What limits visual acuity in retinal implants despite electrode design

Retinal implants promised bionic sight, but the numbers never added up. Devices packed with hundreds or even thousands of electrodes deliver vision that falls dramatically short of what their physical design should allow. A new review published in Biomedical Engineering Letters now dissects this “resolution gap” and argues that the future of artificial vision depends not on cramming more electrodes into smaller spaces, but on solving a web of intertwined biological and biophysical problems that geometry alone cannot fix.

The review, led by Seoyoung Hwang, Hee Soo Jeong, and Sang Beom Jun of Ewha Womans University in Seoul, examines why retinal prostheses—devices that electrically stimulate surviving retinal neurons in patients blinded by photoreceptor degeneration—consistently deliver visual acuity far below the theoretical limits implied by their electrode spacing. The work synthesizes two decades of clinical and engineering data across the major implantation paradigms: epiretinal, subretinal, and suprachoroidal devices.

The biological foundation for these devices rests on a fortunate quirk of retinal degeneration. Diseases such as retinitis pigmentosa and age-related macular degeneration destroy the rods and cones that convert light into neural signals, but the inner retinal circuitry—particularly bipolar cells and retinal ganglion cells, the retina’s output neurons—remains substantially preserved even in advanced disease. This preservation means that electrical stimulation delivered directly to these surviving neurons can, in principle, bypass the lost photoreceptors entirely and reanimate the visual pathway.

Clinical results have validated the concept, though only up to a point. The Argus II epiretinal system, which received both European CE mark approval and FDA Humanitarian Device Exemption approval in the United States, demonstrated that patients with profound blindness could detect large objects, discriminate motion direction, and perform basic orientation tasks. Subretinal implants such as Alpha IMS pushed further, enabling letter recognition and rudimentary form vision in selected patients. More recently, the photovoltaic PRIMA system showed that patterned near-infrared light projected from external glasses could be converted into localized retinal stimulation, restoring measurable central vision in patients with geographic atrophy secondary to AMD.

Yet the acuity numbers tell a sobering story. The 60-electrode Argus II achieves roughly 20/1260 vision at best—a level at which a patient must stand 20 feet from a target that a normally sighted person could read from 1,260 feet. The Alpha IMS achieved Landolt C acuity up to 20/546 in some patients. The PRIMA system produced acuity in the range of 20/460 to 20/550 in patients with centrally placed implants. Normal vision, by comparison, is 20/20. Even the best prosthetic results correspond to a minimum resolvable angle more than twenty times coarser than natural sight.

The puzzle deepens when theoretical predictions are calculated from device geometry. The Argus II array has electrodes spaced approximately 575 micrometers apart, which translates to roughly 2 degrees of visual angle on the retina, since about 288 micrometers of retinal distance corresponds to 1 degree of visual angle. Because at least two adjacent stimulation sites are needed to represent one bright-dark grating cycle, the theoretical resolution limit works out to about 4 degrees per cycle. Subretinal systems such as Alpha IMS and Alpha AMS use far denser pixel architectures, with pixel pitches of just 70 micrometers, predicting theoretical acuities of approximately 20/250 to 20/280. Clinical performance falls well short of these numbers in nearly every case.

The review identifies current spread within retinal tissue as perhaps the single most important biophysical culprit. When an electrode delivers electrical charge, the resulting electric field does not remain confined to the geometric boundary of the metal contact. Current permeates the conductive retinal environment, activating neurons well beyond the intended target region. If the activation field of one electrode overlaps that of its neighbors, adjacent electrodes cannot function as independent visual channels, no matter how tightly they are packed. The functional benefit of a denser array therefore hinges entirely on whether each electrode can produce spatially separable patterns of neural activation—a condition that current spread routinely violates.

Electrode size introduces a second constraint with an unforgiving trade-off. Smaller electrodes offer better geometric confinement of stimulation, but shrinking their area limits the charge that can be safely delivered through them because the available electrochemical surface area shrinks as well. A recent study comparing 10, 20, and 30 micrometer planar iridium oxide electrodes at 50 micrometer pitch found that smaller electrodes exhibited lower stimulation thresholds, while larger ones provided broader usable dynamic ranges within safe charge-injection limits and recruited more ganglion cells. Lower threshold alone does not define an optimal electrode; reliable charge delivery over a wide operating range matters just as much. There is, in effect, a practical lower limit to planar electrode size below which miniaturization stops translating into better functional stimulation.

The distance between electrode and retina compounds these problems. Closer apposition lowers stimulation thresholds and localizes charge delivery, while greater separation demands higher currents that broaden the activated region. In Argus II patients, perceptual thresholds correlated with electrode-retina distance and interfacial properties, and patient-specific computational models confirmed that distance and impedance contribute significantly to threshold variability. The problem is dynamic as well as static: fibrosis, chronic micromotion, and geometric mismatch between planar arrays and the curved retinal surface can all degrade the interface over time, turning initially favorable coupling into an unstable one.

Epiretinal devices face an additional complication: unintended activation of passing ganglion cell axons. Because these devices sit adjacent to the ganglion cell layer, their electrical fields can sweep up axons en route to the optic nerve, producing elongated or distorted phosphenes—the percepts of light that patients report. This weakens the correspondence between electrode position and perceived visual location, degrading spatial specificity even when the array itself is well-designed.

Stimulation parameters add yet another layer of complexity. Pulse waveform, phase duration, and frequency all shape not only thresholds but also the mode of neural recruitment—whether direct ganglion cell activation or network-mediated recruitment through residual retinal circuitry—and the temporal stability of the resulting percept. Longer pulse durations reduce required amplitude but shift the balance between direct and indirect activation pathways. Frequency governs whether a percept appears sustained or fragmented. Argus II operates at 3 to 60 Hz with phase durations of 0.46 or 0.97 milliseconds, while PRIMA’s photovoltaic protocol spans 0.7 to 9.8 milliseconds per phase, illustrating how divergent the paradigms have become across platforms.

The review’s authors argue that these factors interact synergistically rather than independently. Increased electrode-retina distance requires higher amplitude, which broadens current spread, which reduces spatial independence, which defeats the purpose of denser arrays. Similarly, degeneration-associated retinal remodeling elevates thresholds, forcing currents up and activation regions wider, again negating geometric gains from miniaturization.

The clinical landscape reinforces the technical message. Argus II was ultimately discontinued commercially despite its regulatory approvals, illustrating the gap between technical feasibility and sustained clinical deployment. Alpha AMS received CE mark approval but never achieved widespread global adoption. The PRIMA system remains in clinical trials rather than routine use, though recent multicenter data confirmed improved visual acuity over 12 months in patients with AMD-triggered geographic atrophy. Suprachoroidal devices, placed in the space between sclera and choroid, offer surgical advantages but sacrifice spatial selectivity to their greater distance from target neurons.

What emerges from this comprehensive analysis is a reframing of the field’s central challenge. The bottleneck is not electrode density but the functional independence of stimulation sites. Electrode spacing, the authors conclude, must be treated as an important yet imperfect predictor of prosthetic vision—never a direct surrogate for clinical acuity. Next-generation devices will need selective neural recruitment strategies, stable electrode-retina coupling maintained through flexible and conformal substrates, materials with higher charge-injection capacity, stimulation waveforms tuned to the specific retinal circuitry being targeted, and standardized outcome measures that allow honest comparison across platforms.

For the roughly 170 million people worldwide affected by degenerative retinal disease, the resolution gap is more than an academic curiosity—it is the difference between perceiving scattered points of light and reading a page. The Ewha team’s synthesis makes clear that closing that gap demands coordinated innovation across every layer of the system, from electrode material to stimulation algorithm, rather than any single breakthrough in one parameter.