Microstimulation of visual cortex to restore vision

Abstract

This review argues that one reason why a functional visuo-cortical prosthetic device has not been developed to restore even minimal vision to blind individuals is because there is no animal model to guide the design and development of such a device. Over the past 8 years we have been conducting electrical microstimulation experiments on alert behaving monkeys with the aim of better understanding how electrical stimulation of the striate cortex (area V1) affects oculo- and skeleto-motor behaviors. Based on this work and upon review of the literature, we arrive at several conclusions: (1) As with the development of the cochlear implant, the development of a visuo-cortical prosthesis can be accelerated by using animals to test the perceptual effects of microstimulating V1 in intact and blind monkeys. (2) Although a saccade-based paradigm is very convenient for studying the effectiveness of delivering stimulation to V1 to elicit saccadic eye movements, it is less ideal for probing the volitional state of monkeys, as they perceive electrically induced phosphenes. (3) Electrical stimulation of V1 can delay visually guided saccades generated to a punctate target positioned in the receptive field of the stimulated neurons. We call the region of visual space affected by the stimulation a delay field. The study of delay fields has proven to be an efficient way to study the size and shape of phosphenes generated by stimulation of macaque V1. (4) An alternative approach to ascertain what monkeys see during electrical stimulation of V1 is to have them signal the detection of current with a lever press. Monkeys can readily detect currents of 1-2 microA delivered to V1. In order to evoke featured phosphenes currents of under 5 microA will be necessary. (5) Partially lesioning the retinae of monkeys is superior to completely lesioning the retinae when determining how blindness affects phosphene induction. We finish by proposing a future experimental paradigm designed to determine what monkeys see when stimulation is delivered to V1, by assessing how electrical fields generated through multiple electrodes interact for the production of phosphenes, and by depicting a V1 circuit that could mediate electrically induced phosphenes.

Fig. 1.

Left: The percentage of saccade evoked electrically from V1 and made into the receptive field (RF) under six task conditions (right: a–f). Task conditions were randomiy selected with the same block of testing. Each condition was tested for 120 trials (20 trials per condition per block for a total of six blocks). The Z-statistic was used to assess the effect of condition on the visually- and electrically evoked saccades terminating in the receptive field of the stimulated V1 cells. The receptive field of the stimulated neurons at the electrode tip was located 2.6° in eccentricity and 265° of meridian, as schematically illustrated on the extreme right in reference to the position of the fixation spot (fix). A visual target (targ) in conditions (a) and (b) was displayed in the location of the receptive field 100 ms following the termination of the fixation spot. Stimulation (stim) was delivered 130 ms after termination of the fixation spot. A drop of apple juice (j) was delivered after the monkey’s eyes entered the target window (a–d) or immediately after the termination of the fixation spot (e and f). The depth of stimulation was 1.25 mm below the cortical surface. Stimulation current, pulse duration, pulse frequency, and train duration were 30 μA, 0.2 ms, 200 Hz, and 100 ms, respectively, in the form of anode-first pulses. Adapted from Tehovnik et al. (2005a).

Fig. 2.

Double-step paradigm. Monkeys were trained to generate saccadic eye movements to two visual targets flashed in the visual field such that the 2nd target often fell in the visual receptive field of the V1 cells under study during initial fixation (fix, left-top). On some trials the 2nd target was presented outside the visual receptive field at T2′ (left-bottom) or at one of two other locations in the upper left field (not shown). The first target (T1) was flashed for 50 ms and the 2nd target (T2) was flashed for 100 ms. The monkey had 500 ms to acquire the 2nd target in order to obtain a juice reward. On stimulation trials (right), the 2nd target was replaced by a 50-ms train of electrical stimulation. Following stimulation, the monkey always generated the 2nd saccade to the receptive-field location of the stimulated cells at the time of initial fixation (right-top) rather than to a target location outside of the receptive field location (right-bottom). On blank trials in which no stimulation was delivered no saccades were generated to any target locations within the 500-ms choice period. Adapted from Schiller et al. (2005).

Fig. 3.

(A) Percent choice to the receptive-field location affected by stimulation is plotted as a function of the size of visual target (at 10% positive contrast) presented in the right field opposite to that of the receptive field of the directly stimulated V1 cells (right inset). Using 60-μA pulses delivered at 200 Hz embedded in an 80-ms train, the phosphene generated was 0.3° in size as indicated by the 50% crossover point (arrow). (B) Percent choice to the receptive-field location affected by stimulation is plotted as a function of target contrast (at 0.3° in size) presented in the right field opposite to that of the receptive field of the directly stimulated V1 cells (right inset). Using 60-μA pulses delivered at 200Hz embedded in an 80-ms train, the phosphene generated was 7% in contrast as indicated by the 50% crossover point (arrow). The receptive-field location of the stimulated neurons was 2.8° of eccentricity. At the stimulated V1 location the receptive field is about 0.3° in diameter (Dow et al., 1981; Hubel and Wiesel, 1974b) and is modulated by a stimulus contrast of 2–20% (Albrecht and Hamilton, 1982). Adapted from Schiller et al. (2005).

Fig. 4.

Two schemes are illustrated for the electrical evocation of motor responses from the neocortex of monkeys. On the right, the stimulation (stim) evokes a motor response that is non-reward driven and that is described as being mediated outside of the animal’s volitional state as a motor command. On the left, the stimulation evokes a motor response that is reward driven and is a reflection of a monkey’s choice. Using behavioral paradigms that make distinctions between these two extremes is central to making clear-cut interpretations about the effects of electrical microstimulation (Chen and Tehovnik, 2007; Tehovnik and Slocum, 2004).

Fig. 5.

(A) A delay field was mapped by presenting a 100-ms train of stimulation at the end of the fixation period (fix) immediately before the presentation of the visual target (targ). The animal was given 500 ms to generate a saccadic eye movement (sacc) to the visual target in order to obtain a juice reward (juice). The visual target [0.2degrees in diameter and at 33% positive contrast (Michaelson)] was presented at various locations with respect to the visual receptive field (RF) of the stimulated V1 cells. (B) Size of the delay field varies with the site of stimulation within the operculum of V1. Size of the delay field is plotted as a function of the eccentricity of the receptive-field center of the V1 cells stimulated. A total of 41 V1 sites (located from 0.9 to 2.0 mm below the cortical surface) were studied. At each site, stimulation was composed of 100-μA anode-first pulses (at 0.2-ms duration) delivered at 200 Hz using a 100-ms train. The solid curve is a regression line representing the data (r = 0.81, n = 41, p<0.01). (C) The relationship between average receptive-field size (in visual-field coordinates) for macaque V1 and receptive-field eccentricity [(I) Hubel and Wiesel, 1974b; (II) Dow et al., 1981] and the relationship between amount of visual field represented by activation of a 0.75-mm diameter region of macaque V1 and receptive-field eccentricity using the complex logarithmic function of Schwartz (1994) (III) are compared to the relationship between the size of the delay field and receptive-field eccentricity using a 100 μA current (from B). Using the delay-field data of (B), a calculation can be made of how far the 100 μA current was spreading in V1. At an eccentricity of 3°, the size of the visual field affected by the 100 μA current was 0.24° (on average). At a 3° eccentricity, 1000 μm of V1 tissue represents about 0.32° of visual field, which is based on the retino-cortical magnification factor of macaque V1 (Hubel and Wiesel, 1974b; Dow et al., 1981; Tootell et al., 1988). Therefore, 100μA affects V1 tissue within 375 μm from the electrode tip (i.e., 0.24°/0.32° × 1000 μm/2). A simplified version of the Schwartz (1994) algorithm was used to compute the function (III) in (C) representing the amount of visual field affected by activating a sphere of tissue with radius of 0.375 mm: Visual-field size = K × E × R, where K is a constant of 0.23°/deg/mm, E (in degrees) the visual-field eccentricity coded by the stimulated cells, and R (in mm) the radial spread of the current, that is, 0.375 mm. Adapted from Tehovnik et al. (2005b).

Fig. 6.

The size of phosphenes evoked from macaque V1 is illustrated for three levels of current: 25, 50, and 100 μA. The electrodes were spaced by 0.5 mm within the neocortex. Based on delay-field data (Tehovnik and Slocum, 2007d) the current was found to spread by 0.19, 0.27, and 0.39 mm from the electrode tip using the current-distance equation: I = 675 μA/mm² × r², where I is the current in microamperes and r the radial spread of current from the electrode tip in millimeters. The current-distance constant of 675 μA/mm² is derived from experimental data (Tehovnik et al., 2006). Visual-field size was computed using a variation of the Schwartz (1994) algorithm: Visual-field size = K × E × R, where K is a constant of 0.23°/deg/mm, E (in degrees) the visual-field eccentricity coded by the stimulated cells, and R (in mm) the radial spread of current derived from the current–distance equation. Notice that for currents under 50 μA the current fields are not overlapped; nor are the evoked phosphenes overlapped for the three eccentricities shown.

Fig. 7.

The current threshold to evoke a detection response on 50% of stimulation trials is plotted as a function of cortical depth. For all experiments (N = 5), a 100-ms train composed of anode-first pulses was delivered at 200 Hz. Stimulation was delivered through a platinum–iridium electrode whose impedance value was maintained at 1.8 MΩ. All stimulation occurred 200 ms before the end of the fixation period. The monkey had 800 ms to register a response by depressing a lever following the termination of the fixation spot. Blank trials having no stimulation were interleaved with stimulation trials 50% of the time. The rate of responding on such trials was less than 10%. The square marker indicates that a detection response was not readily evoked from the site using a maximal current of 3 μA. The lowest threshold of 1.6 μA occurred at 1.6 mm below the cortical surface. Standard errors of the mean are shown.

Fig. 8.

Retinal lesioning. All lesions were made using a NIDEK GYC-2000 532-nm photocoagulation laser. (A) Left: Picture of the right eye fundus immediately after inducing a laser photocoagulation lesion. The lesion appears white. Right: Picture of the right fundus after extraction and fixation in formaldehyde at the end of the experiment, 9 months later. The monkey had been euthanized and perfused with formaldehyde prior to extraction. Note the hyperpigmented scar (white arrow), which corresponds to the retinal lesion. (B) Fifteen-micrometer thick section through the center of the same lesion, stained with hematoxylin-eosin. Note that all retinal layers are essentially completely destroyed at the center of the lesion. (C) Saccade to visual target task used to map the visual-field scotoma induced in a macaque after a homonymous retinal lesion (D). Adjacent saccade targets were spaced by ~0.5°. Each dot represents a successful saccade to target. Note the absence of saccades outlining the area of the induced scotoma. Scale: Degrees. (D) Illustration of the lesions made in both fundi in order to induce a partly homonymous lesion. Left: Lesion in the left eye fundus. Right: Overlay of the right eye fundus (larger) and the left eye fundus (smaller square) illustrating the lesions (white patches marked by the black arrowheads). The left retina (small square overlay) was mirrored along the vertical axis and scaled to make the optical nerves overlap. Black dots outline the left eye lesion and its homonymous location in the right eye, which lies almost entirely over the right eye lesion resulting in an essentially homonymous left visual field scotoma. Note that lesions in each eye were offset from the fovea such that the foveal representation was spared, thereby allowing the animal to fixate accurately. Adapted from Smirnakis et al. (2005). (See Color Plate 23.8 in color plate section.)

Fig. 9.

Detection of V1 stimulation and the target-choice paradigm. Monkeys are required to fixate (fix) a spot of light as targets are presented briefly within each hemifield: the left target (left targ) is situated in the receptive field of the cells under study and the right target (right targ) is positioned in the mirror-opposite hemifield. Following termination of the fixation spot, a monkey must pull the left lever (left lever) if the left target is brighter or larger than the right target and the monkey must pull the right lever (right lever) if the right target is brighter or larger than the left target. A juice reward (juice) is delivered for a correct response. The monkey is free to generate a saccade (sacc) away from the fixation location once the fixation spot has been terminated. On a fraction of trials, electrical stimulation (stim) is delivered to the receptive-field neurons instead of presenting the target (left targ). On stimulation trials all responses are rewarded.

Fig. 10.

(A) The collision method. (Top) By first delivering conditioning pulses (c) to the somal end of a fiber bundle and then by the delivery test pulses (t) to the terminal end of the fiber bundle (middle), the interval at which the threshold to evoke a behavioral response doubles (for reduced c–t intervals) marks the collision interval between the action potentials evoked by the c and t pulses (bottom). When t pulses are delivered outside the collision interval the threshold to evoke a behavioral response is reduced since both c and t pulses now contribute to the elicitation of behavior. Once the collision interval is determined, the conduction time between the somal and terminal electrode can be computed by subtracting the refractory period at the terminal electrode from the collision interval. Refractory periods are ascertained by delivering c and t pulse through a single electrode (not shown). As with the collision interval, the refractory period is the c–t interval at which the threshold doubles. For complete methodological details see Yeomans and Tehovnik (1988). (B) Phosphene circuit. A laminated portion of V1 is shown including laminae II–VI. Vertically aligned pyramidal fibers are proposed to mediate phosphenes, which are extinguished by GABAergic interneurons (GABA) (Tehovnik and Slocum, 2007d). Pyramidal fibers from superficial V1 (pyrl) terminate on pyramidal fibers from deep V1 (pyr2) (Peters and Sethares, 1991). The latter send projections to the motor system. To determine whether vertically aligned pyramidal fibers mediate phosphenes the collision test could be performed by delivering pulses through the top and bottom electrode (i.e., stim_1 and stim_2, respectively). Additionally, pharmacological agents could be infused at the bottom site (stim_2) to study their effects on the threshold to evoke phosphenes from the top electrode (stim_1).