MOTIVATIONS

The functions of photoreceptors, horizontal cells and bipolar cells are desired as a sub-retinal prosthesis. The method supply power to the implant chip is critical in the procedure of implantation. The power dissipation is necessary to avoid huge heat in the eyeball.

Fig. 1.1 The Human Eye. Reprinted from “Foundations of Physiological Psychology,” by Neil R. Carlson, 1988, p.134.

Fig. 1.2 A Conceptual Epi-Retinal Prosthetic System. Reprinted form “A Neuro-Stimulus Chip with Telemetry Unit for Retinal Prosthetic Device,” by Wentai Liu, and etc, 2000.

Fig. 1.3. The Typically Functional Implantable Epi-Retinal Microsystem. Reptinted from “A Neuro-Stimulus Chip with Telemetry Unit for Retinal Prosthesis Device” by Wentai Liu, and etc., 2000.

Fig. 1.4. The Sub-Retinal Implant. Reprinted from the internet:

“http://www.optobionics.com/theeye.htm” by Mike Zang.

CHAPTER 2

THE ARCHITECTURE OF VERTERATE RETINA

2.1 BACKGROUND

Vision contains abundant information. Therefore, animals devote a large fraction of their nervous system to vision. On the human cerebral cortex, about 55% of the nervous system is contributed to the computations underlying the perception of vision.

Such a large computational system makes human easily perceive the surrounding world and control their actions. Vision is, therefore, one of the most important perception that human has.

The retina is a complex structure, containing a variety of cell types with complex interconnections. According to the researches on biological retina, all vertebrate retinas are organized in the same basic plane where two synaptic layers called outer and inner plexiform layers are intercalated between three cellular layers called photoreceptor unclear layer, bipolar unclear layer, and ganglion cell layer. Fig. 2.1 shows the schematic diagram of the retinal circuitry of the vertebrate retinas. There are five basic classes of retinal neurons in the three cellular layers as shown in Fig. 2.1, namely, the photoreceptor, horizontal, bipolar, amacrine, and ganglion cells.

Light is transduced into electrical potential by the photoreceptors at the top. The primary signal pathway proceeds from the photoreceptors in the photoreceptor unclear layer through the synapse in the outer plexiform layer to the bipolar cells, and then to the retinal ganglion cells, the output cells of the retina. The horizontal cells are located just below the photoreceptors whereas the amacrine cells just below the ganglion cells.

Both horizontal cells and amacrine cells spread across a large area of the retina to form layers transverse to the primary signal flow. The signals carried by the neurons in the retina can be divided into two types called graded potentials and action potentials. Graded potentials are observed in photoreceptor, horizontal, and bipolar cells whereas action potentials in amacrine and ganglion cells. Fig. 2.2 shows the intracellular response of typical recordings from a number of different cell types in the mudpuppy retina [56].

The properties of the five basic cells are described in the following subsections.

2.2 THE PHOTORECEPTOR CELLS

When light passes through the cornea, the pupil, the lens, the vitreous humor, and the transparent retina, it is finally captured by the most distal retina cells of the

photoreceptors, the input cells of the retina. The photoreceptor cells respond to the captured light only with the graded hyperpolarizing potentials as shown in the first row of Fig. 2.3. The potentials are used to activate the following neurons in the retina.

All vertebrates so far examined appear to have at least two types of photoreceptors, which often can be classified as rod cells and cone cells. [57]. The human retina contains approximately 120 million rods and 6 million cones. Rods are more sensitive and used in moonlight. Cones are less sensitive and operated over a broad range of light intensities. Cones also convey color information.

2.3 THE HORIZONTAL CELLS

The responses of the horizontal cells are activated by the outputs of the photoreceptor-horizontal synapse in the outer plexiform layer. The horizontal cells respond to the inputs from the photoreceptor cells only with the graded hyperpolarizing potentials. There are large gap junctions existing among the horizontal cells. So when the current form the photoreceptor is injected through the synapse into horizontal cell, the injected current flows into the neighboring cells horizontally. This electrical coupling enables the horizontal cells to have a very large receptive field which is defined as the portion of the visual field to which a single neuron responds. The intracellular response of the horizontal cell from the mudpuppy retina are shown in the second row of Fig. 2.3.

At the layer of the horizontal cells, two functions are performed. Firstly, the clear image signals passed form the photoreceptors are slowed in times so that the outline blurs and rapid movements are transferred imperfectly. This is equivalent to transient low-pass filtering. Secondly, the spatial low-pass filtering takes place because the incoming excitation from each photoreceptor is spread into the adjacent region over this horizontal layer. Consequently, what the low-pass filtering of the horizontal layer does is to transform the photoreceptor output response into a function of average illumination in both time and space.

2.4 THE BIPOLAR CELLS

The responses of the bipolar cells are initiated by the outputs of both photoreceptor cells and horizontal cells through the synapses in the outer plexiform layer. The bipolar cells respond to the inputs from both photoreceptor cells and horizontal cells not only with graded hyperpolarizing potential but also with graded

depolarizing potentials.

According to the responses to the retinal illumination, there are two types of bipolar cells observed. The first type of cell is that hyperpolarizes in response to the illumination of the center of the cell’s receptive field. The second type of cell is that depolarizes in response to the center illumination. The former cells are called the hyperpolarizing or off-center bipolar cells whereas the later are called the depolarizing or on-center bipolar cells.

In the off-center bipolar cells, the photoreceptor-bipolar synapses are excitatory synapse whereas the horizontal-bipolar synapses are inhibitory synapse. Therefore, the off-center bipolar cells are hyperpolarized strongly in a graded and sustained fashion by the outputs of the photoreceptors cells whereas these bipolar cell potentials are antagonized by the outputs of the horizontal cells. Because the horizontal cells have a large receptive field and lateral extent in the outer plexiform layer than do the bipolar cells, a center-surround receptive organization is observed in the bipolar-cell response which is a center-surround receptive organization is observed in the bipolar-cell response which is a center-hyperpolarizing surround-depolarizing response. The intracellular responses of this type of cells from the mudpuppy retina are shown in the third row of Fig. 2.3. In the on-center bipolar cells, the photoreceptor-bipolar synapses are inhibitory synapses whereas the horizontal-bipolar synapses are excitatory synapses. The response of such bipolar cells is a center-depolarizing surround-hyperpolarizing response.

The off-center and on-center bipolar cells come in pairs in the retina. Therefore they can offer and add benefit: activity in paired populations falls on the opposite sides of a light-dark boundary, and thus signals a line edge. A line edge is nothing more than a boundary between light and dark.

2.5 THE AMACRINE CELLS

In most retinas, two basic types of amacrine cell responses, transient and sustained, have been observed. More is known about transient amacrine cells than the sustained ones. The transient amacrine cells usually give on- and off-responses to the illumination presented anywhere in their receptive fields and are also very responsive to moving stimuli. They always respond by depolarizing and are the first neurons along the visual pathway to respond primarily with transient and depolarizing potentials. The transient amacrine cells generally do no show a center-surround antagonistic receptive field organization. Action potentials are often seen superimposed o their depolarizing on- and off- responses. The fourth row of Fig. 2.3.

shows the intracellular responses of the transient amacrine cell.

2.6 THE GANGLION CELLS

Most reports describe two basic kinds of ganglion cells which are the output cells of the retina: those that give sustained on-center or off-center responses and show an antagonistic surround and those that respond to flashes of light with transient on-off responses. The first type of cell is called the on-center ganglion cell or the off-center ganglion cell. The second type of cell is called the on-off ganglion cell. The responses of both two types of cells in the mudpuppy retina are shown in the last two rows of Fig. 2.3 where the responses of the sustained type are from an on-center ganglion cell.

As shown in the fourth row of Fig. 2.3, the on-off ganglion cells give transient responses at the onset and cessation of stimulation, much as the transient amacrine cells do. These recordings show that the on-off ganglion cells would receive much, if not all, of their synaptic inputs from the transient amacrine cells. In addition to the transient responses at the onset and cessation of stimulation, the on-off ganglion cells respond very well to motion. Moreover, they may show direction-sensitive responses.

According to the results of several experiments [58]-[59], these direction-sensitive responses are suggested to be mediated by two kinds of transient amacrine cells called excitatory and inhibitory amacrine cells.

Fig. 2.4 shows two schemes suggesting how direction-sensitive responses may be mediated by excitatory and inhibitory transient amacrine cells [60]-[61]. In the top scheme of Fig. 2.4, the inhibitory amacrine cell AI makes its synapses on the excitatory amacrine cell AE. In the lower scheme of Fig. 2.4, the inhibitory amacrine cell makes its junctions directly on the ganglion cell G dendrites. Movement of a spot of light in the preferred direction activates first the inhibitory amacrine cell, thus causes the inhibition of the excitatory amacrine cell (top scheme) or of the ganglion cell (bottom scehme) and the cancellation of excitatory input to the ganglion cell.

2.7 A SUMMARY OF FUNCTIONAL ORGANIZATION OF THE RETINA

Fig.2.5 summarizes much information and many ideas concerning the functional organization of the retina [62]. In summary, the outer plexiform layer of the retina responds mainly to both static and spatial aspects of illumination. The neurons in this layer consisting of the photoreceptors, the horizontal cells, and the bipolar cells, respond to stimuli primarily with sustained, graded potentials. The antagonistic center-surround organization at the level of the bipolar cells accentuates contrast in the retinal image. The on- and off-center ganglion cells, receiving much of their input directly from either center-depolarizing or center-hyperpolarizing bipolar cells, reflect

this basic center-surround receptive field organization established in the outer plexiform layer.

The inner plexiform layer, on the other hand, responds more to the dynamic or temporal aspects of photic stimuli. Both transient amacrine and on-off ganglion cells accentuate the changes in the retinal illumination and respond vigorously to moving stimuli. Interactions in the inner plexiform layer underlie the motion- and direction- sensitive responses of the on-off ganglion cells and the orientation-preferring response of some on- and off-center ganglion cells.

Fig. 2.1. The schematic diagram of the retinal circuitry of the vertebrate retinas.

Reprinted from “foundations of Physiological Psychology,” by Nwil R. Carlson, 1989, p. 136.

Fig. 2.2 Intracellular recording from the mudpuppy retina showing the difference in response of a given cell type to a 100-μm spot and to annuli of 0.5 and 1.0 mm in diameter, all at the same intensity. Reprinted from “ The Retina: An

Approachable Part of the Brain,” by J. E. Dowling, Cambridge, MA: Belknap Press of Harvard University Press, 1987, p. 84.

Fig. 2.3. Two schemes suggesting how direction-sensitive responses may be mediated by excitatory and inhibitory amacrine cells in the retina. In the top scheme, the inhibitory amacrine cell AI makes its synapses on the excitatory amacrine cell AE processes. In the lower scheme, the inhibitory amacrine cell makes its junctions directly on the ganglion cell G dendrites. (^○: excitatory synapses; ^●: inhibitory synapses; B: bipolar cell). Reprinted from “ The Retina: An

Approachable Part of the Brain,” by J. E. Dowling, Cambridge, MA: Belknap Press of Harvard University Press, 1987, p. 114.

Fig. 2.4. Summary scheme of the synaptic interactions that occur in the retina and that underlie the receptive field properties of the on-center, off-center, and on-off ganglion cells. (^○: excitatory synapses; ^●: inhibitory synapses; ^△: reciprocal synapses). Reprinted from “ The Retina: An Approachable Part of the Brain,”

by J. E. Dowling, Cambridge, MA: Belknap Press of Harvard University Press, 1987, p. 118.

CHAPTER 3

THE DESIGN, CHARACTERISTICS AND ANALYSIS OF ARTIFICIAL RETINA

3.1 THE CHARACTERISTICS OF PN JUNCTION DIODE AS ARTIFICIAL RETINA

In 1990, Dr. Gene de Juan and Dr. Mark Humayun first demonstrated that retinal ganglion cells could be electrically stimulated without penetrating the retinal surface [63]. Afterwards, bypassing the damaged retina with an artificial retina is the goal of many researchers around the world.

In certain diseases of the retina, the photoreceptor cells do not respond to light and thus causes blindness. If only the photoreceptor, light sensing part of retina, is damaged, the vision can be restored by replacing the photoreceptor only. The photoreceptor converts the light into electrical signal as the pn junction diode does.

The absorption of light inside the diode creates electron-hole pairs, as pictured in Fig.

3.1 [64], where LP, LN and W are referred to as the minority carrier holes in an n-type material, the minority carrier electrons in a p-type material and depletion width of pn diode respectively.

Minority carriers, holes in n-diffusion and electrons in p-diffusion, and carriers photogenerated within the depletion region are swept by the electric field to the opposite side of the junction, thereby contributing an added reverse-going component to the current through the diode. If the photogeneration rate (GL) is assumed to be uniform through the diode, the added component due to light (IL) should be equal to –g times the electron-hole pairs photogenerated per second in the volume A(LN+W+LP), or depletion region is typically small compared to LN+LP. If W is negligible, IL becomes

independent of the applied bias. The light-on I-V characteristics are therefore expected to be essentially identical to the dark I-V characteristics, except the light-on curves are translated downward, move in the –I direction, along the current axis.

Moreover, because IL is proportional to GL, the downward translation of the characteristics should increase in proportion to the intensity of the incident illumination. The described form of the photodiode I-V characteristics is illustrated in Fig. 3.2.

Due to the function of pn junction diode, transforming light into electrical signal, the pn junction diode in solar cell mode can be used as subretinal synthesis to replace photoreceptor in retina. There is, however, linearity problem, that is, the pn junction diode in solar cell mode cannot differentiate the light intensity in a wide range. Fig.

3.3 shows

According to the characteristics of pn junction diode, there are two different usages while applying pn junction diode to artificial retina. One is giving no bias, and the pn junction diode works as solar cell. The other is applying positive bias, and the pn junction diode function as CMOS image sensor. To differentiate the performance of theses two cases, two kinds of chips are designed and analyzed.

Firstly, a p+-n-well diode array is designed and fabricated. Because it is not clearly understood which kind of signal is transferred in mammalian retina, four different diode arrays are designed which are pictured in Fig. 3.3. Fig. 3.3 (a) shows a diode with n-type node floating and p-type node connected to an electrode to stimulate the cell while Fig. 3.3 (b) shows a diode with p-type node floating and p-type node connected to an electrode. When the light irradiates on the diode, the voltage across the two node of the diode would changes and thus a voltage signal stimulates the cell. Fig. 3.3 (c) and (d) picture a diode with one node connecting to ground and the other connect to an electrode to stimulate the cell. Because a closed loop is formed, the diode would send current of different magnitudes according to the light intensity.

Secondly, the other utilization of p+-n-well diode is pictured in Fig. 3.4.One pixel contains a p+-n-well diode and a current mirror with current gain of 8 formed by PMOS. There are two electrodes in each pixel. The electrode connected to the pn diode provides the diode a close loop so that a photocurrent can be generated. The output current from the current mirror pass through the other electrode to stimulus the retinal cells. The current mirror serves as two functions. One is to bias the pn junction diode to ensure that the diode work at the third quadrant just as the CMOS sensor does. The other function is to amplify the photocurrent generated by the pn junction diode so that the area of pn junction diode can be made smaller. When the pn junction diode senses the light, a photocurrent is generated and amplified by a current mirror.

The pn junction diode operates in the third quadrant without doubt because n-type node connected to higher potential (closer to power supply). The amplified photocurrent stimulates the retinal cell through the electrode.

3.2 ANALYSIS OF THE PSEUDO-BJT-BASED SILICON RETINA

Although some patients can restore their vision by replacing photoreceptors, other patient may need replacing more retina cells, including bipolar cells and horizontal cells. Recall that the bipolar cell processes image edge detection function, and the horizontal cell process smoothing function.

The proposed silicon retina is based on Pseudo-BJT (PBJT) [65]. The PBJT is composed of two MOSFETs. Fig. 3.5 (a) shows the pseudo npn BJT. It possess two n-channel enhanced mode MOSFETs with the source and gate of MOSFEET M1 and the gate of MOSFET M2 connected at the same node, B. This structure is the same with the popular current-mirror arrangement. Therefore, the current gain approximately equals to the W/L ratio between the transistor M1 and M2. The input node B of the current mirror acts as the base of the PBJT. In the same way, the pseudo pnp BJT can be derived by two enhanced mode p-channel MOSFETs via similar arrangement as shown in Fig. 3.5 (b).

The photocurrent generated by pn junction diode is usually small, and therefore the PBJT work in the subthreshold region. The drain current in the subthreshold region can be expressed as [66]

where Vgs is the gate-to-source potential, Vds is the drain-to-source potential, VT=kT/q is 26mV, Kx depends on process parameter, W/L is the geometric ratio of the MOS,η is the subthreshold swing parameter [66]. Based on Eq. (3.3), the current gain of

Examining Eq. (3.4), the current gain beta is larger than desired value in low-induced current level due to the difference between Vds of master and slave MOSFETs. When the induced current becomes larger, the Vds of master and slave

MOSFETs are close to each other. Thus the current gain beta approaches to the desired value. In order to reduce the affect of subthreshold swing parameter, the slave

MOSFETs are close to each other. Thus the current gain beta approaches to the desired value. In order to reduce the affect of subthreshold swing parameter, the slave