32nd Annual International Conference of the IEEE EMBS Buenos Aires, Argentina, August 31 - September 4, 2010

A Continuum Model of the Retinal Network and its Response to Electrical Stimulation
Shijie Yin, Nigel H. Lovell, Senior Member, IEEE, Gregg J. Suaning, Member, IEEE, Socrates Dokos, Member, IEEE
AbstractA continuum network model of the retina is presented, consisting of an active implementation of the retinal ganglion cell tissue layer and passive implementation of deeper cell layers. The retinal ganglion cell layer receives excitatory presynaptic inputs from the bipolar layer and inhibitory presynaptic inputs from the amacrine layer. Simulations were performed to investigate the behavior of retinal tissue activation with epiretinal and suprachoroidal electrode stimulation. The results indicated the presence of both early and late onset action potentials consistent with experimental findings.

I. INTRODUCTION

etinitis pigmentosa and age-related macular degeneration are two of the most prevalent retinal diseases and leading causes of blindness in developed countries [1]. Both of these diseases result in degeneration of the retina, predominantly the outer retina where the photoreceptors are located, leading to eventual loss of vision. The inner retina however is largely left intact, even in patients who have been clinically blind for many years [2, 3], raising the possibility that these inner retinal neurons can be electrically stimulated by nearby electrodes to elicit light perception [4, 5]. The development of an intraocular vision prosthesis is currently being pursued by several research groups around the world [6, 7, 8]. A particular focus has been on eliciting meaningful visual percepts through extracellular stimulation. Electrode placement strategies have been explored on the epiretinal surface, sub retinal space and suprachoroidal space. Previous modeling studies on the effects of electrical stimulation have been limited in their scope [9, 10]. These isolated studies have only taken into account the behavior of the retinal ganglion cells (RGCs) in response to electrical stimulation and have largely ignored the presence of the rest of the retinal network and passive propagation through retinal layers. We believe these network effects play a significant role in shaping the spiking activity of the RGC layer and therefore the spatial and temporal responses to excitation at the cortical level. It has been shown that excitation of the cells in the inner nuclear layer, including the bipolar and amacrine cells, produce excitatory and inhibitory inputs respectively to the ganglion cells [11]. The presence of these so-called late onset or secondary spikes have been

observed in a variety of experimental studies [12, 13, 14]. Other groups have attempted to incorporate these presynaptic inputs by modeling the retina as a discrete network [15, 16]. These models however ignore the active membrane properties of the RGCs. We have developed a continuum model of the retinal network which includes both passive retinal properties and active ganglion cell behavior. Synaptic inputs from bipolar and amacrine cells are also incorporated in shaping the spiking activity at the ganglion level. A generalized retinal network model of the ON system is presented where parameters have been compiled from experimental data from various studies. Results are shown for both epiretinal and suprachoroidal electrode placement. II. METHODS A. Model Formulation To investigate patterns of retinal tissue activation due to electrical stimulation, we have developed a 3D finite-element model comprised of vitreous fluid and the neural retina (Fig. 1). The retinal section consists of an active domain of RGCs, and two passive domains including that of the inner nuclear layer (INL), (bipolar, horizontal, wide-field and narrow-field amacrine cells) and sub retinal space (modeled with cone photoreceptors). Additional conductive tissue layers in the model include the inner plexiform layer (IPL), outer plexiform layer (OPL), outer nuclear layer (ONL), retinal pigment epithelium (RPE) and choroid. A single pair of circular disc electrodes is used to apply electrical stimulation in this study; one active and one return. In all simulations the return electrode is connected to ground (zero volts). The electrodes have a radius of 0.2 mm and are placed in the vitreous fluid, 200 m above the upper surface of the retina for epiretinal stimulation and placed 426 m below the retinal pigment epithelium for suprachoroidal stimulation (based on choroid thickness from in vivo measurements of the human retina [17]). The extracellular voltage distribution Ve within the vitreous region was governed by Poisson's equation (1) ( b V e ) = I where

b is

the conductivity of the bulk vitreous medium

and I is the volume current density source (A/m3) injected into the vitreous at a given location. The value of b was taken to be 1.25 S/m [18]. Within the active retinal ganglion cell region, an adaptation of the bidomain equations was applied (Fig. 2 left panel).

S. Yin, S. Dokos, N. H. Lovell and G. J. Suaning are with the Graduate School of Biomedical Engineering, University of New South Wales, Sydney, NSW 2052, Australia (email: n.lovell@ unsw.edu.au).

978-1-4244-4124-2/10/$25.00 2010 Crown

2077

where the extracellular retinal conductivity

e of

the

remaining layers are derived from [18, 19], and Im is the RGC membrane current density per unit volume, given by V (3) I m = C m m + J ion + i sn = g r (Vr Vi ) t and Vm is the transmembrane potential given by (4) Vm = Vi Ve z y x Cm is the membrane capacitance and is 1 F/cm2 for all cell types, consistent with previous epiretinal stimulation models [20] and also cellular models [10, 21]. denotes the surface to volume ratio of the ganglion cell layer. We adopted a ganglion cell density of 2000 cells/mm2 [22] and determined the value as 9.25x10-4 m-1 by assuming a spherical ganglion cell with a soma diameter of 18 m. This also assumes that extracellular currents stimulate only the soma of the ganglion cell [15] as threshold for activation is lowest near the soma or axon hillock [9]. Jion is determined based on the ionic formulation of Fohlmeister and Miller [10]. The synaptic current, isn denotes input from deeper layers: excitatory input feeds in from the bipolar cells and inhibitory input arrives from wide-field amacrine cells. (5) isn = Psn g sn (Vm E sn ) where gsn is the conductance of the synapse, Esn is the reversal potential of the synaptically gated channel and Psn represents the delayed response from presynaptic potentials on the conductance of the synaptic channel. Psn is determined by the synaptic transfer function with a center operating point of V50 and steepness parameter 50. Synaptic transfer parameter values were derived from [15]

Fig. 1. Schematic diagram of the retinal model with two circular disc electrodes located epiretinally (solid line) and suprachoroidally (dashed line). Dimensions of the rectangular domain are 4 x 4 x 0.805 mm. Thickness of retinal layers are: vitreous (200 m), RGC layer (22 m), IPL (23 m), INL (27 m), OPL (16 m), ONL (31 m), sub retinal space (40 m), RPE (20 m) and choroid (426 m) [17, 18]. Other key dimensions include: electrode radius 0.2 mm; distance between electrode centers 1 mm. Electrodes are placed centrally perpendicular to Y and symmetrically located about the mid-plane perpendicular to X.

Ve

Ve

Cm

Jion

isn

Cm

Rm

Vi
gr Vr

Vi
gr Vr

dPsn p Psn (t ) = dt sn
where p is,

(6)

Fig. 2. RC circuit of the active RGC membrane potential (left), RC circuit of the passive membrane potential implementation (right). Cm represents the membrane capacitance, Jion is the ionic current per unit area through gated channels in the cells soma, isn represents the presynaptic current inputs from the retinal network and gr is the resistive tie connecting the intracellular potential Vi to a resting potential Vr, Rm is the specific membrane resistance.

p =

1 + exp(V pre V50 50 )

exp(V pre V50 50 )

(7)

The intracellular potential Vi, of each cell is resistively tied to a resting potential Vr, representing the intracellular potential in the more distal portions of the neuron (axon and dendrites). As a consequence, the intracellular potential is not able to float freely with the extracellular potential during extracellular stimulation. This model allows for local excitation and activation of individual ganglion cells without spread of activation to neighboring cells, important for eliciting focal excitation of phosphenes. The extracellular voltage is determined from (2) e 2Ve = I m

Vpre is the average presynaptic potential integrated over the area of the dendritic receptive field. We assumed a dendritic receptive field size of 100 m for the ON ganglion cell, this is similar to previous modeling studies of ON-alpha ganglion cell receptive fields [23]. We modeled the remainder of the retinal cells with a passive implementation, where the membrane current Im was expressed as, dVm Vm Vr (8) Im = C m dt + R = g r (Vr Vi ) m where Rm is the specific membrane resistance. For parameter consistency, we have adopted values from [15] for all relevant cell types.

2078

B. Numerical Implementation Simulations were conducted using COMSOL Multiphysics Version 3.5a (COMSOL AB, Sweden) on a Quad Core AMD Opteron Windows 64 server with 127 GB of RAM. Simulation time varied from 2 to 4 hours depending on server capacity and solver settings. III. RESULTS Simulation results are shown in Fig. 3 for epiretinal and suprachoroidal electrode placements. In both simulations a monophasic current pulse of 0.0159 mC/cm2 charge density was applied at t = 1 ms with a pulse width of 0.1 ms. RGC activation patterns are also presented 0.1 ms after stimulus onset. The output shown in Fig. 3 for both stimulus strategies was recorded 215 m below the vitreous surface at the center of the stimulating electrode. The epiretinal simulation elicited an early onset action potential (AP) as a result of extracellular stimulus as well as a late onset AP. The suprachoroidal simulation however displayed a subthreshold response due to the extracellular stimulus, but it induced two late onset APs. Initial inspection based on input currents from the bipolar and amacrine cell layers suggest that these late onset APs are of presynaptic origin. Removal of these currents eliminated the late onset spiking activity, this was true for both studies conducted (results not shown). The latency of the early and late onset AP (measured from

stimulus onset to peak of the AP) for the epiretinal case was 0.8 ms and 6.2 ms respectively. For the two suprachoroidal APs, these were 2.7 ms and 6.5 ms. IV. DISCUSSION The epiretinal case was observed to have a lower current threshold for AP generation compared to suprachoroidal, which displayed a subthreshold response. This is consistent with previous experimental studies. Current threshold densities for epiretinal stimulation have been reported to be 0.093 mC/cm2 in rabbits [13] and 0.073 mC/cm2 in rat RGCs [14]. For suprachoroidal stimulation a more disparate range of values were observed from 0.042 mC/cm2 to 0.375 mC/cm2 [7, 8, 24]. Variations between reported values are largely due to differences in the experimental protocol and threshold definitions. Our current density was observed to be lower than that reported in the above experimental studies. These threshold differences can be accounted for in part by our use of bipolar monophasic stimulation, as it has been suggested in prior studies that monophasic pulses have a lower threshold than biphasic [8, 25], and also the choice of parameter values we have adopted from various experimental studies. The occurrence of early onset or stimulus-locked APs as well as late onset APs has been observed in several experiments. Depending on the particular study, variations

Fig. 3. Simulation results 215 m below the vitreous surface in the RGC sub domain, stimulus applied at 1 ms (arrowhead) of 0.1 ms pulse width with a simulation time of 10 ms. A constant 0.2 mA monophasic epiretinal stimulus, left: transmembrane potential (mV), right: transmembrane potential (mV) at 0.1 ms after stimulus application, showing activation pattern of the RGC layer. B constant 0.2 mA monophasic suprachoroidal stimulus, left: transmembrane potential (mV), right: transmembrane potential (mV) at 0.1 ms after stimulus application, showing activation pattern of the RGC layer.

2079

exist in the classification of latency, in some instances as low as 0.7 ms [13] for epiretinal stimulation. This was also seen in [14] where the majority of short latency responses occurred within 1 ms. In the same study, secondary spikes were observed to occur with a latency of 5 ms. Interestingly this was also the case with sub retinal stimulation [12]. The latencies observed in these studies for both early and late onset spikes are consistent with our model behavior under similar pulse widths. For suprachoroidal studies, latency was measured to the first peak of the electrically evoked response (EEP) ranging from 9 ms to 25 ms [7, 8]. Since we did not simulate EEP output, direct latency comparisons could not be made. The activation pattern of the epiretinal simulation shows a much more focal response than that of suprachoroidal. We believe this is a consequence of current flowing through the highly resistive retinal pigment epithelium layer, resulting in a spread of charge throughout the deeper retinal layers. This behavior was also observed in suprachoroidal bipolar stimulation of the cat retina [8]. As we have not modeled the complete retinal architecture for the ON system, there is an absence of inhibitory inputs present in the model. This is seen in the suprachoroidal study as the direct stimulation of these deeper layers have resulted in more pronounced presynaptic excitatory currents facilitating multiple late onset APs as shown in Fig. 3. V. CONCLUSION This study was undertaken to establish a basis for a retinal model that can incorporate both active and passive properties of the retina using published experimental data. We intend to expand the model's retinal architecture in future studies, as well as incorporate more complex electrode arrangements to help in the design of vision prosthetic devices.

[6] [7]

[8]

[9]

[10]

[11] [12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

REFERENCES
[1] Klaver, C.C.W, Wolfs, R.C.W., Vingerling, J.R., Hofman, A., and de Juan, P.T.V.M. Age-specific prevalence and causes of blindness and visual impairment in an older population. Arch. Ophthalmol. 116:653-658, 1998. Stone, J.L., Barlow, W.E., Humayun, M.S., de Juan, E., Milam, A.H. Morphometric analysis of macular photoreceptors and ranglion cells in retinas with retinitis pigmentosa. Arch. Ophthalmol. 110:16341639, 1992. Santos, A., Humayun, M.S., de Juan, E., Greenberg, R.J., Marsh, M.J., Klock, I.B., Milam, A.H. Preservation of the inner retina in retinitis pigmentosa. A morphometric analysis. Arch. Ophthalmol. 115:511-515, 1997. Humayun, M.S., de Juan, Jr., E., Dagnelie, G., Greenberg, R.J., Propst, R.H., Phillips, D.H. Visual perception elicited by electrical stimulation of retina in blind humans. Arch. Ophthalmol. 114(1):4046, 1996. Rizzo, J.F. III, Wyatt, J., Loewenstein, J., Kelly, S., Shire, D. Methods and perceptual thresholds for short-term electrical stimulation of human retina with microelectrode arrays. Invest. Ophthalmol. Vis. Sci. 44(12):5355-5361, 2003. [22]

[2]

[23]

[24]

[3]

[4]

[25]

Wyatt, J., Rizzo, J.F. III. Ocular implants for the blind. IEEE Spectr. 3:47-53, 1996. Sakaguchi, H., Fujikado, T., Fang, X., Kanda, H., Osanai, M., Nakauchi, K., Ikuno, Y., Kamei, M., Yagi, T., Nishimura, S., Ohji, M., Yagi, T., Tano, Y. Transretinal electrical stimulation with a suprachoroidal multichannel electrode in rabbit eyes. Japan J. Ophthalmol. 48:256-61, 2004. Wong, Y.T., Chen, S.C., Seo, J.M. Morley, J.W., Lovell, N.H., Suaning, G.J. Focal activation of the feline retina via a suprachoroidal electrode array. Vision Res. 49:825-833, 2009. Greenberg, R.J., Velte, T.J., Humayun, M.S., Scarlartis, G.N., de Juan, Jr., E. A computational model of electrical stimulation of the retinal ganglion cell. IEEE Trans. on Biomedical Eng. 46(5):505-514, 1999. Fohlmeister, J.F., Miller, R.F. Impulse encoding mechanisms of ganglion cells in the tiger salamander retina. J. Neurophysiol., 78:1935-1947, 1997. Wssle, H., Boycott, B.B. Functional architecture of the mammalian retina. Physiological Reviews, 71(2):447-480, 1991. Tsai, D., Morley, J.W., Suaning, G.J., Lovell, N.H. Direct activation and temporal response properties of rabbit retinal ganglion cells following subretinal stimulation. Journal of Neurophysiology, 102(5):2982-2993, 2009. Fried, S.I., Hsueh, H.A., Werblin, F.S. A method for generating precise temporal patterns of retinal spiking using prosthetic stimulation. Journal of Neurophysiology, 95(2):970-978, 2006. Sekirnjak, C., Hottowy, P., Sher, A., Dabrowski, W., Litke, A.M., Chichilnisky, E.J. Electrical stimulation of mammalian retinal ganglion cells with multielectrode arrays. Journal of Neurophysiology, 95:3311-3327, 2006. Cottaris, N.P., Elfar, S.D. How the retinal network reacts to epiretinal stimulation to form the prosthetic visual input to the cortex. J. Neural Eng. 2:S74-S90, 2005. Teeters, J., Jacobs, A., Werblin, F. How neural interactions form neural responses in the salamander retina. Journal of Comp. Neurosci., 4:5-27, 1997. Brown, J.S., Flitcroft, D.I., Gui-shuang, Y., Francis, E.L., Schmid, G.F., Quinn, G.E., Stone, R.A. In vivo human choroidal thickness measurements: evidence for diurnal fluctuations. Invest. Ophthalmol. Vis. Sci. 50(1):5-12, 2009. Karwoski, C.J., Xu, X. Current source-density analysis of lightevoked field potentials in rabbit retina. Vis. Neurosci., 16:369-377, 1999. Brindley, G.S. The passive electrical properties of the frogs retina, choroid and sclera for radial fields and currents. J. Physiol., 134:339352, 1956. Dokos, S., Suaning, G.J., Lovell, N.H. A bidomain model of epiretinal stimulation. IEEE Trans. Neural System and Rehabilitation Eng. 13(2): 137-146, 2005. Coleman, P., Miller, R.F. Measurement of passive membrane parameters with whole-cell recordings from neurons in the intact amphibian retina. Journal of Neurophysiology, 61(1):218-230, 1989. Oyster, C.W., Takahashi, E.S., Hurst, D.C. Density, soma size, and regional distribution of rabbit retinal ganglion cells. The Journal of Neuroscience, 1(12):1331-1346, 1981. Freed, M., Smith, R.G., Sterling, P. Computational model of the onalpha ganglion cell receptive field based on bipolar cell circuitry. Proc. Natl, Acad. Sci. USA, 89:236-240, 1992. Yamauchi, Y., Franco, L.M., Jackson, D.J., Naber, J.F., Ziv, R.O., Rizzo, J.F. III., Kaplan, H.J., Enzmann, V. Comparison of electrically evoked cortical potential thresholds generated with subretinal or suprachoroidal placement of a microelectrode array in the rabbit. J. Neural Eng., 2:S48-S56, 2005. Jensen, R.J., Rizzo, J.F. III. Activation of ganglion cell in wild-type and rd1 mouse retinas with monophasic and biphasic current pulses. J. Neural Eng. 6:1-7, 2009.

[5]

2080