Accommodation responses following contrast adaptation

The current study explored the effects of contrast adaptation on the accommodation response (AR), using low- and high-pass filtered video clips as stimuli. Ten young myopic (mean ± standard deviation: -2.91 ± 1.36D) and 10 near emmetropic subjects (-0.19 ± 0.14D) participated in the study. The AR was monitored under monocular viewing conditions using an eccentric infrared photorefractor. A 2-stage procedure was used: (1) the minimum spatial frequency content necessary to produce a proper individual AR; and (2) the AR was compared before and after adaptation to low-pass (s = -0.5), control (s = 0) and high-pass (s = +0.5) filtered videos. We found that (1) the average threshold Sinc-blur of both myopes and emmetropes necessary to evoke accommodation was (mean ± standard deviation) λ = 7.40 ± 4.05 cpd. Myopes required a higher Sinc blur (average, 10.00 ± 4.05 cpd) compared to emmetropes (average, 4.80 ± 1.60 cpd). (2) Adaptation to low-pass filtered videos increased the AR by 0.41 ± 0.33D in the myopic group and reduced it in the emmetropic group by 0.31 ± 0.25D. Adaptation to high pass-filtered videos induced similar changes in both refractive groups (an increase of 0.41 ± 0.40D and 0.46 ± 0.29D for myopes and emmetropes, respectively). Our measurements show that the human AR can be modified by spatial frequency selective contrast adaptation although these were short-term effects. The perhaps most striking finding was that adaptation to low pass filtered videos had opposite effects on the AR in emmetropes and myopes. It remains to be studied whether these differences were a consequence of myopia or a contributing factor in myopia development.


Introduction
As a result of changes in viewing distance and accommodation tonus, the plane of focus in the retinal image is continuously shifting in depth (Garcia, Ohlendorf, Schaeffel, & Wahl, 2018, Sebastian, Burge, & Geisler, 2015, Sprague, Cooper, Reissier, Yellapragada, & Banks, 2016. Because of the resulting changes in focus, as well as of variations of contrast in different visual scenes, the image contrast at the level of the photoreceptors varies considerably over time (Rucker & Kruger, 2006). To optimize the detection of contrasts, the contrast sensitivity function undergoes continuous adaptation, mainly in the lower spatial frequency range ( (Kohn, 2007, Webster, 2015, see review by Schaeffel (2017)). In addition to contrast adaptation, the visual system adapts continuously to virtually all retinal image features like luminance, color, or motion (Ghosh, Zheleznyak, Barbot, Jung, & Yoon, 2017). The general advantage of adaptation is that the response function is moved into the range were optimal signal detection is possible (Baccus & Meister, 2002, Ghosh et al., 2017, Webster, 2015. Contrast adaptation refers to the continuous change in contrast sensitivity and determines perception of contrast in a visual scene (Baccus & Meister, 2002). Two forms of contrast-related adaptations are known: contrast gain control, which occurs within very short timescales of about 100 ms, and adaptation to contrast that occurs over a larger timescale in the range of 1-10 s (Baccus & Meister, 2002) or even more than 10 s (Chander & Chichilnisky, 2001, Kim & Rieke, 2001, Smirnakis, Berry, Warland, Bialek, & Meister, 1997. This process is closely linked to blur adaptation, which refers to the change in perceived blur after a short period of blur exposure (Vera-Diaz, Woods, & Peli, 2017). The speed of this neural process (Greenlee, Georgeson, Magnussen, & Harris, 1991, Webster, Georgeson, & Webster, 2002, and the correlation between the contrast sensitivity function and adaptation magnitudes are reasons why it is believed that blur adaptation is a direct result of contrast adaptation (Haun & Peli, 2013).
Studies of adaptation and its influence on the accommodation response (AR) have not shown a clear trend. While several authors have previously tackled this question (George & Rosenfield, 2002;Cufflin, Hazel, & Mallen, 2007, Vera-Diaz, Gwiazda, Thorn, & Held, 2004, the evidence remains controversial. For example, (George & Rosenfield, 2002) and Cufflin et al. (2007) found no effect of adaptation to blur on the accommodative response or its gradient. In contrast, Vera-Diaz et al. (2004) reported that adaptation to a Gaussian blur, using 0.2 Bangerter foils, resulted in an altered accommodative response in myopes but not in emmetropes. However, Bangerter foils produce a qualitatively different blur than optical defocus due to the lack of spurious resolution and random phase shifts (Perez, Archer, & Artal, 2010). It is well known that defocus blur is the main trigger to elicit an AR, but also detection of the sign of defocus may further improve the AR. Directional information can be extracted by the visual system using cues from longitudinal chromatic aberrations (Cholewiak, Love, & Banks, 2018), spherical aberration , astigmatism (Ohlendorf, Tabernero, & Schaeffel, 2011), and other higher order monochromatic aberrations (Wilson, Decker, & Roorda, 2002). Furthermore, the principle of trial and error may be implemented, using accommodation microfluctuations (Charman & Heron, 1988). However, all these cues are limited by the optical and neuronal depth of focus of the eye (Held, Cooper, O'Brien, & Banks, 2010).
To gain more insight into the role of contrast adaptation on AR and its plasticity, we studied accommodation to spatially filtered videos. Furthermore, we explored potential changes in accommodation when subjects were pre-adapted to contrast by extended viewing of low-pass or high-pass filtered video clips.

Subjects
Twenty young adult (range: 21 to 29 years; mean age: 25.45 ± 2.64 years), ten myopic (mean age: 25.20 ± 2.89, mean spherical equivalent refraction: −2.91 ± 1.36D) and ten emmetropic subjects (mean age: 25.70 ± 2.33, mean spherical equivalent refraction: −0.19 ± 0.14D) took part in the experiment. Subjects with myopic spherical refractive error > 6.0D, astigmatism > 2.0D, accommodation disorders, active ocular pathologies or simultaneous participation in other experimental trials were excluded from the study. All participants had uncorrected or corrected visual acuities of 0.0 logMAR or better. The experiment followed the tenets of the declaration of Helsinki of 1964 and approval from the ethical board committee of the University of Tübingen was obtained for this investigation. Informed consent was collected from all subjects after indications and potential consequences of the study had been explained in detail.

Pre-experiment protocol
During the first visit, the pre-experiment protocol of all subjects included an optometric examination and individual calibration of the photorefractor, since accommodation response was measured by means of a customized eccentric photorefractor and an individual calibration was required to reduce errors below 0.25D. As previously described in literature, photorefraction calibration involves the calculation of an individual conversion factor and an individual offset, using ophthalmic trial lenses and dynamic streak retinoscopy (Schaeffel, Wilhelm, & Zrenner, 1993, Seidemann & Schaeffel, 2003. Optometric examination included the following procedures: uncorrected distance visual acuity (UDVA), corrected distance visual acuity (CDVA) and manifest refraction (objective: i.Profiler®plus, Carl Zeiss Vision GmbH, Aalen, Germany; subjective: ZEISS Visuphor 500®, Carl Zeiss Vision GmbH, Aalen, Germany). Subjective refraction was used to cluster refractive groups: myopia was defined as spherical equivalent ≤−0.50D, emmetropia as ≥−0.50D to +0.50D and hyperopia as > +0.50D (Flitcroft et al., 2019).

Stimuli
Visual stimuli were video clips from an open source movie, whose blur and contrast were manipulated digitally in the Fourier domain by filtering with a Sinc function and modifying the slope (s) of the amplitude spectrum, respectively. Two video clips were extracted with a duration of 5 s (used in Phase (1) and (2) of the experiment) or 60 s (used in Phase (2) as adaptor). Each of the frames were extracted and processed digitally, involving a total of 1560 individual frames with a format of 1280x720 pixels. Color information was removed, and image size was measured in terms of degrees of visual angle, derived from viewing distance and size of the display. To remove high spatial frequency components originating from sharp edges, frame margins were smoothened by means of a Sinc window as described before (Sanz Diez, Ohlendorf, Schaeffel, & Wahl, 2019). Sinc window and image frames were equated and therefore no oscillation effects were detectable at the edges. Applying a Sinc window had the same effect as a Gaussian or Hamming window where intensity was maximum in the center and progressively reduced towards the periphery where it reached zero values.
As described by Sanz Diez et al. (2019), blur levels were manipulated using a Sinc function. Six levels of lambda were tested: 1, 2, 4, 8, 16, 32 cycles per degree (cpd). Contrast was manipulated in a second step by modifying the slope of the amplitude spectrum of the image frame (Webster et al., 2002). The original amplitude spectrum was multiplied by f s using 3 different slopes (s): −0.5/0.0/+0.5, to enhance or reduce the sharpness of the original image. Consequently, with negative slopes the amplitudes of the spatial frequency spectrum declined more pronounced towards high spatial frequencies and, with positive slopes, the spatial frequency spectrum became flatter, enhancing amplitudes at higher spatial frequencies (Murray & Bex, 2010, Vera-Diaz, Woods, & Peli, 2010, Webster et al., 2002. As proposed by Webster et al. (2002), as well as by Vera-Diaz et al. (2010), filtered images were adjusted to contain the same root mean square contrast and luminance.
Finally, digitally manipulated frames were converted into a video file. MATLAB R2016a (The MathWorks, Massachusetts, United States) and Psychtoolbox-3 (Brainard, 1997, Pelli, 1997 were used for video and frame processing.

Setup
A liquid crystal display (LCD) was used to present the video stimuli at a screen with a resolution of 2048x1536 pixels and a refresh rate of 60 Hz. A forehead and chin rest were used to control head position and keep the viewing distance at 50 cm where the visual stimulus subtended a vertical and horizonal visual angle of 8°. Image size was adjusted to account for individual differences in spectacle magnification for each myopic subject. An eccentric infrared photorefractor (Camera DMK 23UM021, Cosmica Pentax 50 mm lens fixed to 1 m distance) was positioned at 100 cm distance from the subject to measure the AR. Sampling rate was 60 Hz. Under the use of a neutral density filter (1.2 ND filter, LEE Filters, United Kingdom), the mean luminance value was 6.05 ± 0.16 cd/m 2 at the end points of the lambda levels used.

Procedures
Experiments consisted of two phases ( Fig. 1): In phase (1) the minimum spatial frequency content necessary to induce a proper AR in each of the subjects was determined. As stated above Sinc-blur lambda was modified in six different steps, from maximum to minimum blur and AR was continuously recorded for 5 s under each condition. To avoid adaptation effects, each stimulus was followed by a blank interval of 100 ms (Coltheart, 1980). As in previous studies (Sanz Diez et al., 2019), a change in refraction of 0.30D was set as threshold for a significant change in AR. A Lambda of 1 cpd was defined as baseline (Fig. 1).
In phase (2), the AR was measured before, during and after adaptation to low-pass (s = −0.5), control (s = 0.0) and high-pass (s = +0.5) filtered video clips. The adaptation period lasted 60 s.
Before and after the adaptation period, a 5-second video clip with the amount of lambda calculated in the first part of the experiment was shown. This process was repeated under each of the three contrast slopes conditions (s = −0.5/0.0/+0.5) and in a randomized order. A comparison between AR before and after adaptation to contrast was conducted (Fig. 1).

Statistics
A Kolmogorov-Smirnov test was employed to assess normality of the data. Statistical analyses were done using the appropriate tests (Student t test or Wilcoxon-signed rank test). Differences were considered statistically significant when the associated p-values were lower than 0.05. Results are provided as mean ± standard deviation. All analyses were performed using MATLAB R2016a and SPSS 22.0 software.

Phase (1) -AR and Sinc-blur filtered videos
The inter-individual variability of the amount of Sinc-blur needed to obtain an accurate AR was analyzed in Phase 1. On average, a Sinc-blur of λ = 7.40 ± 4.05 cpd (median: 8 cpd, interquartile range: 4-8 cpd) was necessary to stimulate accommodation. In Phase 1, the individual spherical equivalent refraction error and the individual amounts of lambda were significantly correlated (Pearson correlation coefficient 0.89, covariance = −6.36, R 2 = 0.79, 95% CI −0.96 to −0.73; p < 0.001), showing that subjects with higher myopic refractive errors needed higher amounts of spatial frequency to reach similar AR values.

Phase (2)effect of contrast adaptation on AR
Mean accommodation responses (ARs) before, during and after contrast adaptation under the three conditions are shown in Fig. 3. ARs changed differently in both refractive groups, significantly after adaptation was induced by watching low-pass filtered videos. The AR increased by 0.41 ± 0.33D in the myopic group but an inverse trend was found in emmetropes with a reduction of the AR by 0.31 ± 0.25D (p < 0.01 for both groups, Wilcoxon test). After adaptation to highpass videos, both groups showed similar increases in the AR (0.41 ± 0.40D and 0.46 ± 0.29D for myopes and emmetropes, respectively; p < 0.01 for both groups, Wilcoxon test). Under the control conditions without adaptation, the AR remained unchanged in both refractive groups (p = 0.816 for myopes and p = 0.474 for emmetropes, Wilcoxon test).

Discussion
We have analyzed the effects of contrast adaptation on the AR in young adult human subjects. Two major findings were made: (1) a mean Sinc-blur level of λ = 7.40 ± 4.05 cpd was required to induce a significant AR; interestingly, emmetropic subjects required less high spatial frequency content than myopes to achieve similar AR to myopes, (2) myopes displayed an increased AR after 60 s exposure to a low-pass filtered video, while emmetropes displayed a reduced AR.

AR and blur sensitivity
Our observations are consistent with previous studies which conclude that myopes interpret and adapt to blur differently than emmetropes (Rosenfield & Abraham-Cohen, 1999, Strang, Winn, & Bradley, 1998, Turatto et al., 1999. However, there is still controversy about the origins of these differences, and how they affect the AR. Studies agree that myopes show less sensitivity to blur (Rosenfield & Abraham-Cohen, 1999), compared to emmetropes. Since uncorrected myopes are continuously adapted to defocus (Rosenfield, Hong, & George, 2004), their contrast sensitivity may be affected especially at high spatial frequencies. Alternatively, contrast sensitivity can be altered by intrinsic changes in the retina generated during myopia development, which can lead to inaccurate processing and sensitivity of high spatial frequencies. Reduced blur sensitivity in myopes could Fig. 2. Raw accommodation traces collected during Phase (1) of the experiments for both emmetropes (left, n = 10) and myopes (right, n = 10). The abscissa shows Sincblur levels (λ) in cycles per degree. Accommodation responses are presented on the ordinate in diopters. Blue thick lines denote averages of accommodation responses of all subjects in each group. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Fig. 3. Mean ARs during Phase (2) of the experiments in both refractive groups (emmetropes -upper graphs, myopes -lower graphs). The three graphs on the right show the changes in ARs before (PRE) and after (POST) contrast adaptation. Ordinates show ARs in diopters. Abscissae show adaptational stages. Error bars represent standard deviations. Red, black and green traces and markers denote ARs during adaptation to low-pass filtered (s = -0.5), control (s = 0) or high-pass filtered (s = +0.5) stimuli, respectively. The dashed lines delimit the time window where contrast adaptation occurred. P values: p < 0.05 (*); p < 0.01 (**); p < 0.001 (***). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) P. Sanz Diez, et al. Vision Research 170 (2020) 12-17 explain why they were not very responsive to an increase in the low to mid spatial frequency range in the videos used in Phase (1) of the experiments and why they needed more energy in the high spatial frequency band for a comparable AR. Ciuffreda (1991) described that optimal ARs require a stationary target with high contrast. In general, if a fixated target is out of focus, causing a blurred and low contrast image in the fovea, an AR is elicited to optimize the detection of fine details (Heath, 1956). Furthermore, an optimal ARs can be expected only under closed loop conditions (Grosvenor, 2007). In our experiments, the feedback was partially opened due to the use of digitally blurred targets. Here, we found that energy must be available in the mid-range of spatial frequencies to elicit an appropriate AR.

AR following contrast adaptation
Visual adaptation to a variety of visual features has been extensively studied (Webster, 2015). It refers to a temporary change in sensitivity to a certain visual aspect to keep normalizing visual experience and to maximize sensitivity (Kohn, 2007, Webster, 2015. In the current study, Phase (2) showed that myopes accommodated differently than emmetropes during and after adaptation to a low-pass filtered video. While the AR of myopes increased over time, becoming more accurate, emmetropes displayed a reduced AR. Since myopes required higher spatial frequency content for appropriate ARs (as shown in Phase 1), it could be that their accommodation is more controlled by higher spatial frequencies than in emmetropes who are more dependent on energy in the mid spatial frequency range. After they had adapted to the mid spatial frequencies, their sensitivity in this spatial frequency range was reduced, partially removing critical information to drive their AR (as shown in Phase 2).
There were other studies on this topic before, where the effects of blur adaptation on the AR in emmetropes and myopes were studied in detail (Cufflin et al., 2007, Vera-Diaz et al., 2004, George & Rosenfield, 2002. Two studies did not observe any effects of contrast adaptation on accommodation (Cufflin et al., 2007, George & Rosenfield, 2002. George and Rosenfield (2002) used + 2.5D lenses (in addition to the subjects' distance correction), imposed for a 3-hour adaptation period. Accommodation targets ranged from −2.5 to −5.0D. Cufflin et al. (2007) used + 1.0D and + 3.0D imposed defocus and a period of 45 min of adaptation and presented accommodation targets at 6 different distance vergences (0.0D, −1.0D, −2.0D, −3.0D, −4.0D and −4.5D). In contrast, Vera-Diaz et al. (2004) found that Gaussian blur (induced by 0.2 Bangerter foils, reducing stimulus contrast by about 75% (Perez et al., 2010)) increased ARs in myopes, but not in emmetropes. Our findings are comparable to those of Vera-Diaz et al. (2004). In their studies, the AR increased by 0.29D in myopes after 3 min of adaptation when subjects were looking at a target at 0.33 m distance. Different from Vera-Diaz et al. (2004), we modified the slope of the amplitude spectrum (Webster et al., 2002) and used Sinc-blur digitally filtered videos (Murray & Bex, 2010). The AR in myopes increased by 0.41D after 60 sec of adaptation but declined by 0.31D in emmetropes, while Vera-Diaz et al. (2004) reported an increase by 0.06 D also in emmetropes. It can be that the different procedures to induce contrast adaptation can account for the differences in the emmetropic group. It is important to note that the phase spectra might be different in both cases, as explained by Murray and Bex (2010).
In summary our study provides further evidence that the mid-range spatial frequency spectrum is very important in driving the AR and that contrast adaptation in this spatial frequency range can have short-term effects on the AR. Our findings demonstrate that accommodation can be modified and improved by contrast adaptation. Perhaps pre-processing of visual stimuli to optimize contrast in the mid-range of the spatial frequency spectrum could be used to reduce the lag of accommodation, for instance in virtual reality systems with augmented reality displays.

Timescale of contrast adaptation aftereffect
Following adaptation, subsequent effects on the response can be observed over a wide time scale, from milliseconds to several seconds later. As a rule, a prolonged adaptation phase produces stronger aftereffects (Kohn, 2007), although this general rule is not always satisfied (Priebe, Churchland, & Lisberger, 2002). In regard to contrast adaptation, the time course of the aftereffects has been analyzed in several studies related to visual acuity, contrast sensitivity and accommodative response. Ohlendorf and Schaeffel (2009) observed that contrast sensitivity had increased by about 30 percent after the adaptation imposed by myopic defocus. This effect remained for 3 min after a 10-minute adaptation period. 5 min were necessary for the contrast sensitivity to return to its baseline levels. Studies on visual acuity showed that the improvement observed after blur adaptation persisted for a period of 4 min (Khan, Dawson, Mankowska, Cufflin, & Mallen, 2013). Furthermore, adaptation processes related to spectral composition of light or color, have shown prolonged post-adaptation effects (Delahunt, Webster, Ma, & Werner, 2004, Neitz, Carroll, Yamauchi, Neitz, & Williams, 2002. Vera-Diaz et al. (2004) reported a progressive increase in accommodative response during a 3-minute blur adaptation period. This effect continued for a 2 min-period until it returned to the initial level. It may be that the type of measurement used after adaptation influenced the duration of the adaptation effect. In our study we used an adaptation period of 60 s and a post-adaptation period of 5 s. Post-adaptation time length could be considered as a study limitation since the maximum duration of the observed effect was not studied. Further experiments under different post-adaptation time periods will be required in order to obtain more information about the time course of the observed adaptation effect on the accommodative response.

Limitation of the study
The optical transfer function determines how contrast at different spatial frequencies is transmitted by the optical system. In different individuals, spatial frequency and contrast distributions are mainly determined by uncorrected higher order aberrations, lag of accommodation and pupil diameter. In the current experiments, pupil sizes were very stable across subjects and experiments [phase (1): 7.18 ± 0.49 mm and 7.20 ± 0.66 mm for emmetropes and myopes, respectively; phase (2): emmetropes (s = −0.5: 7.19 ± 0.50 mm; s = 0.0: 7.19 ± 0.49 mm; s = +0.5: 7.19 ± 0.49 mm) and myopes (s = −0.5: 7.17 ± 0.67 mm; s = 0.0: 7.18 ± 0.65 mm; s = +0.5: 7.17 ± 0.66 mm)]. However, the lag of accommodation varied, as described earlier. A demanding way to cope with these potential confounders would have been to measure pupil diameter, lag of accommodation and higher order ocular aberrations in real time and adjust the spatial frequency spectrum in a closed feedback loop. If spatial frequency spectra were controlled individually, the impact of these confounding variables would have been reduced.

Conclusions
We have studied the short-term effects of contrast adaptation on accommodation and the effects of refractive state. While adaptation clearly modified the accommodation amplitudes, the most striking finding was that it affected accommodation differently in emmetropes and myopes. These results suggest that contrast at different spatial frequencies may be differently weighted by the accommodation system in the two refractive groups. Whether these differences were a consequence of existing myopia or were present from the beginning and contributed to the development of myopia needs to be determined in future studies.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.