Free
Research Article  |   May 2010
Visual performance with real-life tasks under Adaptive-Optics ocular aberration correction
Author Affiliations
Journal of Vision May 2010, Vol.10, 19. doi:10.1167/10.5.19
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to Subscribers Only
      Sign In or Create an Account ×
    • Get Citation

      Lucie Sawides, Enrique Gambra, Daniel Pascual, Carlos Dorronsoro, Susana Marcos; Visual performance with real-life tasks under Adaptive-Optics ocular aberration correction. Journal of Vision 2010;10(5):19. doi: 10.1167/10.5.19.

      Download citation file:


      © 2016 Association for Research in Vision and Ophthalmology.

      ×
  • Supplements
Abstract

We measured the effect of the correction of the natural aberrations of the eye by means of adaptive optics on the subject's performance on three different visual tasks: subjective sharpness assessment of natural images, familiar face recognition, and facial expression recognition. Images were presented through a dedicated psychophysical channel and viewed through an electromagnetic deformable mirror. Experiments were performed on 17 normal subjects. Ocular aberrations (astigmatism and higher order aberrations) were reduced on average from 0.366 ± 0.154 to 0.101 ± 0.055 μm for a 5-mm pupil diameter. On average, subjects considered to be sharper 84 ± 14% of the images viewed under AO correction, and there was a significant correlation between the amount of corrected aberrations and the percentage of images that the subject considered sharper when observed under AO-corrected aberrations. In all eyes (except one), AO correction improved familiar face recognition, by a factor of ×1.13 ± 0.12 on average. However, AO correction did not improve systematically facial expression recognition.

Introduction
The impact of optical aberrations on visual performance or, alternatively, the benefits of correcting ocular aberrations on vision are open questions in visual optics. The problem is of practical relevance, as several correction alternatives (i.e., refractive surgery, contact lenses, intraocular implants) address the possibility of compensating not only for defocus and astigmatism, but also, at least partially, for other high order aberrations (HOA) (Marcos, 2001; Marcos, Barbero, & Jiménez-Alfaro, 2005; McLellan, Prieto, Marcos, & Burns 2006; Mrochen, Kaemmerer, & Seiler, 2001; Piers, Weeber, Artal, & Norrby, 2007). The correction of HOA in the eye leads to an improvement in retinal image quality as well as visual function tested by standard tests. The study presented here addresses whether correcting optical aberrations results in a better perceptual quality of natural images and in an improvement of functional visual tasks. 
Adaptive optics is an ideal technique to manipulate the retinal image quality of the eye (Liang, Williams, & Miller, 1997). In combination with a psychophysical channel, adaptive optics has become a useful tool to test visual performance. To date, most studies that have looked into the effects of aberrations on vision used standard visual tests, primarily visual acuity and contrast sensitivity (Marcos, Sawides, Gambra, & Dorronsoro, 2008; Piers, Manzanera, Prieto, Gorceix, & Artal, 2007). Several studies correlated visual performance and optical quality metrics derived from the wave aberrations (either directly or through simulations of aberrated visual charts) (Applegate, Ballentine, Gross, Sarver, & Sarver, 2003; Applegate, Marsack, & Thibos, 2006; Jimenez, Ortiz, Hita, & Soler, 2008; Levy, Segal, Avni, & Zadok, 2005; Marsack, Thibos, & Applegate, 2004). Adaptive optics allows a better manipulation of the optical aberrations than simulations, allowing correction of the natural aberrations of the eye. Artal et al. (2004) found that correcting HOA produced a decrease of the minimum angle of resolution (MAR) for polychromatic high contrast targets (by a factor of 1.16). Yoon and Williams (2002) found a significant decrease in the logMAR by a factor of 1.2 for high luminance polychromatic targets and of 1.6 for dim monochromatic light. Dalimier, Dainty, and Barbur (2008) found visual benefits ranging from ×1 to ×1.7 depending on the luminance, using 15 arc min Landolt C target, which they define as being a functional visual task. In a recent study, we found that polychromatic high contrast visual acuity increased when aberrations were corrected for a large range of luminances and both contrast polarities (an average factor of 1.29 for black letters on white background and of 1.13 for white letters on black background; Marcos et al., 2008). While in particular the latter studies aimed at testing a wider range of conditions (in luminance and contrast), performance on clinical tests such as orientation discrimination in visual acuity represents a limited assessment of visual performance in the real world. However, and particularly as the correction of HOA is being considered in the clinical practice, it is important to investigate whether such correction will have a positive impact on daily life activities. 
The importance of using natural images and of assessing visual performance using tasks that are closer to the daily experience of subjects has been widely recognized. Although the physical constraints of the adaptive optics equipment does not allow to perform in situ tests in complex environments, as those undertaken in other areas (i.e., assessment of low vision devices (Peli, Goldstein, Young, Trempe, & Buzney, 1991; Peli, Lee, Trempe, & Buzney, 1994)), at least it is possible to perform visual tests using images of natural scenes (landscapes, buildings, faces, etc.). 
An obvious way to evaluate the benefits of optical correction is the subjects' subjective image sharpness assessment between the corrected and original image, where the subject has to choose between the two images which one he or she considered of better quality. These methods are widely used in radiology, and other areas where image processing techniques to compress images are used (Gur et al., 1997; Slone et al., 2000). Alternatively, one can test whether the ability to perform a given visual task is improved upon correction, or change with aging or disease. For example, Owsley and Sloane (1987) used identification of faces and daily objects like lamps, bicycles, or road signs to assess contrast thresholds in patients. Bullimore, Bailey, and Wacker (1991) tested face and facial expression recognition on healthy and age-related macular degeneration (ARMD) patients by determining the equivalent viewing distance required for recognition of identity and expression. They found that in normal patients face and expression recognition was similar but in ARMD patients, the decline in performance was due predominantly to difficulties in identity recognition. The equivalent viewing distance required for recognition of identity and of expression was very similar except in subjects with very poor visual acuity, for whom recognizing expressions became easier than recognizing identity. Peli et al. (1991) used a celebrity face recognition test to evaluate the visual improvement in low vision patients with digitally enhanced images. The data were analyzed in terms of the area under the receiver operating characteristic (ROC) curves for the two conditions natural and enhanced (Peli et al., 1991). ROC analysis has been shown to be a powerful tool to compare techniques or methods, and the statistical study of the ROC curves is well documented (Hanley, 1988; Hanley & McNeil, 1982, 1983; Metz, 1978, 2006, 2007; Metz, Wang, & Kronman, 1984). 
Face perception is perhaps one of the most highly developed visual skills in humans and it is mediated by high level cognitive processes. Face recognition is accomplished since a very early age, while facial expression recognition is essential for social communication. There have been debates on the role of spatial frequency on face and facial expression recognition (Costen, Parker, & Craw, 1996; Halit, de Haan, Schyns, & Johnson, 2006; Owsley & Sloane, 1987; Peli et al., 1991; Ruiz-Soler & Beltran, 2006). Correction of HOA will likely produce a contrast enhancement of the critical frequency band for face recognition (4–8 cycles/face: Peli et al., 1991; 8–16 cycles/face: Gold, Bennett, & Sekuler, 1999; Näsänen, 1999). Different spatial frequency sensitivities have been found between face identification and emotional expression recognition tasks (Posamentier & Abdi, 2003; Vuilleumier, Armony, Driver, & Dolan, 2003). Previous studies have examined the influence of image blur by digital band-pass filtering the images (Costen et al., 1996). The possibility of undertaking psychophysical experiments without the retinal blur produced by optical aberrations allows exploring these fundamental questions further, as the effects will be dominated purely by neural factors. 
In an effort to understand the parameters (neural and optical) that influence the effect of aberration correction on visual performance and whether correcting aberrations represents a real benefit in real world tasks, we examined the subjective image sharpness assessment (with natural or corrected aberrations) on complex natural scenes and the differences in familiar face recognition and facial expression recognition with and without aberration correction in normal individuals. 
Methods
Adaptive optics setup
Figure 1 shows a view of the adaptive optics setup. A detailed description of the system has been presented in recent studies where it was used to evaluate the improvement of visual acuity, at different luminance and contrast polarities, upon correction of ocular aberrations (Marcos et al., 2008) and to measure the accommodative response under manipulated aberrations (Gambra, Sawides, Dorronsoro, & Marcos, 2009). The primary components of the system are a Hartmann–Shack wave front sensor (HASO 32 OEM, Imagine Eyes, France) composed by a matrix of 32 × 32 microlenses with 3.6-mm effective diameter and an electromagnetic deformable mirror (MIRAO, Imagine Eyes, France) with 52 actuators in a 15-mm effective diameter. A motorized Badal system is used to compensate spherical error. A cold mirror behind the wave front sensor allows inserting visual stimulus channels in the deformable mirror path, so that the subject can perform psychophysical tasks under controlled optical aberrations. A flip mount mirror allows selecting between two different psychophysical channels. 
Figure 1
 
Adaptive optics setup with the two synchronized computers, one controlling the AO system (deformable mirror and Hartmann–Shack wave front sensor) and the Badal system the other controlling the psychophysical channel (ViSaGe platform and monitor).
Figure 1
 
Adaptive optics setup with the two synchronized computers, one controlling the AO system (deformable mirror and Hartmann–Shack wave front sensor) and the Badal system the other controlling the psychophysical channel (ViSaGe platform and monitor).
The first channel is identical to the previous study of Marcos et al. (2008): A 12 mm × 9 mm SVGA OLED minidisplay (LiteEye 400) is used for fixation during measurement and correction of the subject's aberration, where a white and black Maltese cross was projected. The second psychophysical channel was specifically developed for this study and allows the presentation of gray-scale images on a 12 × 16 inches Mitsubishi Monitor controlled with the ViSaGe psychophysical platform (Cambridge Research System, UK) and custom routines written in Matlab. The calibration of the monitor was performed using the ViSaGe platform and the ColorCAL colorimeter of Cambridge Research System. We performed the gamma correction using the ViSaGe platform and applied the correction to the natural images and faces presented. To control and validate the gamma correction applied, we performed visual measurements of the gamma function (following a procedure similar to that implemented in Psychtoolbox's VisualGammaDemo; Pelli, 1997). The effective maximum luminance of the monitor (after light losses in the system) was between 10 and 50 cd/m2
The system was controlled using custom routines written in Visual C++ and Matlab from two different computers, one controlling the AO system (Deformable mirror, Hartmann–Shack wave front sensor) and the Badal system the other controlling the ViSaGe psychophysical platform and the Mitsubishi Monitor. Computers were synchronized for a rapid presentation of visual stimuli under controlled aberrations (with the AO mirror) and best spherical refractive error correction (with a Badal system). 
Correction of ocular aberrations
The performance of the wave front sensor was validated using artificial eyes with known HOA as explained in previous publications (Gambra, Sawides, Dorronsoro, Llorente, & Marcos, 2008; Llorente, Barbero, Cano, Dorronsoro, & Marcos, 2004; Llorente, Diaz-Santana, Lara-Saucedo, & Marcos, 2003). Discrepancies in the measured aberrations were typically less than 5% with respect to nominal values. The calibration of the deformable mirror and the construction of the interaction matrix and command control matrix were performed using an artificial eye, as described in detail in our previous studies (Gambra et al., 2009; Marcos et al., 2008). In a closed loop operation, the residual wave front is continuously measured and command controls are continuously sent to the deformable mirror to keep the correction of astigmatism and HOA in real time (13 Hz). To perform a static correction, we stop the loop when the root mean square (RMS) of the wave front (excluding tilt and defocus) does not longer decrease, and we save the deformable mirror state with the voltage applied to each actuator for a future use. 
Natural and face images
The images presented to the subjects were acquired using a photographic digital camera (Canon PowerShot) with an original resolution of 4M pixels. All the photographs were converted to grayscale. Whether a gamma compensation is applied to the images saved from the digital camera is proprietary information. However, we performed a control experiment in 5 subjects using images presented in the display without the gamma correction to verify that this aspect did not affect the results. Some examples of natural scenes (typically landscapes, trees, buildings, street scenes, etc.) are shown in Figure 2. Grayscale face photographs were taken of volunteers, showing neutral, angry, and happy expressions. The subjects included people that were known to the participating observers (in most cases colleagues from the institute) and people unknown to them. The faces were cropped to remove background and the identification was predominantly dependent on facial features. For the facial expression recognition, the faces which were not sufficiently expressive were discarded. Figure 3 shows four examples of face images used in the face recognition experiments (familiar/unfamiliar and happy/angry). 
Figure 2
 
Examples of natural scenes from the subjective image sharpness assessment experiment.
Figure 2
 
Examples of natural scenes from the subjective image sharpness assessment experiment.
Figure 3
 
Examples of facial images from the familiar face and facial expression recognition experiments.
Figure 3
 
Examples of facial images from the familiar face and facial expression recognition experiments.
Images were presented at optical infinity for the observers. The natural scene images (640 × 480 pixels) subtended vertically 1.4 deg. The face images (320 × 240 pixels) subtended vertically 0.7 deg and therefore a normal subject would have its maximum contrast sensitivity (CSF) at about 3.5 cycles/face and a theoretical limiting resolution larger than 28 cycles/face. 
Subjective image sharpness assessment experiment
The subjective image sharpness assessment experiment consisted on a two-alternative forced-choice test. Thirty-four natural images were presented to the subject. On each trial, both an AO-corrected and a natural image were presented (with the appropriate best spherical refractive error correction) in a randomized order. There was no adaptation time to gray scale at the beginning. The image was presented during 3 s in both conditions (AO and noAO), without any grayscale screen in between, and then disappeared into a black screen until the subject make his/her choice. The subject's task was to choose the one of two that appeared sharper to him/her. 
Familiar face recognition experiment
The familiar face recognition involved presentation of 33 faces (15 familiar and 18 unfamiliar faces) to the subjects. The images were presented randomly viewed with the subject's natural aberrations and with the AO correction (with the appropriated best spherical refractive error correction). Each image was presented during 3 s in one condition and each image appeared twice throughout the experiment (for both AO and noAO conditions). There was no adaptation time to gray scale before the test, and between images there was a black screen during which the subject made his/her choice. The subject provided a graded response from 1 to 6 for their level of confidence in recognizing the face (1/6 for definitely familiar/unfamiliar, 2/5 for probably familiar/unfamiliar, 3/4 for a lower level of confidence on the face being familiar/unfamiliar). 
Facial expression recognition experiment
Fifty-two different facial expressions (26 happy and 26 angry faces) were randomly presented to the subjects, half of the images viewed under the subject's natural aberrations and the other half with the AO correction (with the appropriated best spherical refractive error correction in each case). The image presentation lasted 3 s. Each image was presented only in one condition (AO or noAO) during the experiment. There was no adaptation time to gray scale before the test and there was a black screen between two images during which the subject made his/her choice. The subject's choices were similar to those in the face recognition experiment, providing a graded response from 1 to 6 (1/6 for definitely happy/angry, 2/5 for probably happy/angry, 3/4 for a lower level of confidence on the face being happy/angry). 
Experimental protocols
The measurements were performed with subject's natural pupil (which ranged between 4.3 and 6.0 mm; a 6-mm artificial pupil was placed to avoid effective pupils larger than this diameter), and each experiment was conducted on one session for each subject, lasting around two hours. 
The subject's pupil was aligned to the system using a bite bar and the pupil was centered and focused. The subject was then asked to adjust the best subjective focus under his/her natural aberrations (noAO), starting from a myopic defocus by controlling the Badal system with a keyboard while he or she looked at a high contrast Maltese cross on the minidisplay. The resolution of the Badal system was 0.125 D. Natural aberrations (astigmatism and HOA) were measured and corrected in a closed loop adaptive optics operation. Then, the subject was asked again to adjust the Badal system position that provides the best subjective focus for this AO-corrected condition. Prior control experiments showed that the displacement of the Badal system did not introduce additional aberrations in the system (Gambra et al., 2008). An AO correction was deemed satisfactory if the residual aberration was bellow 0.2 μm. In most cases, the residual was less than 0.15 μm (RMS error correction of 71 ± 7%) except for one eye, which was that showing the highest amount of natural aberrations. A closed-loop correction (at a rate of 13 Hz) was typically achieved in 15 iterations. The state of the mirror that achieved this correction was saved and applied during the measurements when required in the psychophysical protocol. Psychophysical measurements were performed under static corrections of aberrations, as continuous dynamic correction would have involved continuous viewing of the spot test and discomfort to the subject (particularly given the relatively long duration of the test). 
Experiments involving the three described visual tests (subjective image sharpness assessment, familiar face recognition and facial expression recognition) were performed in the same session for each subject. Each test was only performed once to avoid learning bias. To ensure proper centration, the natural pupil was continuously monitored and the pupil was realigned if a shift of more than 0.5 mm was observed during the measurement. Aberration measurements were performed immediately before and after each test to ensure AO correction. A new closed-loop AO correction was performed, and the measurement repeated, if the percentage of correction had fallen by 15% from the initial value. 
Subjects
Seventeen young subjects aged 24 to 38 years (28.1 ± 4.8) participated in the experiment (17 participated in the subjective image sharpness assessment and facial expression recognition tests; 12 participated in familiar face recognition test). Spherical errors ranged between −5 and 4 D (−1.16 ± 1.13 D). Astigmatism ranged between 0 and 0.12 D (0.05 ± 0.03 D on average). HOA accounted for almost 70% of the RMS (excluding defocus and tilt). Subjects signed a consent form approved by the institutional review boards after they had been informed on the nature of the study and possible consequences. All protocols met the tenets of the Declaration of Helsinki. Four of the subjects were trained and the others were naïve subjects. 
Data analysis
Aberrations
Wave aberrations were fitted by 7th order Zernike polynomial expansions. Tilt and residual defocus term in the Zernike polynomial expansion (which was consistent with longitudinal chromatic aberration between visible and IR light; Llorente et al., 2003) were set to zero. Optical quality was evaluated in terms of RMS wave front error (excluding tilt and defocus) and Strehl ratio. 
Subjective image sharpness assessment experiment
Results of subjective image sharpness assessment were analyzed as percentage of images considered sharper with and without AO correction and presented for each subject. 
Face and facial expression recognition
Results of face (familiar and expression) recognition were analyzed in terms of the area under receiver operating characteristic (ROC) curves (Au_ROC), for the two conditions (natural and AO-corrected aberrations) where subjects have to give a score between 1 and 6 depending on their level of confidence (1 if he or she is confident that the face was familiar—or happy—and 6 if he or she is confident that the face was unfamiliar—or angry—). The ROC curves were generated using the subject's confidence ratings to calculate the probabilities of identifying a familiar face as familiar ( p(familiar ∣ familiar)) (true positive) and an unfamiliar face as familiar ( p(unfamiliar ∣ familiar)) (false positive). These probabilities were calculated as the fraction of faces identified with a certain confidence rating. For the first point of the ROC curve, the fraction of familiar faces receiving the rating 1 was plotted against the fraction of unfamiliar faces receiving the same rating. The following points of the curve were calculated as the cumulative fraction of the subsequent ratings. Therefore, the ROC curves represent the true-positive fraction versus false-positive fraction (Metz, 1978, 2006, 2007), i.e., the probability of correctly identifying a face as familiar versus the probability of incorrectly identifying an unfamiliar face as familiar. The same analysis is performed for happy/angry faces. The area under the ROC curves was measured with a trapezoidal method using the raw data. Perfect performance corresponds to an Au_ROC = 1. Inability to recognize any face or expression, i.e., completely arbitrary responses, will produce Au_ROC = 0.5. For some calculations and illustration purposes, we subtract the offset and defined Au_ROC′ = Au_ROC − 0.5. The analysis was performed separately for images viewed through natural aberrations and AO-corrected aberrations. The area under ROC curves was taken as a measure of recognition in each condition. 
The areas under the ROC curves in both conditions, Au_ROC_noAO and Au_ROC_AO, were statistically compared. For both recognition experiments, the statistical analysis was based on the standard error ( SE) of the difference between Au_ROC_noAO and Au_ROC_AO and follows the procedures of Hanley and McNeil (1982, 1983), which is similar to Wilcoxon (or Mann–Whitney). As the familiar face recognition experiment was conducted by presenting the images twice throughout the experiment (with both conditions, noAO and AO), the correlation between Au_ROC_noAO and Au_ROC_AO was taking into account, performing a correlated ROC analysis as described by Hanley and McNeil (1983). In the facial expression recognition experiment, we presented images only once (with either one or the other condition, noAO or AO). No correlation existed between Au_ROC_noAO and Au_ROC_AO, and therefore we applied a standard bivariate statistical analysis to calculate the Standard Error SE
To calculate the SE of one ROC curve, we used Equation 1, from previous literature (Hanley & McNeil, 1982): 
SE=Au\_ROC×(1Au\_ROC)+(n21)×(Q1Au\_ROC2)+(n11)×(Q2Au\_ROC2)n1n2,
(1)
where n1 is the number of unfamiliar/angry faces (in the familiar face and facial expression recognition, respectively), n2 is the number of familiar/happy faces (in the familiar face and facial expression recognition, respectively), and Q1 and Q2 are calculated as Equations 2 and 3: 
Q1=Au\_ROC2Au\_ROC,
(2)
 
Q1=2×Au\_ROC21+Au\_ROC.
(3)
 
We calculated the standard error of the difference ( SE (Au_ROC_noAO − Au_ROC_AO)), annotated SE(diff) in the Equation 4 for the facial expression recognition experiment (where different set of images were used in the two conditions AO, noAO) and Equation 5 for familiar face recognition where the same sets of images were used in both conditions and where Au_ROC_AO and Au_ROC_noAO are likely to be correlated.  
S E ( d i f f ) = S E 2 ( A u \_ R O C \_ n o A O ) + S E 2 ( A u \_ R O C \_ A O ),
(4)
 
S E ( d i f f ) = S E 2 ( A u \_ R O C \_ n o A O ) + S E 2 ( A u \_ R O C \_ A O ) 2 . r . S E ( A u \_ R O C \_ n o A O ) . S E ( A u \_ R O C \_ A O ),
(5)
where r is a quantity representing the correlation introduced between the two areas by studying the same sample of faces (Hanley & McNeil, 1983). 
Once we have the standard error of the difference in areas, we can calculate the z statistic using Equation 6:  
z = A u \_ R O C \_ n o A O A u \_ R O C \_ A O S E ( d i f f ).
(6)
 
If z is above a critical level, we accept that the difference between the two ROC curves is significant. Typically, for a confidence level of 95%, the critical level is set at 1.96, which correspond to a type I error probability ( p-value, two tails) of 0.05 as a criterion for a significant difference. 
Potential improvements in performance were analyzed in terms of ratios and differences of Au_ROC′ (AO vs noAO), and also in terms of Gain. Gain was defined as the actual improvement normalized to the potential maximum improvement, as expressed by Equation 7  
G a i n = A u \_ R O C \_ A O A u \_ R O C \_ n o A O 0.5 A u \_ R O C \_ n o A O .
(7)
 
Results
Best corrected ocular aberrations
Figure 4 shows RMS wave front error (excluding tilt and defocus) for all 17 subjects of the study before and after AO correction of aberrations. Data are for 5-mm pupil diameters, except for subject S8 that had a 4.3-mm pupil diameter. Subject S1 performed the measurements wearing her soft contact lenses. 
Figure 4
 
RMS wave front error (excluding tilt and defocus) for all 17 subjects before and after AO correction. Data are for 5-mm pupil diameters (except for S8, who had a 4.3-mm pupil diameter).
Figure 4
 
RMS wave front error (excluding tilt and defocus) for all 17 subjects before and after AO correction. Data are for 5-mm pupil diameters (except for S8, who had a 4.3-mm pupil diameter).
On average, RMS (excluding tilts and defocus) decreased from 0.366 ± 0.154 to 0.101 ± 0.055 μm, with an average correction of 72 ± 7%. The estimated RMS excluding tilts, defocus and astigmatism was 0.244 ± 0.113 μm. The average RMS values were calculated for 5-mm pupil diameters, except for S8 who had a 4.3-mm pupil. 
For the subjects participating in the control experiment, RMS (excluding tilts and defocus) decreased, on average, from 0.417 ± 0.218 to 0.078 ± 0.037 μm, with an average correction of 80 ± 10%, for 5-mm pupil. 
Subjective image sharpness assessment experiment
Figure 5 shows the percentage of images considered sharper with and without AO by each subject. On average, subjects considered sharper 84% of images viewed with AO. 
Figure 5
 
Percentage of images considered sharper with or without AO for each subject.
Figure 5
 
Percentage of images considered sharper with or without AO for each subject.
In the control experiment, where no gamma correction had been applied in the display, the percentage of images which were considered sharper by the subjects when AO was applied was 78% on average. 
We explored potential correlations between the subjective image sharpness assessment and the amount of corrected aberrations. Figure 6 shows the correlation plots and corresponding regression lines between the percentage of images considered sharper with AO correction and the optical quality improvement in terms of RMS ( Figure 6A) or Strehl ratio at best focus ( Figure 6B) for a 5-mm pupil (4.3-mm pupil for S8). We found a significant correlation between the percentage of images considered sharper with AO and the amount of corrected aberrations in both cases ( R = −0.7, p < 0.002 for RMS and R = 0.5, p < 0.05 for Strehl ratio). 
Figure 6
 
Correlation between the percentage of images considered sharper with AO and the optical quality improvement in terms of RMS (A) or Strehl ratio at best focus (B) for a 5-mm pupil (4.3-mm pupil for S8).
Figure 6
 
Correlation between the percentage of images considered sharper with AO and the optical quality improvement in terms of RMS (A) or Strehl ratio at best focus (B) for a 5-mm pupil (4.3-mm pupil for S8).
This indicates that the subjects who experienced larger improvement in their optics considered sharper a larger number of images viewed under adaptive optics correction than under natural aberrations. 
Familiar face recognition
Au_ROC_noAO ranged from 0.69 to 0.98 and Au_ROC_AO from 0.74 to 1.0 across subjects. Figure 7 shows an example of response operating characteristic (ROC) curves for four representative subjects. In terms of area under ROC, although the overall performance is very different, we observe an improvement in familiar face recognition with AO in all four subjects. 
Figure 7
 
Examples of ROC curves, probability of correctly identifying a face as familiar versus the probability of incorrectly identifying an unfamiliar face as familiar, for four subjects. The inset shows the ratio of the area under the curves.
Figure 7
 
Examples of ROC curves, probability of correctly identifying a face as familiar versus the probability of incorrectly identifying an unfamiliar face as familiar, for four subjects. The inset shows the ratio of the area under the curves.
In the control experiment (no gamma correction in the display), we found an average increase in the area under ROC curves by a factor of ×1.07. 
As Au_ROC = 0.5 corresponds to arbitrary responses, an offset of 0.5 was subtracted from the computed areas. Au_ROC′ then correspond to the area under ROC without the 0.5 offset. Figure 8 shows the change in the areas under ROC curves, under natural and AO-corrected aberrations for the 12 subjects that participated in the test (after the 0.5 offset subtraction). Except for one subject (S7, with Au_ROC′ ratio of 0.98), there is systematic increase in the area under ROC curve with AO correction. On average, Au_ROC′ increased by ×1.13 ± 0.12 with correction. 
Figure 8
 
Ratio of the area under ROC (AO-corrected/natural aberrations) for familiar face recognition (after 0.5 offset subtraction). The inset shows the difference in areas under ROC (AO—natural aberrations) for familiar face recognition. Error bars correspond to the standard error ( SE) calculated with the statistical analysis. Double stars indicate a significant difference with a 95% level of confidence and single stars for a level of confidence higher than 70%.
Figure 8
 
Ratio of the area under ROC (AO-corrected/natural aberrations) for familiar face recognition (after 0.5 offset subtraction). The inset shows the difference in areas under ROC (AO—natural aberrations) for familiar face recognition. Error bars correspond to the standard error ( SE) calculated with the statistical analysis. Double stars indicate a significant difference with a 95% level of confidence and single stars for a level of confidence higher than 70%.
A paired t-test showed a statistical significant difference between the ROC curves under natural and AO-corrected aberrations across all subjects ( p = 0.0027). 
Statistical calculation using Equation 6 showed a statistically significant difference between the two areas under ROC curves (AO-corrected and natural aberration), with a level of confidence of 95% in one eye (S3) and two other eyes with a level of confidence more than 70%. The inset of Figure 8 shows the difference in the areas under ROC curves for each subject, their standard errors and the cases where the difference is statistically significant. 
Facial expression recognition
Figure 9 shows examples of ROCs representing the probability to recognize a happy face when it is happy, for four subjects, the same as in the familiar face recognition experiment. We did not see a systematic improvement with the AO correction. Au_ROC_noAO ranged from 0.85 to 1.0 and Au_ROC_AO ranged from 0.86 to 1.0 across all subjects. 
Figure 9
 
Example of ROC curves, probability of correctly identifying a happy facial expression versus the probability of incorrectly identifying an angry expression as a happy one, for four patients. The inset shows the ratio of the area under the curves.
Figure 9
 
Example of ROC curves, probability of correctly identifying a happy facial expression versus the probability of incorrectly identifying an angry expression as a happy one, for four patients. The inset shows the ratio of the area under the curves.
In the control experiment, we found an average decrease of the Au_ROC_AO/Au_ROC_noAO ratio by a factor of ×0.93. 
Figure 10 shows the change in the areas under the ROC curves, under natural and AO-corrected aberrations for all subjects (after the 0.5-offset subtraction). In 9 out of 17 subjects, AO correction improved facial expression recognition, but in 8 subjects the ratio was below 1. On average, Au_ROC′ ratio was ×1.01 ± 0.11. A paired t-test did not show a statistically significant difference between the areas under ROC curves, under natural and AO-corrected aberrations across subjects ( p-value = 0.8315). 
Figure 10
 
Ratio of the area under ROC (with AO/natural aberrations) for facial expression recognition (after 0.5 offset subtraction). The inset shows the difference in areas under ROC (AO—natural aberrations) for familiar face recognition and errors bar correspond to the Standard Error ( SE) calculated with the statistical analysis. Single stars indicate a significant difference with a level of confidence higher than 70%.
Figure 10
 
Ratio of the area under ROC (with AO/natural aberrations) for facial expression recognition (after 0.5 offset subtraction). The inset shows the difference in areas under ROC (AO—natural aberrations) for familiar face recognition and errors bar correspond to the Standard Error ( SE) calculated with the statistical analysis. Single stars indicate a significant difference with a level of confidence higher than 70%.
Statistical calculation using Equation 6 did not show any statistically significant difference between the two areas under ROC curves (AO-corrected and natural aberration), with a level of confidence of 95% but a statistically significant difference with a level of confidence higher than 70% for three subjects. The inset of Figure 10 shows the difference in the areas for each subject, their standard errors and the cases where the difference is statistically significant. 
Comparison of familiar and face expression recognition
Figure 11 compares the ratio of the areas under ROCs' for familiar and face expression recognition in all subjects that performed both experiments. Except for subject S7, the improvement in performance for familiar face recognition when the aberrations where corrected was systematically higher than the improvement of facial expression recognition. The difference varied across individuals and the increase in performance in the familiar face recognition task was uncorrelated with the change in performance in the facial expression recognition ( p > 0.8). 
Figure 11
 
Comparing familiar face and facial expression recognition in terms of changes in the area under ROC' with AO correction.
Figure 11
 
Comparing familiar face and facial expression recognition in terms of changes in the area under ROC' with AO correction.
We correlated positive gain ( Equation 7) with optical quality improvement in terms of RMS (AO/noAO) (for 5-mm pupil, and 4.3-mm pupil for S8), as shown in Figure 12. We found a nearly statistically significant correlation ( p = 0.09) for familiar face recognition. For facial expression recognition, the gain with AO correction was always lower than 10% and did not show any correlation with the optical quality improvement in terms of RMS (AO/noAO). 
Figure 12
 
Correlation between gain (defined in Equation 7) and optical quality improvement (in terms of RMS (AO/noAO) for 5-mm pupil (4.3-mm pupil for S8) for familiar face and facial expression recognition tasks.
Figure 12
 
Correlation between gain (defined in Equation 7) and optical quality improvement (in terms of RMS (AO/noAO) for 5-mm pupil (4.3-mm pupil for S8) for familiar face and facial expression recognition tasks.
Discussion
We found that correcting aberrations increases dramatically the subjective impression of sharpness in natural images. On average, subjects identified as sharper more than 80% of the images (up to 100% of the images for some subjects) viewed under AO-corrected aberrations. We did not observe that some image category (i.e., artificial environments or natural landscapes) was more consistently identified as sharper with AO correction than others. The fact that there is a significant correlation between the percentage of images considered sharper with AO correction and the amount of corrected aberrations ( Figure 6) indicates that the correction of aberrations effectively increases contrast and enhances high spatial frequencies which may not be visible in the presence of aberrations. 
Correcting aberrations improved slightly (but systematically in all subjects but one) visual performance of real-life visual tasks such as familiar face recognition. However, although AO correction had a positive impact on facial expression recognition in some subjects, on average performance on this task did not improve. 
Correction of HOA will likely produce a contrast enhancement of the critical frequency band for face recognition (4–8 cycles/face: Peli et al., 1991; 8–16 cycles/face: Gold et al., 1999; Näsänen, 1999). A frequency band of 4–8 cycles/face corresponds to a spatial frequency range of 6–11 cycles/deg in our setting and 8–16 cycles/face to 11–23 cycles/deg. We estimated that the AO correction resulted in an increase of ×1.9 in the volume under the MTF in the lower frequency band and an increase by a factor of ×2.7 in the higher frequency band, on average across subjects. This increase, in contrast and in the spatial frequency content of the image, may have made it easier for the subject to recognize the scene. Moreover, we found that gain in the familiar face recognition tended to be correlated with the optical correction (Figure 12) suggests that subjects use contrast increase and spatial resolution as cues. 
The differences between the responses to the familiar face and facial expression recognition tasks likely arise from fundamental differences in the neural pathways (Bruce & Young, 1986; Young, McWeeny, Hay, & Ellis, 1986) as well as the different facial features critical in each task, even if both tasks are accomplished since a very early age (Erickson & Schulkin, 2002; Grossmann & Vaish, 2008; Nelson, 2001). Familiar face recognition is a more difficult task than facial expression, as indicated by its systematically lower AU_ROCs. 
Studies using functional magnetic resonance imaging (fMRI) in humans have shown that neural responses to repeating the same face identity were greater with intact or high-spatial-frequency face stimuli than with low-frequency faces, regardless of emotional expression. However, responses to fearful expressions were greater for intact or low-frequency faces than for high-frequency faces (Vuilleumier et al., 2003). These results are consistent with our finding that correction of HOA has a higher positive impact on recognition of familiar faces than on recognition of emotional expressions, a task that can be successfully undertaken. Very likely, a higher spatial frequency range is used to recognize a familiar face than a facial expression. Debates are open to determine these critical spatial frequencies. 
The practical relevance of these improvement remains to be evaluated to deem whether correction of higher order aberrations in normal subjects is sufficiently important from a clinical point of view. However, our results indicate that a static correction of aberrations is very significantly appreciated by the subject and may result in improvements of some daily life activities (such as recognition of faces). While we cannot exclude effects of neural adaptation (Webster, Georgesson, & Webster, 2002) to the natural aberrations of the eye, those are potentially secondary in our results, as we did not find a trend toward better performance with the natural aberrations of the eye with respect to the AO-corrected optics. 
Conclusions
Correcting ocular aberrations produced an improvement in visual performance. The subjective impression of sharpness increased significantly when high order aberrations were corrected by means of Adaptive Optics. The improvement of familiar recognition was systematic in all but one subject. Correcting high order aberrations did not improve systematically facial expression recognition. 
Acknowledgments
MEyC FPI Predoctoral Fellowship to LS; CSIC I3P Predoctoral Fellowship to EG; MICINN FIS2008-02065 and PETRI PET-2006-0478, and EURYI-05-102-ES (EURHORCs-ESF) to SM. The authors thank the subjects for their patience to participate in the experiments of this study. 
Commercial relationships: none. 
Corresponding author: Lucie Sawides. 
Email: lucie@io.cfmac.csic.es. 
Address: Instituto de Óptica, CSIC. Serrano 121, 28006, Madrid, Spain. 
References
Applegate R. A. Ballentine C. Gross H. Sarver E. J. Sarver C. A. (2003). Visual acuity as a function of Zernike mode and level of root mean square error. Optometry and Vision Science, 80, 97–105. [PubMed] [CrossRef] [PubMed]
Applegate R. A. Marsack J. D. Thibos L. N. (2006). Metrics of retinal image quality predict visual performance in eyes with 20/17 or better visual acuity. Optometry and Vision Science, 83, 635–640. [PubMed] [Article] [CrossRef] [PubMed]
Artal P. Chen L. Fernandez E. J. Singer B. Manzanera S. Williams D. R. (2004). Neural compensation for the eye's optical aberrations. Journal of Vision, 4, (4):4, 281–287, http://journalofvision.org/content/4/4/4, doi:10.1167/4.4.4. [PubMed] [Article] [CrossRef]
Bruce V. Young A. (1986). Understanding face recognition. British Journal of Psychology, 77, 305–327. [PubMed] [CrossRef] [PubMed]
Bullimore M. A. Bailey I. L. Wacker R. T. (1991). Face recognition in age-related maculopathy. Investigative Ophthalmology and Visual Science, 32, 2020–2029. [PubMed] [Article] [PubMed]
Costen N. P. Parker D. M. Craw I. (1996). Effects of high-pass and low-pass spatial filtering on face identification. Perception & Psychophysics, 58, 602–612. [PubMed] [CrossRef] [PubMed]
Dalimier E. Dainty C. Barbur J. (2008). Effects of higher-order aberrations on contrast acuity as a function of light level. Journal of Modern Optics, 55, 791–803. [CrossRef]
Erickson K. Schulkin J. (2002). Facial expressions of emotion: A cognitive neuroscience perspective. Brain and Cognition, 52, 52–60. [PubMed] [Article] [CrossRef]
Gambra E. Sawides L. Dorronsoro C. Llorente L. Marcos S. (2008). Development, calibration and performance of an electromagnetic mirror based adaptive optics system for visual opticsn Proceedings of the sixth international workshop on adaptive optics for industry and medicine (pp. 322–328). London: Imperial College Press.
Gambra E. Sawides L. Dorronsoro C. Marcos S. (2009). Accommodative lag and fluctuations when optical aberrations are manipulated. Journal of Vision, 9, (6):4, 1–15, http://journalofvision.org/content/9/6/4, doi:10.1167/9.6.4. [PubMed] [Article] [CrossRef] [PubMed]
Gold J. Bennett P. J. Sekuler A. B. (1999). Identification of band-pass filtered letters and faces by human and ideal observers. Vision Research, 39, 3537–3560. [PubMed] [Article] [CrossRef] [PubMed]
Grossmann T. Vaish A. Striano T. Reid V. (2008). Reading faces in infancy: Developing a multi-level analysis of a social stimulus. Social cognition: Development, neuroscience and autism. Chicester: Wiley and Blackwell.
Gur D. Rubin D. A. Kart B. H. Peterson A. M. Fuhrman C. R. Rockette H. E. (1997). Forced choice and ordinal discrete rating assessment of image quality: A comparison. Journal of Digital Imaging, 10, 103–107. [PubMed] [CrossRef] [PubMed]
Halit H. de Haan M. Schyns P. G. Johnson M. H. (2006). Is high-spatial frequency information used in the early stages of face detection? Brain Research, 1117, 154–161. [PubMed] [Article] [CrossRef] [PubMed]
Hanley J. A. (1988). Alternative approach to Receiver Operating Characteristic Analyses. Radiology, 168, 568–570. [CrossRef] [PubMed]
Hanley J. A. McNeil B. J. (1982). The meaning and use of the area under a receiver operating characteristic (ROC curve. Radiology, 143, 29–36. [PubMed] [Article] [CrossRef] [PubMed]
Hanley J. A. McNeil B. J. (1983). A method of comparing the areas under receiver operating characteristic curves derived from the same case. Radiology, 148, 839–843. [PubMed] [Article] [CrossRef] [PubMed]
Jimenez J. R. Ortiz C. Hita E. Soler M. (2008). Correlation between image quality and visual performance. Journal of Modern Optics, 55, 783–790. [CrossRef]
Levy Y. Segal O. Avni I. Zadok D. (2005). Ocular higher-order aberrations in eyes with supernormal vision. American Journal of Ophthalmology, 139, 225–228. [PubMed] [CrossRef] [PubMed]
Liang J. Williams D. R. Miller D. T. (1997). Supernormal vision and high resolution retinal imaging through adaptive optics. Journal of the Optical Society of America A, Optics, Image Science, and Vision, 14, 2884–2892. [PubMed] [CrossRef] [PubMed]
Llorente L. Barbero S. Cano D. Dorronsoro C. Marcos S. (2004). Myopic versus hyperopic eyes: Axial length, corneal shape and optical aberrations. Journal of Vision, 4, (4):5, 288–298, http://journalofvision.org/content/4/4/5, doi:10.1167/4.4.5. [PubMed] [Article] [CrossRef]
Llorente L. Diaz-Santana L. Lara-Saucedo D. Marcos S. (2003). Aberrations of the human eye in visible and near infrared illumination. Optometry and Vision Science, 80, 26–35. [PubMed] [CrossRef] [PubMed]
Marcos S. (2001). Aberrations and Visual Performance following standard laser vision correction. Journal of Refractive Surgery, 17, 596–601. [PubMed]
Marcos S. Barbero S. Jiménez-Alfaro I. (2005). Optical quality and depth-of-field of eyes implanted with spherical and aspheric intraocular lenses. Journal of Refractive Surgery, 21, 223–235. [PubMed] [PubMed]
Marcos S. Sawides L. Gambra E. Dorronsoro C. (2008). Influence of Adaptive-optics ocular aberration correction on visual acuity at different luminances and contrast polarities. Journal of Vision, 8, (13):1, 1–12, http://journalofvision.org/content/8/13/1, doi:10.1167/8.13.1. [PubMed] [Article] [CrossRef]
Marsack J. D. Thibos L. N. Applegate R. A. (2004). Metrics of optical quality derived from wave aberrations predict visual performance. Journal of Vision, 4, (4):8, 322–328, http://journalofvision.org/content/4/4/8, doi:10.1167/4.4.8. [PubMed] [Article] [CrossRef]
McLellan J. S. Prieto P. Marcos S. Burns S. (2006). Are the eye's wave aberrations random? Vision Research, 46, 2546–2553. [CrossRef] [PubMed]
Metz C. E. (1978). Basic principles of ROC analysis. Seminars in Nuclear Medicine, 8, 283–298. [PubMed] [CrossRef] [PubMed]
Metz C. E. (2006). Receiver operating characteristic analysis: A tool for the quantitative evaluation of observer performance and imaging systems. Journal of the American College of Radiology, 3, 413–422. [PubMed] [CrossRef] [PubMed]
Metz C. E. (2007). ROC analysis in medical imaging: A tutorial review of the literature. Radiological Physics and Technology, 1, 2–12. [CrossRef] [PubMed]
Metz C. E. Wang P. L. Kronman H. B. Deconinck F. (1984). A new approach for testing the significance of differences between ROC curves measured from correlated data. Information processing in medical imaging. (pp. 432–445). The Hague: Nijhoff.
Mrochen M. Kaemmerer M. Seiler T. (2001). Clinical results of wavefront-guided laser in situ keratomileusis 3 months after surgery. Journal of Cataract and Refractive Surgery, 27, 201–207. [PubMed] [CrossRef] [PubMed]
Näsänen R. (1999). Spatial frequency bandwidth used in the recognition of facial images. Vision Research, 39, 3824–3833. [PubMed] [Article] [CrossRef] [PubMed]
Nelson C. A. (2001). The development and neural bases of face recognition. Infant and Child Development, 10, 3–18. [CrossRef]
Owsley C. Sloane M. E. (1987). Contrast sensitivity, acuity and the perception of real world targets. British Journal of Ophthalmology, 71, 791–796. [PubMed] [Article] [CrossRef] [PubMed]
Peli E. Goldstein R. B. Young G. M. Trempe C. L. Buzney S. M. (1991). Image enhancement for the visually impaired: Simulations and experimental results. Investigative Ophthalmology and Visual Sciences, 32, 2337–2350. [PubMed]
Peli E. Lee E. Trempe C. L. (1994). Image enhancement for the visually impaired: The effects of enhancement on face recognition. Journal of the Optical Society of America A, Optics, Image Science, and Vision, 11, 1929–1939. [PubMed] [CrossRef] [PubMed]
Pelli D. G. (1997). The VideoToolbox software for visual psychophysics: Transforming numbers into movies. Spatial Vision, 10, 437–442. [PubMed] [CrossRef] [PubMed]
Piers P. A. Manzanera S. Prieto P. M. Gorceix N. Artal P. (2007). Use of adaptive optics to determine the optimal ocular spherical aberration. Journal of Cataract and Refractive Surgery, 33, 1721–1726. [PubMed] [CrossRef] [PubMed]
Piers P. A. Weeber H. A. Artal P. Norrby S. (2007). Theoretical comparison of aberration correcting customized and aspheric intraocular lenses. Journal of Refractive Surgery, 23, 374–384. [PubMed] [PubMed]
Posamentier M. T. Abdi H. (2003). Processing faces and facial expressions. Neuropsychology Review, 13, 113–143. [PubMed] [CrossRef] [PubMed]
Ruiz-Soler M. Beltran F. S. (2006). Face perception: An integrative review of the role of spatial frequencies. Psychological Research, 70, 273–292. [PubMed] [Article] [CrossRef] [PubMed]
Slone R. M. Foos D. H. Whiting B. R. Muka E. Rubin D. A. Pilgram T. K. (2000). Assessment of visually lossless irreversible image compression: Comparison of three methods by using an image-comparison workstation. Radiology, 215, 543–553. [PubMed] [Article] [CrossRef] [PubMed]
Vuilleumier P. Armony J. L. Driver J. (2003). Distinct spatial frequency sensitivities for processing faces and emotional expressions. Nature Neuroscience, 6, 624–631. [PubMed] [CrossRef] [PubMed]
Webster M. A. Georgesson M. A. Webster S. M. (2002). Neural adjustments to image blur. Nature Neuroscience, 5, 839–840. [PubMed] [CrossRef] [PubMed]
Yoon G. Williams D. (2002). Visual performance after correcting the monochromatic and chromatic aberrations of the eye. Journal of the Optical Society of America A, Optics, Image Science, and Vision, 19, 266–275. [PubMed] [CrossRef] [PubMed]
Young A. W. McWeeny K. H. Hay D. C. Ellis A. W. (1986). Matching familiar and unfamiliar faces on identity and expression. Psychological Research, 48, 63–68. [PubMed] [CrossRef] [PubMed]
Figure 1
 
Adaptive optics setup with the two synchronized computers, one controlling the AO system (deformable mirror and Hartmann–Shack wave front sensor) and the Badal system the other controlling the psychophysical channel (ViSaGe platform and monitor).
Figure 1
 
Adaptive optics setup with the two synchronized computers, one controlling the AO system (deformable mirror and Hartmann–Shack wave front sensor) and the Badal system the other controlling the psychophysical channel (ViSaGe platform and monitor).
Figure 2
 
Examples of natural scenes from the subjective image sharpness assessment experiment.
Figure 2
 
Examples of natural scenes from the subjective image sharpness assessment experiment.
Figure 3
 
Examples of facial images from the familiar face and facial expression recognition experiments.
Figure 3
 
Examples of facial images from the familiar face and facial expression recognition experiments.
Figure 4
 
RMS wave front error (excluding tilt and defocus) for all 17 subjects before and after AO correction. Data are for 5-mm pupil diameters (except for S8, who had a 4.3-mm pupil diameter).
Figure 4
 
RMS wave front error (excluding tilt and defocus) for all 17 subjects before and after AO correction. Data are for 5-mm pupil diameters (except for S8, who had a 4.3-mm pupil diameter).
Figure 5
 
Percentage of images considered sharper with or without AO for each subject.
Figure 5
 
Percentage of images considered sharper with or without AO for each subject.
Figure 6
 
Correlation between the percentage of images considered sharper with AO and the optical quality improvement in terms of RMS (A) or Strehl ratio at best focus (B) for a 5-mm pupil (4.3-mm pupil for S8).
Figure 6
 
Correlation between the percentage of images considered sharper with AO and the optical quality improvement in terms of RMS (A) or Strehl ratio at best focus (B) for a 5-mm pupil (4.3-mm pupil for S8).
Figure 7
 
Examples of ROC curves, probability of correctly identifying a face as familiar versus the probability of incorrectly identifying an unfamiliar face as familiar, for four subjects. The inset shows the ratio of the area under the curves.
Figure 7
 
Examples of ROC curves, probability of correctly identifying a face as familiar versus the probability of incorrectly identifying an unfamiliar face as familiar, for four subjects. The inset shows the ratio of the area under the curves.
Figure 8
 
Ratio of the area under ROC (AO-corrected/natural aberrations) for familiar face recognition (after 0.5 offset subtraction). The inset shows the difference in areas under ROC (AO—natural aberrations) for familiar face recognition. Error bars correspond to the standard error ( SE) calculated with the statistical analysis. Double stars indicate a significant difference with a 95% level of confidence and single stars for a level of confidence higher than 70%.
Figure 8
 
Ratio of the area under ROC (AO-corrected/natural aberrations) for familiar face recognition (after 0.5 offset subtraction). The inset shows the difference in areas under ROC (AO—natural aberrations) for familiar face recognition. Error bars correspond to the standard error ( SE) calculated with the statistical analysis. Double stars indicate a significant difference with a 95% level of confidence and single stars for a level of confidence higher than 70%.
Figure 9
 
Example of ROC curves, probability of correctly identifying a happy facial expression versus the probability of incorrectly identifying an angry expression as a happy one, for four patients. The inset shows the ratio of the area under the curves.
Figure 9
 
Example of ROC curves, probability of correctly identifying a happy facial expression versus the probability of incorrectly identifying an angry expression as a happy one, for four patients. The inset shows the ratio of the area under the curves.
Figure 10
 
Ratio of the area under ROC (with AO/natural aberrations) for facial expression recognition (after 0.5 offset subtraction). The inset shows the difference in areas under ROC (AO—natural aberrations) for familiar face recognition and errors bar correspond to the Standard Error ( SE) calculated with the statistical analysis. Single stars indicate a significant difference with a level of confidence higher than 70%.
Figure 10
 
Ratio of the area under ROC (with AO/natural aberrations) for facial expression recognition (after 0.5 offset subtraction). The inset shows the difference in areas under ROC (AO—natural aberrations) for familiar face recognition and errors bar correspond to the Standard Error ( SE) calculated with the statistical analysis. Single stars indicate a significant difference with a level of confidence higher than 70%.
Figure 11
 
Comparing familiar face and facial expression recognition in terms of changes in the area under ROC' with AO correction.
Figure 11
 
Comparing familiar face and facial expression recognition in terms of changes in the area under ROC' with AO correction.
Figure 12
 
Correlation between gain (defined in Equation 7) and optical quality improvement (in terms of RMS (AO/noAO) for 5-mm pupil (4.3-mm pupil for S8) for familiar face and facial expression recognition tasks.
Figure 12
 
Correlation between gain (defined in Equation 7) and optical quality improvement (in terms of RMS (AO/noAO) for 5-mm pupil (4.3-mm pupil for S8) for familiar face and facial expression recognition tasks.
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×