Neural adaptation to the eye’s optics.

The eye’s optics degrade visual inputs transmitted to the retina in two ways: it reduces image contrast and can disrupt the phase (position) of visual signals relative to each other. Although there is evidence that neural adaptation to optical blur helps partially counteract blur-induced contrast reductions, it is unknown whether the visual system can compensate for phase disruptions. Given the importance of phase information in visual recognition and image quality, phase compensation may play a vital role in the brain’s ability to adapt to the eye’s optics to improve perceptual quality. Left panel: Visual inputs corresponding to a high-contrast acuity letter (20/20) and a grayscale image of a spider. Middle panel: Simulated retinal images using the habitual optical quality of a patient with moderate keratoconus–a corneal disease causing severe optical aberrations. Right panels: Simulated images illustrating the benefits of contrast compensation and phase compensation, respectively. Both mechanisms would be needed to optimally enhance perceived image quality.

Perceptual importance of phase information.

(A) Any image can be decomposed into its amplitude spectrum and phase spectrum. The phase spectrum, rather than the amplitude spectrum, contains most of perceptually-relevant information about image structure and is vital for visual recognition. (B) Reconstructed images reflect the phase spectrum, rather than the amplitude spectrum, of original images.

Experimental approach

(A) Schematic illustration of the AO vision simulator. AO correction allows us to bypass optical factors and compare visual functions under similar, fully-corrected optical quality in participants otherwise exposed to varying amounts of optical aberrations. Optical aberrations are measured in real-time by a Shack-Hartmann wavefront sensor. During closed-control loop, a deformable mirror maintains a desired aberration pattern by compensating for the participant’s wavefront aberrations. Wavefront maps and simulated acuity letter acuity are showed for participants with typical optical quality and moderate keratoconus. (B) Compound stimuli consisted of two suprathreshold horizontal gratings with frequencies f and 2f. The appearance of compound stimuli depends on the f-2f relative phase, going from squarewave-like (upper panel) to sawtooth-like stimuli whose luminance peaks points either more “downward” (lower left) or more “upward” (lower right). (C) Perceived phase under full AO-correction in participants with typical optical quality (Exp.3, N=11). Perceived phase was measured by computing the percent of “upward” responses as a function of relative phase, and estimating the point-of-subjective equality (PSE)–i.e., the relative phase at which stimuli appeared aligned. Data were fit with a Cumulative Normal function to compute the PSE (+1.89º; vertical dashed line) with bootstrapped 95%-confidence intervals (95%-CI: [-0.7º +4.4º]; shaded gray area). All participants were tested across ±90º in relative phase, but the exact range of values used across participants in Exp.3 slightly varied (see Methods), as indicated by the size of each data point in this example figure. Error bars correspond to ±1SEM.

Perceived phase shifts with brief exposure to vertical coma.

(A) Trial sequence. Participants judged the appearance of compound gratings under specific amounts of AO-induced vertical coma. (B) Optical theory predictions. We predicted the magnitude and direction of f-2f relative phase shifts for varying amounts of coma and different f frequencies. Dashed area indicates conditions selected for testing in (D-F). (C) Setup validation. We validated the predictions before testing directly from the visual display by capturing images of gratings under AO-induced vertical coma. (D) Perceived phase shifts. Brief exposure to vertical coma resulted in PSE shifts matching optical theory predictions, except at ±0.5µm. (E) Impact of reduced contrast on perceived phase. The attenuation of PSE shifts under ±0.5µm of coma was replicated under AO correction using MTF-adjusted stimuli with ±52.3º phase offset. (F) Perceived PSE estimates plotted relative to optical theory predictions. Individual PSEs correspond to filled colored symbols, while open black circles show group-average PSEs with 95%-CI error bars. As showed in (F), individual participant data were consistent with group-average data plotted in (D,E). Results in (D,E,F) are for compound gratings with a fundamental frequency of 3 cycles/deg (see Fig.S2 for additional results using f: 4 cycles/deg).

Short-term adaptation to vertical coma.

(A) Each experimental session consisted of 41 blocks divided into 3 segments: baseline (pre-adaptation), adaptation, and post-adaptation. Both pre- and post-adaptation were measured under AO correction (PSEpredicted:0º). During blur adaptation, -0.4µm of vertical coma was AO-induced (PSEpredicted:+29.5º). After the first 3 blocks of adaptation (indicated by the * symbol), we started presenting grayscale natural images and checkerboards before each compound grating, serving as cues the visual system could use to detect and adapt to blur. In a control condition, AO correction was maintained during the entire session and a phase offset (+29.5º) was added instead, thus mimicking the impact of vertical coma on perceived phase without observers being exposed to blur. Breaks were allowed at specific time points, as indicated by vertical lines. (B) Trial sequence during the adaptation (and control) segment. (C) Examples of adaptation stimuli. (D–F) Blur adaptation. Before adaptation, psychometric functions and PSE estimates were centered near 0º. Blur exposure resulted in a PSE shift that initially matched optical theory predictions (+29.5º), but then decreased over time. As soon as blur was removed (post-adaptation), an aftereffect in the opposite direction was observed. (G–I) Control experiment. In the absence of blur, PSE shifts matched the added phase offset (+29.5º) and remained stable over time, with post-adaptation PSEs returning immediately to baseline. (D,G) Psychometric curves fitted to group-average data. (E,H) Group-average (filled circles) and individual (open circles) PSE estimates, with solid lines corresponding to power functions fitted to group-average PSEs. To better visualize changes in PSE over time, PSE estimates are adjusted for small biases observed during baseline, thus treating 0º as baseline. (F,I) Schematic representations of perceived phase over time observed in both experimental conditions.

Altered phase congruency following long-term adaptation to poor optical quality.

(A) Relative to typical ‘healthy’ eyes, KC eyes are affected by large amounts of optical aberrations, as illustrated by the wavefront maps. As a result, the visual system of KC patients is constantly exposed to degraded retinal images, as illustrated by simulated acuity letters. AO allows to fully correct all optical aberrations while assessing visual functions, even in KC eyes. (B) Under full AO correction, participants with healthy eyes showed PSEs tightly distributed around 0º, as expected. In contrast, KC patients showed larger PSE shifts, consistent with altered phase congruency. Each data point corresponds to PSE estimates (x-axis) of individual participants (y-axis) plotted with 95%-CIs (see also Fig.S5). (C) The magnitude of the PSE shift correlated with the amount of habitual optical aberrations (e.g., higher-order aberrations–hRMS) that each participant is exposed to in their daily life (see also Fig.S6).