1. Introduction

Accurate assessment of VA in aging populations is a critical component of clinical trials and the path to approval of therapies. There is a wide range of treatments currently under study for diseases that affect aging eyes, particularly age-related macular degeneration (AMD) and diabetes. Approximately 12 million people 40 years and over in the United States have vision impairment, including 1 million who are legally blind and 3 million who have vision impairment after correction [1]. AMD is the chief cause of irreversible visual impairment in industrialized countries outside of SE Asia, and diabetic retinopathy and diabetic macular edema (DME) are the chief causes of visual impairment in the working age adult [2,3]. The number of clinical trials to improve treatments is large and increasing dramatically. Limiting the search to only trials for adult patients registered in the US database Clinicaltrials.gov, there are now 2142 clinical trials for macular degeneration and 936 for macular edema [4]. The dramatic increase in clinical trials on these topics emphasizes their importance: when we submitted the grant proposal for NIH NEI EY030829 less than 2 yr ago, trials numbered 1862 for macular degeneration and 262 for treatment of diabetic retinal disease topics. Maintaining or improving a patient’s visual function, such as measured by VA, remains a key treatment goal and primary outcome measure for these trials.

There is a critical need for accurate tools for outcome measures in clinical trials that are reproducible, reliable and repeatable. These tools will have a large immediate market since visual function is a key component in almost all US clinical trials for diseases particularly common in the aging eye and now amenable to treatment: AMD and diabetic retinopathy and diabetic macular edema [5,6].

There are critical barriers to achieving accuracy of visual acuity in treatment trials, particularly since most methods are aimed at specifying a refraction rather than understanding the retinal sources of decreased VA. Most importantly, methods to measure VA or other metrics of visual function can have serious confounding influences due to the optics of the aging eye. For instance, performance with traditional charts and electronic displays that use Newtonian viewing cannot compensate for irregularities across the pupil plane from cataractous lenses and tear film abnormalities [7]. Small pupils in aging eyes [8] decrease light reaching the retina. The common clinical use of a pinhole aperture can reduce unwanted scattered light and effects of defocus, and with elderly patients, inaccurate centration of the pinhole over the pupil and vertex distance from the eye result in variable measures. In addition, the decrease of target light intensity at the retina can be greater than 1 log unit.

Using Maxwellian view to project the stimuli through a pupil of known size reduces the variability among subjects, since younger subjects are likely to have much larger pupils than the older subjects, whose pupil diameters can be as small as 1-3 mm. WF aberrations are typically characterized by Zernike polynomials [9]. The measured modulation transfer function is improved in some of the younger (20-30 yr) and older (60-70 yr) subjects for a 3 mm pupil compared to a larger diameter pupil [10]. Older subjects have greater amounts of higher order aberrations compared with younger subjects [11]. The finer spatial detail (high spatial frequency information) is transmitted better through a smaller pupil (3 mm) [12], and this was therefore selected for projection of our visual stimuli.

Typically, optical aberrations are not characterized in clinical trials for retinal treatments, despite the overall increased trend of higher order aberrations in healthy aging eyes [11]. Higher order aberrations are greater in the presence of cataractous lens [13,14] and tear film irregularities [15]. Patients are classified in some studies as phakic vs. pseudophakic, but without details of cataract progression in subjects or cataract surgery during a trial [5]. Thus, for longitudinal studies, it becomes unclear whether the retina, the optics, or both are changing during the treatment. Additionally, the retinal locus used for judging the VA targets is uncontrolled with wall charts and electronic (Newtonian) displays.

Subjective refraction prior to the measurement of VA by a trained refractionist is slow and tedious, requiring extensive training and certification. Although subjective refraction is typically specified in clinical trials when possible, it is not uncommon for a trial to use the habitual refraction because the patient cannot perform the procedure.

Subretinal fluid is commonplace in AMD and is also found in diabetic macular edema. Using habitual refraction in patients with subretinal fluid leads to focus errors because the photoreceptor layer is elevated, i.e. the eye can become hyperopic and the amount of defocus is likely to change over time. In cases when the patient has hemorrhage or other vision issues preventing refraction, use of the habitual refraction potentially leads to defocus for the baseline measurements, and the defocus will be in the hyperopic direction in the presence of subretinal fluid. Careful measurement of the defocus and other aberrations could improve the accuracy of VA, making possible distinguishing treatment effects from other sources of variability. While correcting the WF errors during VA measurements is a long term goal, measurement of the WF errors can help distinguish damage to the retina vs. poor optical quality of the visual stimulus on the retina. As part of the model for the VA of each individual patient, changes to VA with induced aberrations contribute to bounds on the variability of the VA measurement. We present an apparatus and method for inducing aberrations and measuring their effect on VA, including both central tendency and variability so that confidence limits on VA can be specified.

2. Methods

2.1 Subjects

Five subjects (2 females, 3 males) were tested, selected to have minimal lower aberrations and too young to be expected to have significant aging changes: 26-37 yr, mean =30 +/- 4.6 yr. A widely used and commercially available VA test provided a monitor-based and Newtonian view VA for these subjects (Smartsystem 2, M&S Technologies, Niles, IL) to demonstrate that the subjects had good VAs. This research, conducted at the Indiana University School of Optometry and Aeon Imaging, LLC, was reviewed by the Institutional Review Board of Indiana University and conforms to the Guidelines of the Declaration of Helsinki on Human Subject Research.

The PVT was designed (OpticStudio, Zemax, Ansys, Canonsburg, PA) and built to study factors in the aging eye that impact the relation between WF aberrations and VA (Fig. 1). The PVT is a 3-channel optical device on a portable breadboard with an adjustable height table. The optical channels include a Visual Display with a deformable mirror to induce optical aberrations in the visual stimulus, a HS sensor to read the patient’s WF aberrations, and a pupil camera and monitor channel to aid in alignment of the subject to the instrument. The visual stimulus and HS sensor channels were aligned to a readily visible 635 nm laser alignment beam.

 

Fig. 1. Photographs of the PVT from the subject’s left (top left panel) and the subject’s right (top right panel) and diagram of the Visual Display channel from Zemax (bottom). Shown is the VA experiment with the letter E. Light from the high resolution visual display is altered by the DM at a pupil plane to produce the induced WF aberrations, altering the light reaching the retina. The operator has two computers, one to run the VA experiment and another to run the vendor software to induce the aberrations. After positioning the subject’s eye using the headrest and pupil monitor, the Badal optometer was used to focus the stimulus on the retina. The Visual Display channel uses the light from an OLED screen of the Pixel 6 phone, which is inherently high contrast, and reduces the magnification and focuses the light on the retina using a series of spherical mirrors (SM), lenses (L1-L5 with L6 being a lenslet array for the Hartmann Shack Sensor (HS) and not to scale), flat mirrors (M), the piezo DM, and finally the Badal optometer that adjusts path length. The field stop (FS) controls the size of retinal image. Pupil position could be monitored with a pupil camera (PC), with the PC used in this experiment moved to the right of the subject as in the top panels. The HS shown is part of the HS + DM kit and was not used in the current experiment. An ancillary HSc was placed temporarily in the entrance pupil when calibrating the induced WF errors.

Download Full Size | PDF

The display channel provided a high pixel density VA target (Pixel 6 Smartphone, Google, Mountainview, CA) in Maxwellian view with a 3 mm pupil. Use of a high pixel density allowed testing down to the sizes of letters required to measure 20/5 VA. Targets are always mapped onto pixel boundaries, therefore eliminating errors from stimulus aliasing or a decrease of contrast from insufficient whole pixels for acuities < 20/20 or intermediate steps. The graphics magnification resulted in a slight variation from our goal of 4 pixels/arc min, resulting in a 20/17 target instead of 20/20, i.e. a change of 0.859. Graphs and statistics are corrected to the calibrated arc min.

The visual display was presented as a second monitor, as described previously [16]. To project the visual stimuli in real time and without changes to the gray scale, commercially available software was used to interface the Pixel phone to the output of the second Windows monitor (spacedesk, v. 1.0.48, datronicsoft, Augsburg, Germany). Visual stimuli appeared crisp and in high contrast when in optimal focus and WF aberrations were not induced. A black on white, centrally located fixation cross was presented for 2 sec prior to each E. The four-alternative (left, right, up, or down), forced-choice trials had continuous viewing of an E. Minimum angle of resolution (MAR) in arc min is measured as the 50% correct point of the psychometric function. Method of constant stimuli provided full psychometric function data for modeling mean and standard deviation. Psychometric functions were fit to 8 successive stimuli per letter size, with steps according to the logMAR scale and including letter sizes that produced 100% correct choices down to chance performance.

Focusing of the visual stimuli was accomplished in a woofer tweeter manner [17]. Each subject was adjusted to best focus using a Badal optometer on a motorized translation stage, i.e., the woofer. This adjustment of optical pathlength is necessary to allow the use of a low cost deformable mirror (DM), which does not have sufficient stroke for the range of focus required by highly myopic and hyperopic subjects, i.e., several diopters. Then a 40-actuator piezo deformable mirror, i.e., the tweeter, having a reflective, continuous surface piezoceramic disk (AOK8, DM and HS kit, Thorlabs) in the display path was set using vendor software control of Zernike terms [9]. The minification in the system resulted in 12 mm of the DM being projected onto to a 3 mm input pupil, including 24 actuators with the peripheral actuators partially illuminated. The DM provided sufficient adjustment to correct for other aberrations. The mirror was flat during adjustment of the Badal optometer. To obtain best focus for the subject, a letter E of 17 arc min was displayed. The experimenter moved the Badal in 0.4 D steps and asked if the letter was sharper or not. Once the best position was found, this process was repeated with 0.2 D steps.

The following conditions were tested, including higher order aberrations with some that mimic aging optical changes, with nominal values of no added aberrations, + 0.25 vertical and horizontal coma, and +0.25, -0.25, + 0.35, and -0.35 spherical aberration, with calibrations of Zernike terms in microns shown below. While larger induced aberrations were available from the DM, these values would likely not be consistent with a physiological range except in extreme pathology. The induced aberrations selected provided significant blur and ghost images to the target letter “E,” but not to the extent that the letter could not be correctly identified with a large increase in size as judged by the experimenters in pilot work. In situ calibrations of Zernike polynomials for the whole system were obtained by placing temporarily a second WF sensor, HS calibration (HSc), (Wfs 30, Thorlabs) at the entrance pupil to the eye for a 3 mm diameter pupil. The illumination of the calibration set-up was provided by a custom tungsten filament light source conjugate to the screen of the display. Both the illumination source and WF for calibration (WFc) were used in the system only temporarily during calibration, since they blocked the visual display and pupil plane for the subject. Calibration data were collected both by flattening the mirror between settings and also collecting WF error data over time. Lower order aberrations, likely due to the entire optical system, were present in the calibrations but decreased by the adjustment of the Badal optometer for each subject. The aberration condition was set prior to performing the entire series of letter sizes to obtain the VA for that condition.

The electronic controls for the head position motors and driving circuitry for the lasers and LEDs are mounted in the racks below the tabletop (bottom of the picture). One computer controls the WF settings, and another controls the visual display and collects subject responses for VA (PsychToolbox, Matlab) [18]. A four-button response box can be used to record the direction to which the letter E points.

2.2 Wavefront data to characterize optical aberrations

WF aberration data for each subject were collected from an FDA approved and commercially available WF measurement device (Pentacam AXL Wave, Oculus Optikgeraete GmbH, Wetzlar, Germany), selected because of the good repeatability [19,20]. The main aberration data were collected for natural pupil conditions at a 3 mm diameter, with the HSP. Comparison data are reported for a 5 mm pupil, which was readily achieved by the young subjects without dilating agents. The dataset included wavefront Pentacam (WFP) measurements for each subject: 5 subjects x 2 datasets per subject, and 2 of the many possible analyses reported (Zernike polynomial values for the 3- and 5-mm pupil data). The WFP measurement device was rotated 90 deg from the visual stimuli in the PVT, i.e., vertical aberrations were horizonal and vice versa. The HS sensor in the Thorlabs kit was not used in this study. As part of the overall capabilities of the PVT, a HS sensor channel is under development, with custom illumination sources including an 835 nm laser and a novel 810 nm LED, which are fiber based. These are seen (Fig. 1) attached to orange cables with the emitting ends obscured by mounts.

3. Results

All subjects had at least best corrected 20/20 VA on the commercial system, with 3 of the non-myopes having 20/15 and 1 having 20/20 + 2. Two subjects had measurable lower order wavefront errors, and the other three had small amounts of lower order aberrations (Fig. 2, Supplement 1). All subjects had small amounts of higher order aberrations (Table 1, Fig. 2). Thus, after using the Badal optometer to correct the focus for each subject, the induced aberrations describe the pattern of light on the retina. There is a 90 deg rotation between the Pentacam and the PVT. Vertical aberrations on the Pentacam are horizontal on the PVT and vice versa. We have plotted the Zernike terms with respect to the subject’s eye, so that horizontal and vertical are matched on the plots. Note that higher order aberrations measured for the subjects’ eyes were small compared with the induced aberrations, and therefore not important for this analysis.

 

Fig. 2. WFP aberrations (microns) for 5 normal subjects, showing that 3 mm pupil data (blue) have minimal higher order wavefront aberrations so that the induced aberrations describe the visual stimuli on the retina. Further, the 5 mm pupil WFP errors (red) are larger than for 3 mm data, despite that the data were from young subjects with little refractive error. Data are truncated at +/- 0.2 microns to expand to increase the visibility of the higher order aberrations. For F804, F1027, and M830, the values for defocus were 0.324, 0.620, and 1.24 microns, and for M826 and M823 only 0.144 and 0.0986 microns. Subject M830 is the myopic subject. The values for defocus at 5 mm were larger than for 3 mm: for F804, F1027m and M830 were 1.07 1.7, and 3.5 microns, and for M826 and M823 were 0.185 and 0.380 microns. Full data are presented in Supplement 1.

Download Full Size | PDF

Table 1. Subject age, sex, refraction, and higher order (HO) and lower order (LO) wavefront aberrations as measured with the Pentacam.

View Table

The piezo DM moved to each induced aberration configuration (Fig. 3) repeatably and stably (Fig. 4), when each configuration followed the flat mirror condition (Table S6, in Supplement 1). The aberration condition was relatively accurate for spherical aberrations, but was larger than expected for the coma conditions, as indicated by the calibrated values on Fig. 3. This allows the comparison of increasing levels of aberrations, such as different levels of spherical aberrations, to test whether aberrations within a range encountered in an aging eye without extreme pathology can combine linearly to alter VA.

 

Fig. 3. Deformable mirror configuration for the 7 nominal induced WF aberrations, used in testing the 5 normal subjects. Labels indicate the calibrated Zernike values for each condition at the Zernike mode intended for the induced aberration. Deformable mirror configurations control the WF for the visual stimuli that enter the pupil of the eye. Top row, left to right, mirror segments in pseudocolor showing nominal aberrations: No added aberrations (0 induced aberrations), + 0.49 microns Y coma, -0.51 microns X coma. Bottom row, left to right: mirror segments in pseudocolor showing nominal spherical aberrations, -0.35, -0.25, + 0.25 and +0.35, that were -0.32, -0.23, + 0.27, and +0.39 microns by calibration with the WFc. The more saturated colors indicate a greater change in the mirror position. Oranges indicate a shift of the mirror surface towards the subject, whereas blues indicate a shift away from the subject. Note that the amount of WF error induced is much larger than the higher order aberrations measured in the subjects. The optical system has a 90 deg rotation between the measurements in Fig. 2, and the induced aberrations with the mirror are for a 3 mm pupil.

Download Full Size | PDF

 

Fig. 4. Calibration of the WF errors through the system using the ancillary HSc sensor for a 3 mm pupil, showing consistency over time. Two sets of 5 measurements each are shown, with each set for each induced aberration condition. Blue symbols = time 1, orange symbols = time 2. WF error data are plotted for the flat mirror with no induced aberrations in the top panel, showing minimal wavefront errors of the device. The Y coma value of +0.49, and X coma of -0.51 microns have the largest wavefront contributions for the Zernike term intended to be induced, but also errors at other Zernike terms. The data are for the induced aberration conditions as shown in the top row of Fig. 3.

Download Full Size | PDF

The stability of the DM for WF errors is shown by the comparison of the means from two sets of 5 WF measurements of the instrument, collected in two periods 13–16 min apart (Fig. 4). The flat mirror condition (0 or no induced aberrations) has minimal measured WF aberrations. The letter E looked clear and high contrast for the flat condition and had blur and ghost images for the induced aberrations conditions. For each induced higher order aberration, there was minimal higher order WF error at each of the other WF condition, as shown in Fig. 4 and in Table S6. The system has unwanted WF aberrations for the lower order aberrations. The Badal optometer is set for each subject to reduce defocus errors when the DM is flat. The nominal induced aberration conditions are potentially influenced by other WF aberration errors, but these errors are small for the higher order aberrations. The size of the induced aberration for each condition of higher order WF aberration that we tested is much larger than the inherent aberrations in all subjects in Fig. 2, i.e. for these subjects there is little effect of innate higher order aberrations. An indication of the expected minimal impact of DM error for the induced lower order WF aberrations, such as defocus, is the good VA results. Only the myopic subject wore prescription lenses and yet the other 4 subjects had VA at or near 20/15. Two of the subjects had WF errors for defocus in the range of that of the induced aberration (up to .5 microns), and yet had excellent VA. Thus, although the lower order aberrations have WF error values that are numerically larger than the higher order WF errors, in fact the impact on VA of the lower order WF errors of this magnitude is not great for these emmetropic subjects.

When measuring VA, the decrease in the size of the letter E led to more errors, as expected. A cumulative normal distribution fits the data well (Fig. 5). The letter size at 50% correct was considered the mean threshold VA. The standard deviation (SD) for each subject was obtained from the fit functions. This provided an upper and lower bound of accuracy of the VA, providing confidence limits for each measurement set for each subject.

 

Fig. 5. Psychometric function for a normal subject with no induced aberration, with proportion of correct responses for the direction of the letter E as a function of size as MAR in arc min. Circles indicate the proportion correct, and the solid line is the fitted function. The subject is the first one in Fig. 2, F804.

Download Full Size | PDF

The VA differed across induced aberration conditions, worsening with induced aberrations as expected (Fig. 6) (ANOVA, 6 df, F = 15.8, p < .0001). Planned comparisons were performed between the no induced aberrations conditions and each of the other 6 conditions. The mean of the no induced aberrations had the MAR smaller compared with -0.51 X coma by 0.616 +/- 0.301 (p = . 0102), + 0.49 Y coma by 0.738 +/- 0.369 (p = .0111), -0.32 spherical aberration by 1.25 +/0.326 (p = .0010), -0.23 spherical aberration by 0.864 +/-0.364 (p = .0061), + 0.27 spherical aberration by 1.12 +/-0.318 (p = .0014), and +0.39 spherical aberration by 1.66 +/-0.232 (p < .0001). All differences for the planned comparison data with spherical aberrations were significant using a Bonferroni-Dunn correction for the number of tests, but the coma data slightly exceeded the significance criterion when all 6 statistics tests between means were performed. However, the VA data were based on repeated measures that provided means and SD values for each subject (Fig. 6). For each subject, each induced aberration condition exceeded the 95% confidence limit based on the no induced aberration condition. That is, for within subjects measures, all induced aberration conditions were significantly greater than the no induced aberrations condition. The standard deviations for individual subjects were sometimes quite large in some of the induced aberration conditions, but not systematically across conditions for a given subject or for all subjects for a given aberration condition. These were much larger than in the 0 induced aberrations condition. Our subjects had not practiced with these larger aberrations, and had little time to adapt to them. Further data will be needed to understand the performance of subjects with aberrations of these types, as most control subjects will not have aberrations of this high magnitude.

 

Fig. 6. Mean and standard deviation of 50% threshold VA for the 5 subjects for the 7 nominal induced aberration conditions. Top row, left to right: No induced aberrations, Y coma +0.49, and X coma -0.51. Bottom row, left to right: -0.32, -0.23, + 0.27 and +0.39 added spherical aberration. Note the VA worsened for subjects in the presence of induced WF aberrations, and typically was worse for the greater WF aberrations of +0.39 vs. + 0.27 or -0.32 vs. -0.23 spherical aberration. For several of the conditions, the most myopic had the worst VA.

Download Full Size | PDF

The VA with positive spherical aberrations, 0, + 0.27, and +0.39, was well-fit by a linear function, differing somewhat among subjects (Fig. 7). Similarly, the VA with negative spherical aberrations, 0, -0.23 and -0.32 was well-fit by a linear function, also differing somewhat among subjects (Fig. 7). The linear regression fits are described by R² ranging from 0.801 to 1.00 for the positive spherical aberrations and 0.859 to 1.0 for the negative spherical aberrations. The R² averaged 0.954 for the 10 linear fits. The slopes were similar across subjects, ranging from 4.26 to 5.32 for positive and -4.77 to -2.35 for negative aberrations. The myope had the worst VA, as indicated by the largest offset in the linear equation, 2.90 in a range down to 1.05. The -0.23 spherical aberration condition had the mean MAR smaller than the -0.32 spherical aberration condition by 0.383 +/-0.228 (p = .0198). The +0.27 spherical aberration condition had the MAR smaller than the +0.39 spherical aberration condition by 0.536 +/- 0.487 (p = .0699). The differences between +0.27 and +0.39 spherical aberrations and between -0.23 and -0.32 spherical aberration were consistent, but not statistically significant when correcting for the number of statistical tests.

 

Fig. 7. VA measured as the 50% threshold of MAR in arc min as a function of induced spherical aberration for 5 normal subjects, in microns as calibrated with the HSc. Left- VA for decreasing amounts of spherical aberration. Right- VA for increasing amounts of spherical aberration. F804- purple diamond, F1027- red circle, M826- blue triangle, M823- black bar, and M830- yellow square.

Download Full Size | PDF

4. Discussion

Our goal is to improve the accuracy of measurements of VA and other visual functions, for purposes of specifically assessing the health of the retina apart from optical issues [21]. The alternative, which is typically the case in clinical trials and in the clinic, is to obtain measurements of VA as a combination of the health of the retina confounded with the consequences of optical effects. One approach is to measure VA for a 3 mm diameter pupil to reduce individual and irrelevant differences in WF error on the retina. Our 5 subjects, who had minimal WF aberrations other than one myopic subject, produced WF data that had fewer errors at 3 compared to 5 mm pupils. Therefore, using the 3 mm pupil to measure VA can lead to fewer unwanted artifacts influencing the VA due to optical aberrations. While a 3 mm diameter pupil does not represent the real-world measurement for young patients except in bright light, a 3 mm or smaller diameter pupil is found in many older patients unless they have had their pupils dilated.

Using the custom-built PVT all 5 subjects produced data that were fit by linear functions across both positive and negative spherical aberration conditions. Higher order polynomial terms, power law, etc., were not needed to fit the data. Thus, the VA task could be considered a metric for how aberrations are combined at the retina, if supported by further testing of other higher order aberrations and subjects with inherent WF errors. However, there is considerable evidence that the sources of WF aberrations from the cornea and internal surfaces of the eye are not independent, and the visual system adapts to individual aberrations over time [11,22,23]. For patients in clinical trials, the specification of lower order and higher order WF aberrations gives more specific information about the impact on VA than the limited categorization of lens status (phakic, pseudophakic, aphakic), which can change over the course of a clinical trial that can last for years [5,6]. Classification of type of cataract provides information about the source of the WF errors, but not how they impact the current VA or potential VA.

To understand the accuracy of the VA measurements for individual subjects, a measure of variability (here the SD) is needed to improve clinical trials and the prognosis for individual patients. Combining patients into groups and specifying only between subject variability, which is often reported in clinical trials, does not give specific information for the individual patient. The accuracy of their VA is needed to determine if they have improved, gotten worse, or essentially stayed the same. In this study, the method of constant stimuli was used because it produces the full psychometric function. In contrast, the output of adaptive methods typically provides only a single number, which is taken as the threshold; the shape of the psychometric function is not given without the use of more complex algorithms [18,24,25]. With many adaptive methods, the standard deviation might not be well-characterized because most of the data points are near threshold rather than defining the tails of the function. The VA measurements in this study provide a step towards specifying the upper bound for VA measurements impacted by WF aberrations. The VA depends on the WF errors and other sources of variability, with mean and standard deviation providing estimates of both the VA and its accuracy.

The three main strengths of this research thus far are the use of a 3-mm pupil and Maxwellian view projection, the calibration of the instrument in situ, and the obtaining of repeat measures for the VA data. The latter is sufficient to provide accurate measurements and statistics on variability, in turn providing Confidence Limits for individual subjects. The method can be used to estimate better whether an individual patient’s VA remains the same, becomes significantly worse, or improves over time or with treatment. It is anticipated that with further development at least partial correction for WF errors will allow decreased variability, and therefore the chance to detect smaller changes in VA.

Another strength is that the method to measure VA is rapid, once alignment is achieved, i.e. under 3 min to obtain mean and variability. Further development into instruments providing robotic alignment and imaging can lead to reduced alignment times and more accurate pupil positioning. This is particularly important in that pupil misalignment can lead to substantial errors of WF measurement or correction. It is anticipated that these methods can lead to removal of the bottleneck for measuring VA in a clinic or treatment trial.

The main weaknesses are that the WF measurement devices are still expensive, and similarly the DM used is a compromise between lowering the cost and delivering an accurate WF correction. The progress to date provides a step towards improved real-world patients who have cataracts or other sources of optical artifacts. Even without correction by a DM, reporting out aberrations gives information about the quality of the stimulation of the retina. This can be part of a model that incorporates information about the quality of the light stimulus, details about retinal damage obtained from imaging studies, fixation stability, locus of fixation, and the statistics of variability. It is anticipated that fewer trials can be used, but with a robust variability measure retained. With further model development, for those patients who do not have adequate assessment of WF, then the Confidence Limits around the VA score can be increased.