It is well known that the human eye suffers from monochromatic aberrations in addition to defocus and astigmatism and that these aberrations degrade the retinal image and consequently limit the visual performance1. Adaptive optics (AO), recently used in the study of the human eye, enables the measurement and the correction of optical aberrations, thus permitting the evaluation of the visual benefit (VB) of correcting higher-order aberrations.

Previous measurements of the impact of the correction of higher-order aberrations on contrast sensitivity and visual acuity have been performed. In terms of contrast sensitivity, Liang et al.2 measured a 6-fold improvement at 27.5 c/deg on 2 subjects, across a 6mm pupil and in monochromatic light, whereas, Yoon and Williams3 measured at 16 c/deg, on 2 subjects and with the same pupil size, an improvement of contrast sensitivity by a factor of 1.9 when only monochromatic aberrations were corrected in polychromatic light. Yoon et al. attempted to predict the VB by calculating the ratio of MTFs computed with and without HO aberrations, they found at 16 c/deg a theoretical polychromatic VB of 2.8, which is quite different from their 1.9 measured VB. In terms of visual acuity, Yoon and Williams3 reported, on 7 subjects across a 6 mm pupil, a 1.2-fold and a 1.4-fold improvement after correcting the monochromatic aberrations in polychromatic light respectively at 575 Td (Troland, unit of retinal illuminance) and 57 Td. Poonja et al.4 measured on 6 subjects, using an Adaptive Optics Scanning Laser Ophtalmoscope (AOSLO) (e.g., in monochromatic light) and a pupil diameter of 5.89mm, a VB of 1.5 with an averaged visual acuity of 0.85 and 0.57 respectively before and after correction of the monochromatic aberrations.

Customized contact lenses, intraocular lenses and refractive surgery are now being developed to correct higher-order aberrations. The benefit from AO correction depends on the pupil size and the initial higher-order aberrations present in the eye. To address this issue, Guirao at al.5 calculated the visual benefit of correcting the monochromatic aberrations of 218 normal eyes and found great variation among eyes with some normal eyes showing almost no benefit and others a benefit higher than 4 at 16 c/deg across a 5.7 mm pupil, the average VB being 2.4.

The visual benefit should also depend on the ametropia6 since cortical and/or retinal factors are involved in the AO-corrected limit of resolution. Rossi et al.6 measured the VA using an AOSLO, in 10 low myopes and 9 emmetropes, the impact of correcting monochromatic aberrations. The myopic group improved from a mean of 0.83’ in the non-AO condition to a mean of 0.61’ with AO (e.g. a VB of 1.37). The emmetropes improved from a mean of 0.8’ without AO to a mean of 0.49’ with AO (e.g. a VB of 1. 64). Retinal and/or cortical factors could have limited the VA in low myopes after AO correction resulting in a lower VB in this population.

Recent studies7,8 have suggested that we might be adapted to our retinal image, which would have an impact on the visual benefit of correcting the eye's aberrations. Artal et al.7 wondered whether the visual system was adapted to the retinal image of its own eye; in other words, is the best overall subjective image quality obtained when a perfect retinal image is presented on the retina or when the retinal image is blurred by the aberrations of the eye's optics, indicating that the neural system is adapted to the particular pattern of its own eyes, and is able to compensate for the effects of this blur. Theirs results support the hypothesis of a neural compensation for the aberrations. Moreover, several studies9-12 demonstrated improvements in visual acuity after a period during which the subjects had been blurred by a defocus. Mon-Williams et al.9 measured, on 15 emmetropic subjects, an average visual acuity increase of 0.1 logMAR after a 30min wearing period of a +1.00 D defocus. George and Rosenfield10 found, on 13 emmetropic and 18 myopic subjects, an average visual acuity increase of 0.2 logMAR after a 2hours wearing period of a +2.50 D defocus. Rosenfield et al.11 measured on 22 slightly myopic subjects an increase of visual acuity of 0.23 logMAR after 3 hours without compensation. Cufflin et al.12 found after blur adaptation of 45 minutes, on 11 emmetropes and 11 early-onset myopes, an improvement in visual acuity of 0.17 logMAR and 0.23 logMAR following adaptation to +1 D and +3 D of defocus respectively. They also noted that acuity changes became significant after 30 minutes of exposure to optical defocus. These experiments show the effect of neural adaptation to a blurred retinal image on objective visual performance as well as on subjective responses. On the contrary, Pesudovs13 did not find any evidence for an adaptation to surgically induced blur until ten weeks after LASIK. In conclusion, these experiments showed that visual acuity improved following a period of adaptation to blur and, consequently, that the optics alone could not explain the changes in performances without assuming that some neural adjustments were interfering with the optics. In this case, a metric derived from wave aberrations measurements might not be able to predict the visual benefit of correcting the higher-order aberrations. However, it is not evident that the situation will be similar with another visual task such as contrast sensitivity.

Previous authors14-18 studied the link between metrics of image quality and visual performances or subjective image quality. The first objective was to determine the accuracy and precision of metrics for predicting the results of conventional sphero-cylindrical refraction from wavefront aberrations14,18. Guirao et al.14 found that image-plane metrics such as the volume under the MTF between 0 and 60 c/deg were able to predict the subjective refraction of the eye. The mean error between predicted and subjective refraction was about 0.1±0.08 D averaged across 6 eyes of 6 normal subjects whereas their five calculated image-plane metrics did not differ by more than 0.03 D. Similarly, Thibos et al.18 compared 33 metrics calculations to subjective refractions performed to the nearest 0.25 D on 200 normal and healthy eyes from 100 subjects. All the mean predicted values varied from −0.50 D to +0.25 D. The variability between various metrics was always lower than the ± 0.75 D test-retest variability in the measurement of refractive error19,20, meaning that the accuracy of metrics to predict subjective image quality should not be assessed this way. Marsack et al.16 used the previous data set to investigate the ability of 31 metrics derived from wave aberrations maps to predict changes in high-contrast logMAR acuity. The visual acuity loss were measured on simulated aberrated charts generated with a commercially available program called CTView© that introduces the aberrations into the charts by directly setting the wavefront Zernike coefficient values and performing a convolution of the resulting point-spread function with an image of an acuity chart. The visual strehl ratio computed in the frequency domain was found to be well correlated (r2=0.81) to the letters lost. However, the comparison between simulated letters and metrics - both calculated using the same input (e.g. the measured wavefront map) - is also questionable. A more rigorous experiment would have compared metrics and measured visual performances in real conditions by using real optics.

Chen et al.15 used matching blur experiment to test the accuracy of various metrics to predict the subjective quality of vision. The subjects had to compare the blur caused by a single aberration to a combination of aberrations. However, as mentioned by the author, blur is not a unity perceptual experience, meaning that the appearance of the blur is different as a function of the origin of the blur. In other words, it appears very difficult to adjust the level of the defocus blur to equalize the blur caused by a LASIK wave aberration, for example.

In order to examine the accuracy of metrics - calculated using a previously published numerical eye model21 including the measurement of the monochromatic aberrations of the eye - and to predict the visual benefit of correcting high-order aberrations, we measured visual performances before and after AO-correction.

Mean contrast sensitivity and minimum angle of resolution (MAR) measured in condition (i) were respectively 22.51 (SEM=2.29) and 0.78’ (SEM=0.04’) corresponding to a Snellen acuity of around 20/15.6. Mean contrast sensitivity and MAR measured in condition (ii) were respectively 33.26 (SEM = 2.95) and 0.62’ (SEM=0.02’) corresponding to a Snellen acuity of 20/12.4. Consequently, the averag visual benefits of correcting the aberrations up to the 5th order were 1.64 (SEM=0.12) and 1.25 (SEM=0.04) respectively in terms of contrast sensitivity and visual acuity. CS and MAR measured in both conditions of correction are represented in figure 2 for all the observers ordered as a function of their age, and the corresponding VB is also shown.

Figure 2.

Measured MAR (a) and contrast sensitivity (b) for all the observers ordered as a function of their age with (grey bars) and without (black bars) AO correction. The upper part of the graph shows the corresponding measured visual benefit defined as the ratio of the visual performances measured in condition (i) and (ii). A ratio higher than 1 means that the performance is improved when correcting all the monochromatic aberrations up to the 5th order (i.e. condition (ii)).

(0.2MB).

CS and MAR measured in conditions (i) and (ii) are shown in figure 3 for the myopic group (n=10) and for the emmetropic group (n=15). Mean contrast sensitivity VB was respectively 1.68 (SEM=0.25) and 1.61 (SEM=0.11) for the myopes and emmetropes. The difference between groups was not significant (P=0.056, t test). In term of visual acuity, the average VB was 1.31 (SEM=0.07) for the myopic group and 1.21 (SEM=0.05) for the emmetropic group. The difference between groups was not significant (P=0.442, t test). There was no difference (P=0.11, t test) in the MAR, measured in condition (ii), between emmetropes and myopes. It should be noticed that the variability of visual performances across individuals was at least twice as large when measured in condition (i). This result can be explained by the variability of the level of HOA across subjects, which was respectively 0.14 and 0.02μm in condition (i) and (ii).

Figure 3.

Averaged measured MAR (a) and contrast sensitivity (b) for the myopic group and the emmetropic group in condition (ii) (dashed green squares) and in condition (i) (open red squares). The box and the error bars represent respectively the standard error and the standard deviation of the mean.

(0.12MB).

Figure 4 shows the correlation for the 25 subjects between the measured visual benefit and the value predicted by different metrics: (a) concerning the contrast sensitivity and (b), (c) and (d) concerning the visual acuity. Measured 16 c/deg-contrast sensitivity visual benefits were well correlated (r2=0.79) with VBs defined as the ratio of the modulation transfer function calculated at 16 c/deg in both conditions of aberration correction. None of the three calculated VB was found to be well correlated to the measured MAR improvement. The highest correlation (r2=0.30) was obtained with the metric where the MTF is weighted by the neural contrast sensitivity function (CSFN). However, the average VB on the population predicted by the intersection between the MTF and the neural threshold curve was comparable to the measured VB (see table 1).

Figure 4.

Correlation between the measured visual benefit and the value predicted by different metrics: (a) concerning the contrast sensitivity and (b), (c) and (d) concerning the visual acuity.

(0.24MB).

We measured a 1.25-fold and a 1.64-fold improvement respectively of the visual acuity of the contrast sensitivity measured at 16 c/deg when measuring the visual performances in condition (ii): dynamic correction of the aberrations up to the 5th order in polychromatic light with a 5.5 mm pupil size. Since the 2nd order aberration were statically corrected in condition (i) and dynamically corrected in condition (ii), this may explain a small part of the visual benefit measured in condition (ii).

As in the Guirao et al.’s calculations5, we measured important variations of VB across eyes with some eyes showing no benefit whereas others obtained a contrast sensitivity benefit higher than 3 or a visual acuity benefit higher than 1.6. Our measured VBs were comparable to the 1.9-fold and 1.2-fold improvement respectively in CS and VA measured by Yoon and Williams3 respectively on 2 and 7 subjects with a 6 mm pupil size. Poonja et al.4 and Rossi et al.6 obtained a visual acuity VB in monochromatic light of around 1.5. The difference of VB between these studies may be explained by the absence or presence of chromatic aberration. In fact, it has been suggested already that the impact of the correction of monochromatic aberrations could be enhanced in the absence of chromatic aberration. In fact, Yoon and Williams, MTF calculations3, based on the measured wave aberrations of 17 normal subjects, showed that the VB of correcting the monochromatic aberrations was twice as large as when computed in absence of chromatic aberration.

We did not find any correlation between the VB and age (see figure 2), between age and the high-order aberration magnitude (r2=0.04), between age and MTF-based metrics (r2<0.1) and nor between age and the visual performances measured in condition (ii) (i.e. r2=0.04 with MAR and r2=0.03 with CS). This last result indicates that the post-receptoral performance did not vary with age and corroborates the results of Dressler and Rassow30 and Kayazawa et al.31 who did not find any significant influence of the age on the neural contrast sensitivity function. While Rossi et al.6 found a significant lower visual benefit in the myopic group, we did not find any statistical difference. The main difference between Rossi's results and ours concerns the MAR measured in condition (ii) in the emmetropic group, since we found an average MAR of 0.62’ instead of 0.49’ for Rossi. Visual stimulus delivery in AOSLO allows the presentation of stimuli of higher optical quality, probably better than ours since we did not avoid the degradation of the retinal image induced by chromatic aberrations, which prevented us from obtaining a visual acuity closer to the limit of resolution of the retina derived from the photoreceptor spacing at the fovea (i.e. Mean= 0.51’, a Snellen acuity of 20/10.2, Yoon and Williams3).

The ratio of the 16 c/deg MTF computed in the two conditions of aberrations correction (i.e. MTFHO / MTFSC) is well correlated with the measured contrast sensitivity VB meaning that this metric is able to predict the effects of aberrations on contrast sensitivity. To our knowledge, only Yoon and Williams3 both measured and calculated the VB in term of contrast sensitivity. However, on the 2 subjects involved in their experiment, measurements and predictions were quite different. Since the contrast sensitivity can be well predicted by a metric based on wave aberration measurements, the neural adaptation to monochromatic aberrations first proposed by Artal7 does not seem to have played such an important role in our experiment. The hypothesis of a cortical adaptation to blur has been suggested to explain subjective preference to one's own aberration. However, this hypothetic effect has never been observed in term of contrast sensitivity measurements. In conclusion, this question requires further investigation.

We did not obtain the same quality of prediction with visual acuity. The best correlations were r2=0.30 and r2=0.27 respectively when computing the ratio of the volumes under the predicted CSFs (i.e. CSFHO / CSFSC) and when calculating the ratio of the intersections between the MTF and the CSFN. Both metrics involved the neural contrast sensitivity function measured by Williams29 on six subjects. Since the CSFN shows large inter-individual variation30-32 the inability to predict visual performances of individuals using data averaged on a population is not surprising. However, the intersection metric predicted an averaged VB of 1.21 (SEM=0.04) compared to the averaged measured VB of 1.25 (SEM = 0.02). This good agreement confirms that general tendencies of a population could be predicted by metrics taking into account a “standard” CSFN, whereas predictions for an individual requires a measurement of its own CSFN. Then, in this case it is better to directly measure the visual performance instead of predicting it via another measurement. An intermediate solution could be the use of typical CSFN of a given population. The typical CSFN should be categorized as a function of parameters such as retinal illuminance, age or ametropia. Consequently, the use of such categorized neural contrast sensitivity functions should improve the accuracy of the prediction of the visual acuity. However, part of the reason behind the better predictability of the CS could be explained by the inherent differences between the demands of the VA and CS tasks. Indeed, the VA task involves a more complex neural process than a task where the subject only needs to detect and recognize the direction of a grid (i.e. CS measurement).

In conclusion, metrics based on wave aberration measurements are able to predict the impact of monochromatic aberrations on contrast sensitivity and should be able to predict its effect on visual acuity with the use of categorized neural contrast sensitivity functions. This simulation should be considered as a useful tool for optical designers since it allows to predict the effect of optics (i.e. impact of aberrations) on visual performances without the need to manufacture prototypes, hence saving time and money.