yin_et_al_2010

Upload: adal-arasan

Post on 02-Apr-2018

214 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/27/2019 Yin_et_al_2010

    1/4

    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. INTRODUCTIONetinitis pigmentosa and age-related macular

    degeneration are two of the most prevalent retinal

    diseases and leading causes of blindness in developedcountries [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 onsetor secondary spikes have been

    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).

    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. METHODSA. 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

    ( ) IVeb = (1)

    where b is the conductivity of the bulk vitreous medium

    andIis 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).

    Shijie Yin, Nigel H. Lovell, Senior Member, IEEE, Gregg J. Suaning,Member, IEEE, Socrates Dokos,Member, IEEE

    A Continuum Model of the Retinal Network and its Response to

    Electrical Stimulation

    R

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

    978-1-4244-4124-2/10/$25.00 2010 2077Crown

  • 7/27/2019 Yin_et_al_2010

    2/4

    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 isnot 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

    mee IV =2 (2)

    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

    ( )irrsnion

    m

    mm VVgiJt

    VCI =

    ++

    = (3)

    and Vm is the transmembrane potential given by

    eim

    VVV = (4)

    Cmis 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 thebipolar cells and inhibitory input arrives from wide-field

    amacrine cells.

    ( )snmsnsnsn EVgPi = (5)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]

    ( )sn

    snsn tPpdt

    dP= (6)

    wherepis,

    ( )5050

    5050

    exp1

    exp

    VV

    VVp

    pre

    pre

    +

    =

    (7)

    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 wasexpressed as,

    ( )irr

    m

    rmmmm VVg

    R

    VV

    dt

    dVCI =

    += (8)

    whereRmis the specific membrane resistance. For parameter

    consistency, we have adopted values from [15] for all

    relevant cell types.

    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

    Cm Jion isn

    gr

    Vr +

    gr

    Vr

    RmCm

    Ve

    Vi Vi

    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 g r is the

    resistive tie connecting the intracellular potential V i to a resting

    potential Vr, Rm is the specific membrane resistance.

    x

    z

    y

    2078

  • 7/27/2019 Yin_et_al_2010

    3/4

    B. Numerical ImplementationSimulations 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. DISCUSSIONThe 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

  • 7/27/2019 Yin_et_al_2010

    4/4

    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 thecomplete 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. CONCLUSIONThis 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 toexpand 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.

    REFERENCES

    [1] Klaver, C.C.W, Wolfs, R.C.W., Vingerling, J.R., Hofman, A., and deJuan, P.T.V.M. Age-specific prevalence and causes of blindness and

    visual impairment in an older population. Arch. Ophthalmol.

    116:653-658, 1998.

    [2] 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. 110:1634-

    1639, 1992.[3] 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.

    [4] 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):40-

    46, 1996.

    [5] Rizzo, J.F. III, Wyatt, J., Loewenstein, J., Kelly, S., Shire, D. Methodsand perceptual thresholds for short-term electrical stimulation of

    human retina with microelectrode arrays. Invest. Ophthalmol. Vis.

    Sci. 44(12):5355-5361, 2003.

    [6] Wyatt, J., Rizzo, J.F. III. Ocular implants for the blind. IEEE Spectr.3:47-53, 1996.

    [7] 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.

    [8] 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.

    [9] Greenberg, R.J., Velte, T.J., Humayun, M.S., Scarlartis, G.N., deJuan, Jr., E. A computational model of electrical stimulation of the

    retinal ganglion cell.IEEE Trans. on Biomedical Eng. 46(5):505-514,

    1999.

    [10] Fohlmeister, J.F., Miller, R.F. Impulse encoding mechanisms ofganglion cells in the tiger salamander retina. J. Neurophysiol.,

    78:1935-1947, 1997.

    [11] Wssle, H., Boycott, B.B. Functional architecture of the mammalianretina. Physiological Reviews, 71(2):447-480, 1991.

    [12] Tsai, D., Morley, J.W., Suaning, G.J., Lovell, N.H. Direct activationand temporal response properties of rabbit retinal ganglion cells

    following subretinal stimulation. Journal of Neurophysiology,

    102(5):2982-2993, 2009.

    [13] Fried, S.I., Hsueh, H.A., Werblin, F.S. A method for generatingprecise temporal patterns of retinal spiking using prosthetic

    stimulation.Journal of Neurophysiology, 95(2):970-978, 2006.

    [14] 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.

    [15] Cottaris, N.P., Elfar, S.D. How the retinal network reacts to epiretinalstimulation to form the prosthetic visual input to the cortex. J. Neural

    Eng. 2:S74-S90, 2005.

    [16] Teeters, J., Jacobs, A., Werblin, F. How neural interactions formneural responses in the salamander retina. Journal of Comp.

    Neurosci., 4:5-27, 1997.

    [17] 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.

    [18] Karwoski, C.J., Xu, X. Current source-density analysis of light-evoked field potentials in rabbit retina. Vis. Neurosci., 16:369-377,

    1999.[19] Brindley, G.S. The passive electrical properties of the frogs retina,

    choroid and sclera for radial fields and currents. J. Physiol., 134:339-

    352, 1956.

    [20] Dokos, S., Suaning, G.J., Lovell, N.H. A bidomain model of epiretinalstimulation. IEEE Trans. Neural System and Rehabilitation Eng.

    13(2): 137-146, 2005.

    [21] Coleman, P., Miller, R.F. Measurement of passive membraneparameters with whole-cell recordings from neurons in the intact

    amphibian retina.Journal of Neurophysiology, 61(1):218-230, 1989.

    [22] Oyster, C.W., Takahashi, E.S., Hurst, D.C. Density, soma size, andregional distribution of rabbit retinal ganglion cells. The Journal of

    Neuroscience, 1(12):1331-1346, 1981.

    [23] Freed, M., Smith, R.G., Sterling, P. Computational model of the on-alpha ganglion cell receptive field based on bipolar cell circuitry.

    Proc. Natl, Acad. Sci. USA, 89:236-240, 1992.

    [24] 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 electricallyevoked cortical potential thresholds generated with subretinal or

    suprachoroidal placement of a microelectrode array in the rabbit. J.

    Neural Eng., 2:S48-S56, 2005.

    [25] Jensen, R.J., Rizzo, J.F. III. Activation of ganglion cell in wild-typeand rd1 mouse retinas with monophasic and biphasic current pulses.

    J. Neural Eng. 6:1-7, 2009.

    2080