words (CCW) and stimulation command words (SCW). The CCW are used at the beginning of a stimulation period usually to specify to the implant the format of the stimulation command words. Once configured, the implant prototype receives SCW which contain stimulation parameters for one or several stimulation sites (LED). Using this information, the controller activates the 16 channels DAC and 16 demultiplexers to activate one or several sites simultaneously. The implant prototype contains 16 channels where each one of them controls the activation and deactivation of 16 LEDs.
Each channel can activate only one site at a time. It means that only 16 LEDs can be on simultaneously (16 channels) for the overall 16x16 LED matrix. Each channel contains a channel controller (inside the FPGA), a demultiplexer to select one LED over 16 and finally a voltage to current (V/I) converter to control the current amplitude in the LED. To activate a particular site, the channel controller starts by sending a request to the DAC controller (inside the FPGA) in order to load the amplitude into the DAC. Then, the channel controller sends the address of the site to stimulate and the ctrl signal to the channel demultiplexer turning on the LED. The 16 LEDs of each channel are arranged in a common cathode configuration. The cathodes are connected to the output of the voltage controlled current source (VCI). Each output of the demultiplexer is connected to the anode of a single LED. When a particular demultiplexer output is activated, it switches from 0 to 5V. This voltage enable a current to pass in the LED. The amplitude of the current is then controlled by the VCI.
In order to transform real life scenes into stimulation information and send it to the implant, the external controller, based around a PC platform computer and an USB (Universal Serial Bus) camera, acquires successive images and convert them into electrical signals. Each image is stored to enable digital processing. The digital information is then sent to a processing unit where the resolution is adapted and the image is enhanced. The remaining steps are the formation of the command words, serial encoding and the transmission to the implant itself. It is also possible to use a pattern generation interface where the command words are formed from internal patterns instead of the image sensor. This option allows to quickly test recognizable patterns like a square, a circle or a cross.
RESULTS
The flexibility of the external controller of the proposed prototype has been demonstrated. It is even more important than the one of the implant since the external controller will have to be able to work with different generations of the MVS implant. Each of them will have its own resolution and serial encoding sequence. The majority of the image processing steps are software based which ensure a maximum of flexibility. Since the external controller is used for MVSP testing and for the upcoming in vivo experiments, it is able to adapt quickly to different generations of the MVS implant prototype. The PC platform was selected for its low cost and high processing power. Figure 3 shows a screen capture of the external controller user interface. It depicts an example of a 176x132 pixels image source (the resolution of the CCD camera) and its corresponding image reduced to 25x25 enhanced image, which is the size of the upcoming MVS implant stimulation matrix. The reduced image resolution can be changed instantly with in the menu list. The lower left part of the dialog specifies the pattern mode with the square, cross, circle and character pattern available.
Figure 3: Screen capture of the external controller software.
The design of the implant prototype is completed and the layout of the printed circuit boards (PCBs) is almost finished. The simulations of the digital part located in a FPGA has been successful. Validation of the analog circuits has been successfully realized on a breadboard with the same discrete components that will be used on the PCBs.
CONCLUSION
The design of a complete prototype of a system has been realized using off the shelf mixed signals electronic components. The prototype allows to validate processing algorithms and stimulation strategies. Its flexibility allows the prototype to become a powerful research tool in cortical visual stimulation.
ACKNOWLEDGEMENT
The authors would like to acknowledge financial support from the Natural Sciences and Engineering Research Council of Canada (NSERC).
REFERENCES
[1] G.S. Brindley, W.S. Lewin,
“The sensations produced by electrical stimulation of the visual cortex”,
Journal of physiology, vol. 196, 1968, pp 479-493.
[2] F.T. Hambrecht, M. Bak, C.V. Kufta, D.K. O'Rourke, E.M. Schmidt, P. Vallabhanath,
“Feasibility of a visual prothesis for the blind
utilizing intracortical microstimulation”,
4th Vienna Int. workshop of functional electrostimulation,
1992, pp 1-8.
[3] A. Boyer, ²Développement d’un
stimulateur miniaturisé dédié à un implant visuel,² mémoire de maîtrise, école
polytechnique, 170 pages, 1995.
[4] K.E.
Jones, R.A. Normann, ²An Advanced
Demultiplexing System for Physiological Stimulation², IEEE Trans. Biomed. Eng.,
vol 44, no 12, December 1997.
[5] J.-F. Harvey, M. Sawan, ²Image
acquisition and reduction dedicated to a visual implant², IEEE-EMBS, November 1996.