Importance of Eye Position on Spatial Localization in Blind Subjects Wearing an Argus II Retinal Prosthesis

Abstract

Purpose.: With a retinal prosthesis connected to a head-mounted camera (camera-connected prosthesis [CC-P]), subjects explore the visual environment through head-scanning movements. As eye and camera misalignment might alter the spatial localization of images generated by the device, we investigated if such misalignment occurs in blind subjects wearing a CC-P and whether it impacts spatial localization, even years after the implantation.

Methods.: We studied three subjects blinded by retinitis pigmentosa, fitted with a CC-P (Argus II) 4 years earlier. Eye/head movements were video recorded as subjects tried to localize a visual target. Pointing coordinates were collected as subjects were requested to orient their gaze toward predetermined directions, and to point their finger to the corresponding perceived spot locations on a touch screen. Finally, subjects were asked to give a history of their everyday behavior while performing visually controlled grasping tasks.

Results.: Misaligned head and gaze directions occurred in all subjects during free visual search. Pointing coordinates were collected in two subjects and showed that median pointing directions shifted toward gaze direction. Reportedly all subjects were unable to accurately determine their eye position, and they developed adapted strategies to perform visually directed movements.

Conclusions.: Eye position affected perceptual localization of images generated by the Argus II prosthesis, and consequently visuomotor coordination, even 4 years following implantation. Affected individuals developed strategies for visually guided movements to attenuate the impact of eye and head misalignment. Our observations provide indications for rehabilitation procedures and for the design of upcoming retinal prostheses. (ClinicalTrials.gov number, NCT00407602.)

Introduction

Retinitis pigmentosa (RP) is a potentially blinding degenerative disease of the retina. The condition results in the destruction of photoreceptors, but characteristically spares other retinal neurons.¹ 

Various biologic and bioelectronic approaches have been proposed to restore some vision in individuals blinded by RP.^2,3 Particular interest has been generated by the development of retinal prostheses.⁴ These devices are designed to electrically stimulate preserved cells in the inner retina and thus to produce a neural signal that is eventually conveyed by optic nerves to the brain, where visual percepts are generated.⁵ 

Clinical trials with retinal prostheses have brought promising results.⁶ Subjects fitted with such systems can identify windows and doors; some of them are able to count fingers and even to read short sentences presented on a computer screen.^6–8 Moreover, retinal prostheses provide invaluable visual information to perform visually directed movements, such as following a layout on the ground or seizing objects.^9–11 

According to their underlying technology, retinal prostheses can be categorized into two general groups. One comprises devices consisting of a photodiode array placed close to the retina,¹² receiving and converting light into electrical impulses. The other comprises systems including a microcamera fitted to a glass frame,⁶ which records pixelated visual information that is eventually converted into an electronic signal; this is then sent to a microelectrode implant fixed onto the retina. With both techniques, the produced electrical signals stimulate preserved retinal cells (essentially retinal bipolar and ganglion cells), which start processing the signal and forward it along anterior visual pathways to the brain, where visual percepts are generated.⁵ 

Over currently available photovoltaic-based systems, camera-connected devices demonstrate the benefits of providing preprocessed information, allowing magnification and other image optimization that extend the functionality of the prosthesis beyond the spatial resolution limited by the electrode number and spacing (Sahel J-A, et al. IOVS 2013;54:ARVO E-Abstract 1389). However, subjects fitted with a camera-connected prosthesis (CC-P) initially show difficulties in visuomotor coordination, such as misdirecting their hand while trying to seize an object using the visual information provided by the device (Sabbah N, Authié CN, Sanda N, Mohand-Said S, Sahel J-A, Safran AB, unpublished observations, 2014). This might be related to the fact that, in contrast to normal individuals who spatially locate the image according to gaze direction, these subjects need to visually locate the image according to the camera (i.e., head) orientation.¹³ Given that experimental manipulation of the eye position alters the localization process of visual percepts in healthy individuals,^14–16 and because eye movements still occur in blind subjects,¹⁷ one might wonder whether gaze shifts potentially modify perceived location of the images generated by the prosthetic device. In more general terms, one might raise the question whether dissociating eye and head directions tends to disturb spatial localization of images produced by such prostheses. This potential information conflict was suggested by some authors to occur in RP subjects fitted with a camera-based prosthetic device,^13,18 but to our knowledge it has not been properly evaluated in affected individuals. This prompted us to explore the issue. Moreover, considering that following implant surgery, CC-P-fitted individuals are advised to attempt maintaining their eyes in primary position, and that neural processing is reorganized following prosthetic implantation,¹⁹ we wondered whether, following years of regular use of the device, some potential information conflict still occurred as a result of eye and head misalignment. This investigation was therefore conducted in individuals who received a CC-P 4 years prior to our examinations. 

We investigated (1) whether head and gaze misalignment occurs during head-free visual search; (2) the impact of this eventual dissociation on the spatial localization process during a task in which eyes and head directions were controlled; and (3) subjects' observations of their own behavior when attempting to seek and grasp an object. 

Study results were expected to provide valuable information on the dissociation of eye and head positions, the strategies used by the subjects to manage the presumably biased image localization processes after 4 years of training, and important indications for optimizing rehabilitation procedures of the affected individuals. These findings might also be taken into account for the design of upcoming generations of retinal prostheses and training protocols. 

Subjects and Methods

The study included three male subjects (subjects 1, 2, and 3), right-handed, aged 53, 61, and 64 years, respectively. They all suffered from advanced RP, with a visual acuity reduced to light perception. Four years prior to study inclusion, they had their right eye fitted with a 60-electrode epiretinal prosthesis (Argus II; Second Sight Medical Products, Sylmar, CA, USA) at the Centre Hospitalier National d'Ophtalmologie des Quinze-Vingts in Paris. This prosthesis is connected to an external device consisting of a video camera and transmitter mounted on eyeglasses and a video processor unit that converts the image into electronic signals.⁶ 

Our study is related to the multicenter feasibility protocol for the Argus II retinal prosthesis system; it was conducted in accordance with the Declaration of Helsinki and applicable regulations for medical device clinical trials in the different countries where the study was conducted (www.clinicaltrials.gov, registration number NCT00407602). The study has been approved by ethics committees and institutional review boards of participating institutions, as well as governmental health agencies in these countries. All subjects signed an informed consent for their participation, including the use of video recordings and photographs for publication. 

During the tests, the subjects were seated, at a viewing distance of 40 cm from the tactile screen (Elotouch 1928L; ELO Touch Solutions, Menlo Park, CA, USA) showing a 4/3 format, 19-inch diagonal, and 1280 × 1024 resolution, subtending 51° horizontally × 40° vertically. The standard deviation of error of the tactile screen is not greater than 8 pixels (i.e., approximately 2.5 mm). On that screen, white targets were displayed over a black background. Targets were 35 pixels wide on the touch screen (i.e., approximately 1 cm, or 1.5° at 40 cm). Therefore, according to its location, the stimulus will activate one to four electrodes of the retinal prosthesis. 

To perform the requested visual tasks, the subjects used the visual information provided by their retinal prosthesis. Tests were conducted in a darkened room. Eye and head movements were recorded by an infrared camera (either a Somikon DV-883.IR [Pearl Diffusion, Selestat, France] or a HDR-XR500 [Sony, Tokyo, Japan]) positioned under the touch screen. 

Misalignment of Head and Gaze Directions During Visual Search

To determine whether obvious misalignment occurred between head and gaze directions during visual search, the examiner triggered the presentation of a white spot at a random location on the screen. With the prosthetic device on, subjects were asked to localize and touch the target on the touch screen with their right index finger. The procedure was repeated 20 times. Head and eye movements were recorded using an infrared camera; recordings were later reviewed to qualitatively determine whether head and gaze directions were misaligned during visual search and target detection. 

Effect of Gaze Orientation on Target Localization

Before starting the evaluation procedure, we trained the subjects to shift their gaze eccentrically without moving the head. As blind individuals are able to orient their gaze direction toward a proprioceptive cue (e.g., their hand),¹⁷ our subjects were requested to direct their eyes to their right index finger positioned by the examiner against the screen on five successive locations, namely, at the center of the screen and then to the right and to the left, up, and down, 4 cm (i.e., approximately 7°) from screen borders. Subjects were requested to memorize these locations for the subsequent evaluation procedure. 

Testing was then conducted as follows. While the subject, using his eyeglass-mounted camera, was asked to visualize a target presented at the center of the screen and to maintain his head still, he was requested to shift his gaze toward five potential directions prescribed by the examiner, that is, at the screen center, up, down, right, and left. After completing each requested gaze shift, the subject had to immediately indicate the perceived position of the visual target by pressing with his right index finger on the touch screen at the corresponding location. The procedure included 50 trials, 10 at each of the five tested locations, performed in a pseudo-random order. Horizontal and vertical screen coordinates of each pointed locations were recorded. 

For data analysis, pointed locations were grouped according to associated gaze directions. A Mann-Whitney test with Bonferroni correction for multiple comparisons was applied between each pointing group for values measured along horizontal and vertical axes, respectively. 

Subject's Appreciation of His Visual Localization Abilities, and Perceived Gaze Position

The tested subject was asked to describe the strategy used in everyday life for localizing and then seizing an object, using the information provided by the implant, and to indicate whether he perceived his gaze orientation. 

Results

Subject 1

Misalignment of Head and Gaze Directions During Visual Search.

Camera recordings demonstrated that during visual search in the head-free condition, contraversive eye movements (i.e., vestibulo-ocular reflex [VOR]) occurred upon head rotations (Fig. 1). Moreover, it was observed that finger pointing was occasionally performed while the CC-P-fitted eye and camera were misaligned. 

Figure 1

 

Subject 1. Vestibulo-ocular reflex observed during search head movements. (A) Before head rotation, eyes are oriented rightward; (B) following head rotation to the right, gaze remains stable in space. Note that, in the left eye, distance (white line) between the limbus and temporal canthus is reduced in (B) compared with (A).

Effect of Gaze Orientation on Target Localization.

When subject 1 was for the first time requested to shift gaze toward a determined direction while keeping head stable, he was unable to carry out the task, and instead turned both eyes and head in a conjugate manner. After several attempts, however, he succeeded in volitionally dissociating gaze and head directions and performing the task. Subject 1 performed the requested finger pointing within the limits of the screen. 

Median pointing directions were shifted toward the gaze direction (Mann-Whitney tests P < 0.01, see Table 1 and Fig. 2). Thus, with left gaze, pointing coordinates were significantly shifted to the left of the center. Conversely, with right gaze, pointing coordinates were significantly shifted to the right. With up and down gaze, locations were respectively significantly shifted upward and downward of the center. 

Figure 2

 

Subject 1. Coordinates of pointing locations, colored according to gaze directions. Colored crosses indicate respective median pointing positions for each tested gaze direction, and dashed ellipses indicate confidence intervals (2 SD). Dotted corners show the limits of the screen.

Table 1

 

Subject 1: Finger-Pointing Coordinates Grouped by Gaze Direction Compared to Those Obtained With Gaze Directed to the Center

Pointing Coordinates	Condition 1	Condition 2	Statistics
Horizontal	Right gaze	Gaze to the center	U = 99, n1 = n2 = 10, P = 0.0002
	Left gaze	Gaze to the center	U = 100, n1 = n2 = 10, P = 0.0001
Vertical	Up gaze	Gaze to the center	U = 100, n1 = n2 = 10, P = 0.0018
	Down gaze	Gaze to the center	U = 100, n1 = n2 = 10, P = 0.0001

 Values recorded along the horizontal axis and those recorded along the vertical are considered independently. U, Mann-Whitney values; n1, first condition sample size; n2, second condition sample size; P, statistical significance.

Independently of this effect of gaze direction, it appeared that the pointing locations of grouped right, center, and left gaze conditions were significantly shifted downward (when compared to the central target position, Mann-Whitney tests U = 23, n = 30, P < 0.01). Also, the pointing locations of grouped up, center, and down conditions were significantly shifted rightward (Mann-Whitney tests U = 462, n = 30, P < 0.01, see Fig. 2). 

Subject's Appreciation of His Visual Localization Abilities, and Perceived Gaze Position.

Subject 1 reported that in everyday life, to accurately localize and then seize an object using the implant image, he first attempted to get the target image in the viewing field, then performed successive small head movements, either horizontally or vertically. If he had not accomplished these preliminary oscillatory head movements, he reportedly failed to grasp the object. Moreover, the subject indicated that he only roughly perceived where his gaze was directed. 

Subject 2

Misalignment of Head and Gaze Directions During Visual Search.

Camera recordings showed that during head-free visual search, head rotations induced reflexive contraversive eye movements (Fig. 3; Supplementary Video S1). Moreover, misaligned head and gaze directions were also observed when the subject detected and intended to point to the target on the screen (Fig. 4; Supplementary Video S2). 

Figure 3

 

Subject 2. Vestibulo-ocular reflex observed following search head movements. (A) Before head rotation, both head and eyes are oriented forward; (B) after head rotation to the left, gaze remains directed forward, being stabilized in space by VOR. Note that, in the right eye, distance (white line) between the limbus and temporal canthus is reduced in (B) compared with (A) (see Supplementary Video S1).

Figure 4

 

Subject 2. Misaligned head and right (CC-P fitted, viewing) eye directions at the time the subject just detected and was intending to point to the target on the screen (see Supplementary Video S2).

Effect of Gaze Orientation on Target Localization.

From the start, subject 2 was able to direct his gaze toward the directions indicated by the examiner. He performed the requested finger-pointing task within the limits of the screen. 

In 7 out of 50 trials, subject 2 did not press firmly enough on the screen when indicating the perceived location, and the corresponding results could not be recorded and taken into account. 

Median pointing directions were shifted toward the gaze direction (Mann-Whitney tests P < 0.01, see Table 2 and Fig. 5), except for the down gaze condition. Thus, along the horizontal axis, as with subject 1, left gaze pointing coordinates were significantly shifted to the left, and right gaze pointing locations were significantly shifted to the right of the center. However, along the vertical axis, the pointing location was not affected for the down gaze condition. 

Figure 5

 

Subject 2. Coordinates of pointing locations, colored according to gaze directions. Colored crosses indicate respective median pointing positions for each tested gaze direction, and dashed ellipses indicate confidence intervals (2 SD). Dotted corners show the limits of the screen.

Table 2

 

Subject 2: Finger-Pointing Coordinates Grouped by Gaze Direction Compared to Those Obtained With Gaze Directed to the Center

Pointing Coordinates	Condition 1	Condition 2	Statistics
Horizontal	Right gaze	Gaze to the center	U = 100, n1 = n2 = 10, P = 0.0012
	Left gaze	Gaze to the center	U = 99, n1 = n2 = 10, P = 0.0024
Vertical	Up gaze	Gaze to the center	U = 100, n1 = 8, n2 = 10, P = 0.0005
	Down gaze	Gaze to the center	U = 42, n1 = 5, n2 = 10, P = 0.3996

 Values recorded along the horizontal axis and those recorded along the vertical axis are considered independently. U, Mann-Whitney values; n1, first condition sample size; n2, second condition sample size; P, statistical significance.

Independently of this effect of gaze direction, it appeared that the pointing locations of grouped right, center, and left gaze conditions were significantly shifted downward (when compared to the central target position, Mann-Whitney tests U = 400, n = 31, P < 0.01). In addition, the pointing locations of grouped up, center, and down conditions were significantly shifted rightward (Mann-Whitney tests U = 258, n = 23, P < 0.01, see Fig. 5). 

Subject's Appreciation of His Visual Localization Abilities, and Perceived Gaze Position.

In everyday life, in order to localize an object using the implant-generated image and to then seize that item, subject 2 reportedly first searched the target through rather ample head scans. Then, when the target had been perceived, he performed smaller horizontal head movements to more precisely pinpoint target location. The subject added that he proceeded this way because if he directed his hand as soon as he perceived the object, he would consistently fail to reach it. 

He further indicated that he did not perceive accurately where his eyes were directed, except when he drove his eyes to a very eccentric position, as he then experienced a feeling of intraorbital tension. 

Subject 3

Misalignment of Head and Gaze Directions During Visual Search.

During search head movements, subject 3 showed VOR contraversive ocular movements (Fig. 6; Supplementary Video S3). Moreover, video recording demonstrated the occurrence of camera and eye misalignment at the time the subject detected the stimulus and was about to reach it (Fig. 7; Supplementary Video S4). 

Figure 6

 

Subject 3. Vestibulo-ocular reflex observed following search head movements. (A) Before head rotation. (B) After head rotation to the right, eyes perform a contraversive leftward movement, as shown by the fact that in the left eye, distance (white line) between the limbus and temporal canthus is reduced in (B) compared with (A). Note, however, that VOR gain is less than 1, as in (B), corneal light reflex is slightly shifted to the right compared with (A). Gaze does not exhibit a noteworthy shift when compared to the magnitude of head rotation (see Supplementary Video S3).

Figure 7

 

Subject 3. Camera and right (CC-P fitted) eye are misaligned at the time the subject detected the stimulus and was about to reach it (see Supplementary Video S4).

Effect of Gaze Orientation on Target Localization.

Subject 3 was unable to carry out the requested tasks. During both training and evaluation phases, he was incapable of volitionally dissociating eye and head orientations in a dependable manner. Very rarely, he did succeed in shifting only his eyes laterally, but then his gaze swiftly drifted back to the primary position before he was able to point to the perceived target. 

Subject's Appreciation of His Own Visual Localization Abilities, and Perceived Gaze Position.

Subject 3 was unable to describe any strategy used in his everyday life to reliably reach an object perceived with his prosthetic device. He added that he could not perceive the position of his eyes. 

Discussion

Our results demonstrated that in two of the three subjects fitted with a camera-connected retinal prosthesis, (1) periods of head and gaze misalignment occurred during head-free visual search; (2) the perceptual location of the image was affected by gaze position and, consequently, the conflict between head (i.e., camera) and gaze information affected visuomotor coordination even 4 years after implantation; and (3) adaptive strategies were developed to partly overcome that inconvenience. The third subject (subject 3) was unable to consistently perform some requested tasks and therefore could not be adequately investigated. 

Impact of gaze direction on visual localization has been investigated in healthy individuals. Gauthier and colleagues^14,15 showed that occluding one eye and then deviating that eye induced a hand-pointing error to a visual target seen by the uncovered eye. They noted that the perceived target location was shifted in the direction to which the manipulated eye was deviated. The same authors later obtained similar results by rotating instead of deviating the eye.¹⁶ They suggested that proprioceptors in the deviated eye muscles were responsible for location misjudgment.^14,16 Goodwin and colleagues²⁰ produced an illusion of visual target movement by vibrating extraocular muscles in a sighted individual; and again the phenomenon was ascribed to the activation of proprioceptive receptors, leading the individual to misjudge target location. Similar localization errors have been observed in clinical conditions. Lewis and Zee²¹ reported altered visuospatial perception in a patient with trigeminal-oculomotor synkinesis trying to point to visual targets, a phenomenon also attributed to proprioceptive disturbances. 

An additional mechanism known as efference copy, which takes into account the message generated by the cerebral command to modify gaze direction, is also considered to contribute to perceptual integration of gaze direction, hence to perceptual localization of the image.^22,23 

We therefore assume that in our testing procedure, when asking tested subjects to shift gaze, both proprioceptive information and efference copy contributed to perceived gaze direction and hence to visual localization process. 

During head-free visual search, we distinctly observed periods of camera and gaze misalignment (Figs. 1, 3, 4, 6, 7; Supplementary Videos S1, S2, S3, and S4). At least some of them appeared to be generated by VOR upon head/camera scans. In healthy subjects, VOR is intended to stabilize gaze in space during head rotations, and consequently to stabilize the visual percept of the environment. In spite of its apparent uselessness when visual function is lost, VOR still occurs in individuals with late acquired blindness, although VOR gain is then reduced as compared to the gain in sighted subjects.²⁴ The importance of vestibular function in the process of visuospatial localization is well illustrated by the occurrence, in vestibular disorders, of illusory oscillatory movements of the image (so-called oscillopsia) related to changes of eye position in space.²⁵ 

It was therefore expected that VOR would be observed in our subjects and was conceivable that occurrence of such eye movements could impact visual localization. Obviously, it would be interesting to measure the effects of such reflexive ocular movements on spatial localization. However, quantifying VOR and correlating eye deviation with the degree of bias in spatial perception was not feasible in our study as a result of limitations in available eye-tracking systems. Infrared eye trackers require a calibration procedure that cannot be properly performed in blind individuals,²⁶ and the use of magnetic search coils is not acceptable in subjects wearing a retinal prosthesis, as the conjunctiva of affected individuals is vulnerable.⁶ We therefore decided to video record eye and head movements. This procedure demonstrated that ocular/camera misalignment periods obviously occurred during both target search and finger-pointing preparation, and has shed some light on the nature of the oculomotor mechanisms involved. However, due to the irregularity of head movements and occasional periods of eye masking, for example, by blinks, video recordings did not enable us to determine the proportion of time when misalignment was obvious. 

Subjects' reports on strategies used when searching an object using the CC-P and eventually attempting to seize it were most informative with regard to the presumed effect of contraversive eye movements following head rotation. Indeed, both subjects 1 and 2 indicated that their grasping attempts were consistently misdirected if performed as soon as the object was detected. In such a condition, before taking hold of the seen object, they would perform several back and forth head turns, regressive in amplitude, before attempting to seize the item. 

We therefore believe that in our subjects, ocular rotation—opposite to the search head movement by dissociating eye and head (i.e., camera) directions—provoked a misestimation of the image position. Subsequent regressive back and forth head turns presumably contributed to bringing gaze closer to its primary position, aligned with the camera direction, thus providing more accurate information on the image position in space. It is, however, also conceivable that these to-and-fro head movements were conducted also to optimize delineation of the image perceived in the seeing window offered by the retinal implant,²⁷ as well as to refresh the visual percept generated by the prosthesis, which often swiftly fades upon stable fixation.²⁸ 

Noteworthy observations were additionally made when subjects were first requested to shift gaze to a prescribed direction while keeping their heads stable. 

The cause of the slight global downward and rightward bias observed in our patients is speculative. To optimize perceptual spatial localization of the information provided by the prosthesis, following implantation surgery the patients underwent a rehabilitation procedure consisting of the following two phases. First, a computerized processing developed by the manufacturer (Second Sight Medical Products) allowed aligning the center of the field of view of the camera with the field of view processed by the retinal prosthesis. Second, the patients were trained to keep their gaze in primary position and perform pointing tasks to random visual targets recorded by the head-mounted camera, being informed after each trial by auditory feedback on errors observed. They thus progressively improved accuracy in their visuomotor coordination. The global bias observed in the pointing task conducted in our study possibly reflected a slightly shifted system positioning (e.g., retinal prosthesis position in the retina, camera settings) or a trend of patient's resting gaze position to shift in such conditions. However, importantly, this slight global shift does not question the significance of our observations regarding the effect of gaze/camera misalignment on spatial localization. 

Two of the tested subjects (subjects 1 and 3) were found to be initially incapable of complying with the request, and instead turned both eyes and head toward the indicated direction. After several attempts, however, they both eventually succeeded in voluntarily dissociating gaze and head directions. In contrast, the third individual (subject 2) showed no apparent difficulty in performing, from the start, the requested isolated ocular movement. It is conceivable that the reinforced coupling of eye and head positions expressed, on the one hand, in subjects 1 and 3 by the relative difficulty of voluntarily dissociating them and, on the other hand, in subjects 1 and 2 by to-and-fro regressive head turns following visual search, might represent adaptive processes in the motor control, preventing the occurrence of erroneous image localization. While the former would contribute to preventing the dissociation of eye and head directions, the latter would help in realigning them when dissociated. 

Since implantation was performed 4 years prior to our study, subjects were repeatedly advised to keep their gaze straight ahead, thus avoid dissociating eye and head movements, which was presumed to generate distorted spatial orientation. That training might have contributed to the development of the adaptive process observed here, although our findings showed that the individuals investigated were unable to indicate their eye position, or at best only roughly, and consequently to precisely control gaze direction. The lack of gaze direction control could be another argument to explain the dissociation of gaze and head positions when subjects were about to detect and reach a target during visual search. Moreover, these subjects demonstrated VOR, dissociating eye and head positions in an uncontrolled manner. These findings show the limits of adaptive processes that could be developed by the subjects after 4 years of training. 

When gaze was laterally directed, two of the tested subjects (subjects 1 and 2) were able to maintain eccentric eye position, whereas the remaining one (subject 3) could not prevent his eyes from rapidly drifting back to the primary, straight-ahead position. As a result, perceptual image shift according to gaze changes could not be evaluated in this subject. This observation is in accordance with previous studies reporting that late blinds commonly exhibit difficulties in maintaining an eccentric gaze position.¹⁷ 

Our findings demonstrate, for the first time quantitatively, the importance of eye and camera axis alignment for blind individuals fitted with a camera-connected retinal prosthesis. They also identify conditions that contribute to achieving this alignment, those breaking it, and the adaptive strategies developed to manage situations when information provided on spatial image localization is conflicting, 4 years after implantation. 

This study was conducted with a limited patient population, since the Argus II is an innovative retinal prosthetic device implanted in only few patients in the world. However, these observations can have a significant impact with regard to optimization of the functional rehabilitative procedures of affected individuals. Moreover, in the design of future generations of CC-P, these issues should be addressed. From that perspective, a suggestion has been to develop devices including a head-mounted wide-angle camera showing part of the image contingent to gaze direction, or using a microcamera implanted in the eye.^13,29 

Acknowledgments

The authors thank Alexandre Leseigneur, OD, and Céline Chaumette, OD, for their assistance during test performance; Johan Lebrun, engineer, for computer programming; Gregoire Cosandey, PhD, Avi Caspi, PhD, Jessy Dorn, PhD, Brian Coley, PhD, Mark Wexler, PhD, and William H. Seiple, PhD, for valuable discussion; and Katia Marazova, MD, PhD, and Anne-Fleur Barfuss, PhD, for editorial assistance. 

Presented in part at the annual meeting of the Association for Research in Vision and Ophthalmology, Orlando, Florida, United States, May 2014. 

Supported by French state funds managed by the Agence Nationale de la Recherche (ANR) within the Investissements d'Avenir program (ANR-11-IDEX-0004-02) and by a grant from Humanis. This work was performed within the framework of the Labex LIFESENSES (ANR-10-LABX-65). The authors alone are responsible for the content and writing of the paper. 

Disclosure: N. Sabbah, None; C.N. Authié, None; N. Sanda, None; S. Mohand-Said, None; J.-A. Sahel, Pixium Vision (C), GenSight Biologics (C), Sanofi-Fovea (C), Genesignal (C); A.B. Safran, None