Vous êtes sur la page 1sur 44

This article was downloaded by: [Universiti Putra Malaysia]

On: 20 March 2013, At: 23:46


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Journal of Modern Optics
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/tmop20
Adaptive optics and vision
K.M. Hampson
a
a
Bradford School of Optometry & Vision Science, University of
Bradford, Bradford, UK
Version of record first published: 19 Dec 2008.
To cite this article: K.M. Hampson (2008): Adaptive optics and vision, Journal of Modern Optics,
55:21, 3425-3467
To link to this article: http://dx.doi.org/10.1080/09500340802541777
PLEASE SCROLL DOWN FOR ARTICLE
Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-
conditions
This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation
that the contents will be complete or accurate or up to date. The accuracy of any
instructions, formulae, and drug doses should be independently verified with primary
sources. The publisher shall not be liable for any loss, actions, claims, proceedings,
demand, or costs or damages whatsoever or howsoever caused arising directly or
indirectly in connection with or arising out of the use of this material.
Journal of Modern Optics
Vol. 55, No. 21, 10 December 2008, 34253467
TOPICAL REVIEW
Adaptive optics and vision
K.M. Hampson
*
Bradford School of Optometry & Vision Science, University of Bradford, Bradford, UK
(Received 30 June 2008; final version received 7 October 2008)
It has long been recognised that the optical quality of the human eye is far from
diffraction limited. This affects our visual acuity and severely limits the resolution
at which images of the living retina can be obtained. Adaptive optics is
a technique that can correct for the eyes aberrations and provide diffraction
limited resolution. The origins of the technique lie in astronomy, but it was
successfully adapted to the human eye just over 10 years ago. Since then there
have been rapid developments in the field of adaptive optics and vision science.
In vivo images of the retina can now be routinely achieved with unprecedented
resolution. Sophisticated experiments can be performed to gain a deeper
knowledge of the interaction of neural retinal architecture and visual perception.
This article presents the theory behind adaptive optics for the human eye and
reviews the developments in this field to date.
Keywords: adaptive optics; vision
1. Introduction
The human eye can be considered to be a biological adaptive optics (AO) system. An AO
system is essentially one in which the optical element is adapted to correct for the image
blur present at that point in time. Such a system consists of three main elements: the
sensor, corrector and controller. A diagram illustrating the main components of the eye is
shown in Figure 1.
The retina, which contains an array of image sampling elements called photoreceptors,
is the sensor of the eyes AO system. This is nourished by the layers of blood vessels
contained in the choroid. The cornea and lens provide the eye with its refractive power.
The majority of the power is provided by the cornea and is around 40 D. (The focal power
in diopters D is the inverse of the focal length in metres.) The lens has the ability to change
its power over a range of around 10 D so that objects at different distances can be brought
into focus. Hence, the lens can be considered to be the corrector of the eyes AO system.
To focus closer objects the lens power is increased by contraction of the ciliary body
a process known as accommodation. This is illustrated in Figure 2. In an AO system the
ciliary body is analogous to the actuator of the corrective device. The controller of the
system is the brain, which after receiving information via the optic nerve, determines by
how much to alter the power of the lens in order to minimise blur.
*Email: k.m.hampson@bradford.ac.uk
ISSN 09500340 print/ISSN 13623044 online
2008 Taylor & Francis
DOI: 10.1080/09500340802541777
http://www.informaworld.com
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

The performance of the eyes AO system is not perfect. Although the lens aims to
maintain as sharp an image as possible on the retina there are limits as to what shape it can
take on. The main aberration the lens can reduce is that of defocus. However, in the eye
there are a host of other imaging degrading aberrations. These aberrations are often more
severe than in a man-made optical system owing to tilts and decentrations in the optical
Iris
Cornea
Optic nerve
Fovea
Ciliary body
Sclera
Choroid
~ 5
~ 22 mm
Sensor
Controller
Corrector
Retina
Brain
Lens
Figure 1. Schematic of the human eye highlighting the components that are analogous to an
adaptive optics system. (The colour version of this figure is included in the online version of the
journal.)
Contraction of ciliary body
Zonules
Accommodated
Relaxed
Object at 8
~ 10 cm (Young healthy eye)
Figure 2. Process of accommodation. To bring closer objects into focus on the retina, the power of
the lens is increased by contraction of the ciliary body. (The colour version of this figure is included
in the online version of the journal.)
3426 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

components and minor imperfections in their shape. Furthermore, the fovea, which is the
part of the retina that provides the eye with its highest spatial resolution is typically off-
axis by around 5

and the pupil is decentered. The poorness of the eyes optical quality has
long been known. More than a century ago Hermann von Helmholtz, a renowned German
physicist and inventor of the ophthalmoscope, made the following comment about the
quality of the eyes optics:
Now it is not too much to say that if an optician wanted to sell me an instrument with all these
defects I should think myself quite justified in blaming his carelessness in the strongest terms
and giving him back the instrument. [1]
Defocus and astigmatism have been routinely corrected for by spectacle lenses since the
thirteenth and nineteenth centuries, respectively [2]. However, this is not the case for the
additional so-called higher-order monochromatic aberrations that the eye suffers from
such as spherical aberration and coma. Leaving these additional aberrations uncorrected
inevitably limits both the quality of the image formed upon the retina and the resolution at
which the retina can be imaged in vivo. Hence, reducing visual acuity and limiting our
ability to monitor the development and progression of retinal disease in the living eye at
the cellular level. Attempts have been made to overcome the effect of these additional
aberrations somewhat using customised contact lenses and refractive surgery. A review of
these techniques can be found in [3,4]. These methods, however, impart only a static
correction. The eye is a biological optical system in which the aberrations are continually
fluctuating on small times scales due to a variety of factors such as blood flow [5,6]. Hence,
in order to maintain a perfect optical correction, a dynamic optical component is required.
In effect another more effective AO system is needed to assist that of the eyes.
The origins of man-made AO systems lie in astronomy. The idea was first proposed in
1953 by Babcock to compensate in real time for the deleterious effects of atmospheric
turbulence on image quality in ground-based telescopes [7]. Owing to the lack of
appropriate technology, it was not until 1977 that an AO system for astronomical
applications was first realised in practice [8]. AO was first applied to the human eye in 1989
by Dreher and colleagues [9]. They used a deformable mirror to improve the resolution of
a scanning laser ophthalmoscope. The mirror was used to impart only a static correction of
astigmatism in one subjects eye, however, based on their spectacle prescription. Although
resolution was improved, Artal and Navarro were the first to demonstrate that it was
possible to obtain objective information of the spacing within the photoreceptor mosaic
in vivo [10]. This was achieved by processing images of specular reflection. The first direct
observations of the photoreceptor mosaic were obtained by Miller and colleagues
using a high-resolution fundus camera [11]. Defocus and astigmatism were minimised using
trial lenses but higher-order aberrations remained uncorrected. Subsequently photo-
receptors were resolved only in young individuals with good optical quality.
Five years after the first implementation of AO in the eye by Dreher and colleagues,
Liang et al. demonstrated the feasibility of the ShackHartmann wavefront sensor for
ocular aberration measurements [12]. This proved to be a key development in the field of
AO in the human eye. This simple objective device had the capability of rapidly measuring
the eyes higher-order wavefront aberrations with great accuracy, which is the first key
step in AO. An early pioneer in capturing the eyes wavefront was Smirnov [13].
Using a subjective vernier-type task Smirnov was able to characterise the higher-order
Journal of Modern Optics 3427
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

aberrations of the eye. He recognised that in principle it was possible to improve the
optical quality of the eye by correcting the calculated wavefront using contact lenses.
However, the method and analysis of the data proved extremely time consuming.
Following his work many investigators developed a variety of methods to more rapidly
and accurately capture the eyes wavefront [14]. Aside from its simplicity, the main
advantage of the ShackHartmann sensor was that it was already a well-established
technique used in astronomical AO systems, making it a natural device of choice for AO in
the eye.
The ShackHartmann sensor was first incorporated into an AOsystemfor vision science
in 1997 by Liang and colleagues [15]. Using a deformable mirror they were able to correct
the higher-order aberrations of the eyes of their subjects. Retinal images with unprecedented
resolution capable of resolving individual photoreceptors were readily obtained and the
benefit to vision of correcting higher-order aberrations was demonstrated. This system,
however, did not track the dynamic changes in the aberrations. Such systems began to
emerge in 2001 [16], and as a result further improvements in retinal image quality were
demonstrated [17]. Since that time the field of AO in vision science has been rapidly growing
with more and more systems being developed. Systems are becoming faster and achieving
better and better aberration correction. They are also becoming more compact and less
expensive, allowing the technology to move towards being incorporated in a clinical setting.
The aim of this review is to give both an overview of the theory behind AO in the eye
and the current developments to date. This article first presents in more detail the optical
properties of the human eye and the requirements and benefit of AO. Then the principles
of AO are discussed. The main components used in systems are presented and their
advantages and disadvantages are outlined. Following this, AO is discussed in the context
of retinal imaging. Specifically the different imaging modalities are presented and the
current performance of systems to date. The knowledge investigators have gained about
the retina through imaging with AO is also presented. AO is then discussed in the context
of psychophysical applications and the results of experiments performed so far are given.
Following this other applications of AO are presented and finally the future of AO in the
eye is discussed.
2. Properties of aberrations in the eye
Like any optical system, the resolution of the eye is determined by both
diffraction and aberrations. The goal of any AO system is to correct the aberrations
so that resolution is limited by diffraction only. Before constructing an AO system for the
eye it is of importance to know the properties of the aberrations. This will determine
factors such as the spatial resolution, dynamic range, and speed of both the sensor and
corrector that is required to reach and maintain the diffraction limit.
The Rayleigh resolution criterion states that for a diffraction limited system, two point
sources can just be resolved if the peak of the image of one lies on the first minimum of
the other as illustrated in Figure 3. This distance is effectively equal to the width of the
diffraction-limited intensity point spread function (PSF) and is given by
PSF
width

1.22zf
nD
,
1
3428 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

where z is the wavelength, f is the focal length, n is the refractive index and D is the pupil
diameter. Hence, for a fixed wavelength, the larger the pupil the smaller the width of the
PSF and the higher the resolution.
Figure 4 shows how the width of the PSF varies with pupil diameter D, taking z as
550 nm, f as 22.2 mm and n as 1.33. The cone photoreceptors, which provide us with our
colour vision, are separated by around 2 mm at the fovea [18]. For an eye limited only by
2 3 4 5 6 7 8
1
1.5
2
2.5
3
3.5
4
4.5
5
5.5
6
Pupil diameter (mm)
P
S
F
w
i
d
t
h

(

m
)
Spacing & diameter
of foveal cones
Actual diffraction
limit of eye
Figure 4. Variation in the width of the PSF with pupil diameter, for a diffraction-limited eye and
wavelength of 550 nm. Increasing pupil diameter decreases the effective width of the PSF and so
increases resolution. (The colour version of this figure is included in the online version of the
journal.)
Minimum separation to be resolved
as two separate sources
PSF width
Figure 3. Rayleigh resolution criterion. In a diffraction-limited system two point sources can be
resolved if they are separated by the effective width of their images. (The colour version of this figure
is included in the online version of the journal.)
Journal of Modern Optics 3429
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

diffraction the pupil needs to be greater than around 5.5 mm in diameter for these to be
resolved. However, studies have shown that the eye approaches the diffraction limit for
a pupil diameter of around 3 mm, but beyond this image quality is limited by aberrations
[19]. Hence, in order to image foveal cones reliably in a human eye of typical optical quality,
an AO system is required. Owing to technical limitations AO systems never achieve
a perfect correction and reduce the aberrations to zero. However, a system can be
considered diffraction limited if the Strehl ratio reaches 0.8. This is the ratio of the
aberrated PSF height to the PSF height of an equivalent diffraction-limited eye. The Strehl
ratio of a typical human eye with only defocus and astigmatism corrected can be less than
0.1 [19].
There are two main classes of aberrations: monochromatic and chromatic.
Monochromatic aberrations are considered to be those that occur as a result of rays
having different optical path lengths for different locations in the pupil. This is due to
variations in refractive index and optical irregularities. Although their magnitude can
depend upon wavelength, they are distinct from chromatic aberrations which are evident
in polychromatic light. Chromatic aberrations are due to dispersion, i.e. the dependence of
refractive index on wavelength. AO manipulates monochromatic aberrations and so the
main focus of this section is on the properties of these types of aberrations. Chromatic
aberrations and their effect on AO will also be discussed briefly.
2.1. Monochromatic aberrations
Monochromatic aberrations are described by the monochromatic wave aberration
function W. This determines the deviation of the actual wavefront from some perfect
wavefront at each point in the pupil. Owing to the inaccessibility of the eyes image space
the aberrations are defined in object space. In this case, for an eye of perfect optical
quality, the emerging wavefront will be plane, and so the geometry of the wave aberration
function is as shown in Figure 5. If the wavefront is phase advanced W is positive, and
negative if it is phase retarded. In order to reduce the optical aberrations in the eye AO acts
on this wavefront.
2.1.1. Zernike polynomials
The aberrated wavefront can be described by a sum of polynomials. The accepted
convention is to use Zernike polynomials as they are orthogonal over the unit circle.
Principal components analysis has demonstrated them to be an efficient basis for
describing the eyes wave aberration function [20]. There are several variations on the way
in which these polynomials are formulated with regards to the numbering, normalisation,
and angle convention. The convention used in ophthalmic optics is that recommended by
the Optical Society of America Taskforce [21]. The wavefront can be written as
W,,
X
1
i0
a
i
Z
i
,, , 2
where , is the radial coordinate ranging from 0 to 1, is the azimuthal component
ranging from 0 to 2, and a
i
is the coefficient of the Zernike polynomial Z
i
usually
3430 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

specified in micrometers. The angle convention follows that used in clinical practice when
refracting a patient. Each Zernike polynomial is commonly identified using a double
indexing scheme where n is the highest power (order) of the radial polynomial, and m
describes the azimuthal frequency of the sinusoidal component.
Z
m
n
,,
N
m
n
R
jmj
n
, cos m; for m ! 0
N
m
n
R
jmj
n
, sinm; for m50
( )
3
where R
jmj
n
, is the radial dependent component given by
R
jmj
n
,
X
njmj,2
s0
1
s
n s!
s!0.5n jmj s!0.5n jmj s!
,
n2s
4
and the normalisation factor N
jmj
n
is given by
N
jmj
n

2n 1
1 o
m0

1,2
,
5
o
nm

1, if m 0,
0, otherwise,

6
m takes the values n, n 2, n 4, . . . , n.
The Zernike polynomials up to and including fifth radial order are shown in Figure 6.
Modes with a radial order up to two are commonly referred to as low-order modes and are
routinely corrected for using conventional methods such as spectacle lenses. The primary
Pupil plane
Z
Aberrated wavefront
Plane wavefront
Retina
W
Figure 5. Geometry of the monochromatic wave aberration function for the human eye.
The wavefront is the deviation from a plane wave measured at each point in the pupil plane.
(The colour version of this figure is included in the online version of the journal.)
Journal of Modern Optics 3431
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

goal of AO is to correct for the so-called higher-order terms. Tip and tilt are analogous to
prismatic corrections. Spectacle parameters are calculated from the second-order terms as
follows:
Sphere
43
1,2
a
0
2
26
1,2
a
2
2

2
a
2
2

2
h i
1,2
R
2
p
,
7
Cylinder
46
1,2
a
2
2

2
a
2
2

2
h i
1,2
R
2
p
,
8
Axis
arctana
2
2
,a
2
2

2
.
9
A common metric to describe the level of aberrations present is to use the rms
wavefront error given by
rms
X
N
i3
a
2
i
!
1,2
.
10
Figure 6. Pyramid of Zernike polynomials used to describe the monochromatic wavefront
aberrations of the eye. Polynomials up to and including fifth radial order are included.
(The colour version of this figure is included in the online version of the journal.)
3432 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

When determining the optical quality from the rms wavefront error, tip and tilt are often
removed as they merely represent a shift in the image which is of no consequence to its
quality. The residual rms wavefront error is often the number quoted to signify the spatial
performance of an AO system. An approximation of the Strehl ratio s based on the rms
wavefront error is given by
s exp2po
2
, 11
where o is the rms wavefront error in waves. In order to reach a Strehl ratio of 0.8 and
so consider the system to be diffraction limited
rms 5
z
14
. 12
This is referred to as the Mare chal criterion [22]. For a wavelength of 550 nm this is equal
to an rms of 0.04 mm.
2.1.2. Aberrations in the population
The level of higher-order aberrations depends upon a variety of factors. They increase with
pupil size [23]. They also vary with age [23] and accommodation level [24]. There are also
reports on variations with refractive error. A review of aberrations and myopia can be
found in [25]. There have been several studies carried out to determine the distribution of
Zernikes in a large population such as [20,26]. Plainis and Pallikaris carried out a recent
population study where their subjects were restricted to emmetropes (i.e. those with
minimal defocus or astigmatism). They found the average higher-order rms wavefront
error across 393 eyes measured across a 6 mm pupil to be 0.26 mm [27]. This corresponds to
an equivalent defocus of 0.2 D. The age range of the subjects was 33 5 years and the
measurements were carried out with the eye focussed at infinity. They found a population
average of zero for all Zernike terms except for spherical aberration and oblique trefoil.
Like other reports they also found bilateral symmetry between the left and right eyes of the
same subject. Even though the subjects were emmetropic, second-order terms were the
most dominant. Again confirming the results of previous studies.
2.1.3. Aberration dynamics
When all other factors are fixed, i.e. when a subject is fixating steadily on a stationary
target, all aberrations display dynamic behaviour. The magnitude of the fluctuations are of
the order of fractions of a micrometer. The dynamics of the higher-order aberrations were
first characterised by Hofer and colleagues in 2001 [28]. Prior to this only microfluctua-
tions in focus had been charactersied. They found measurable fluctuations out to around
56 Hz. Based on these results they suggested that an AO system capable of correcting
fluctuations of up to 12 Hz would be sufficient to yield diffraction limited resolution over
a dilated pupil. Using a much faster sensor, Diaz-Santana and colleagues found
measurable power out to around 30 Hz and demonstrated the advantage of an AO
system which is capable of correcting aberration fluctuations up to this higher
frequency [29]. Figure 7 shows the power spectral density function of the rms wavefront
error for one subject focussed at infinity, measured over a 4.2 mm pupil at a rate of 20 Hz.
Journal of Modern Optics 3433
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

As can be seen the power spectrum is approximately a straight line with a slope of 1.5 as
is found for most subjects [28].
Several sources have been identified as contributing to the dynamic behaviour of the
aberrations such as the cardiopulmonary system [5,6]. Many studies have also
demonstrated an effect of tear film fluctuations, see for example [30]. However, these
factors cannot fully explain the dynamic properties of the aberrations and so there are
others yet to be determined. Possibilities include small fixational eye movements known as
microtremor [31], and fluctuations in pupil size [32]. Ongoing efforts are being made to
both model the aberration dynamics, see for example [33], and also generate them using
a model eye [34]. Such efforts will assist in optimising the dynamic performance of AO
systems.
2.1.4. Wavelength dependence
The monochromatic aberrations are also affected by wavelength. This has been investigated
by several research groups. Two groups to date have developed polychromatic Shack
Hartmann wavefront sensors capable of measuring the eyes monochromatic aberrations
across a range of wavelengths in the visible spectrum [35,36]. The main effect of wavelength
is on the Zernike defocus term, although other aberrations are also minimally affected.
The effect of wavelength is an important parameter to consider. Often the light used for
wavefront sensing is in the near-infrared region. But when carrying out psychophysical
experiments, for example, the stimulus will be in the visible region of the spectrum. This
difference needs to be accounted for. Using the ShackHartmann sensor the difference in
these two light regimes has been found to be equivalent to around 0.72 D [37].
2.1.5. Off-axis aberrations and anisoplanatism
Numerous studies have demonstrated that aberrations vary with retinal eccentricity, see for
example [38,39]. The area over which the aberrations remain relatively constant is called the
isoplanatic patch. Bedggood and colleagues have found that it is around 0.8

at the fovea
[40]. This has important consequences for conventional AO as this means that
when imaging the retina for example, a clear image can only be obtained over this limited
region. Strategies are being developed to try to overcome this limitation. These shall be
discussed later.
Log frequency (Hz)
10
Slope ~ 1.5
L
o
g

p
o
w
e
r

(

m
2
/
H
z
)
Figure 7. Typical rms wavefront error power spectral density function for one subject measured over
a 4.2 mm pupil. (The colour version of this figure is included in the online version of the journal.)
3434 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

2.2. Chromatic aberration
The two types of chromatic aberration are shown in Figure 8. Longitudinal chromatic
aberration is effectively the variation in focus with wavelength. This is in the range of 2 D
across the visible spectrum. Transverse chromatic aberration is the variation with
wavelength of the location of the image in the image plane. It can be thought of as the
wavelength dependence of tip/tilt. This is variable across subjects.
In terms of using AO to improve visual performance, the benefit is limited by the fact
that the world we view around us is polychromatic. The impact of chromatic aberration
when performing retinal imaging with AO depends upon the imaging modality used. If
a broad-band light source is required then chromatic aberration needs to be minimised.
This can be achieved with a specially designed lens.
2.3. Other factors affecting image quality
Light scatter is another issue that reduces image quality. This occurs from both the ocular
media and the retina. Although this is not a major issue in a normal young eye, it is when
scatter is increased owing to refractive surgery or increasing age. This will decrease the
performance of the imaging system and also visual performance.
3. Principles of adaptive optics
An AO system consists of three main components: the sensor, the corrector, and the
controller. A general AO system is shown in Figure 9. A laser is used to form a small point
source on the retina. This light is reflected and passes back through the eyes pupil where it
is aberrated. This is the light source for the wavefront sensor, which is in a plane conjugate
to the eyes pupil. From the sensing data the control computer determines the signals
required to drive the corrector, which is also in a conjugate pupil plane.
3.1. Sensor
Wavefront sensing is the first key step in AO. There are several parameters that need to be
considered when choosing a sensor. The first is the speed of the sensor. It needs to be fast
enough to capture the aberration dynamics if one wants to correct for them in real time.
Hence, it needs to sample at around 2030 Hz. This inevitably requires the sensor to be
objective. It must also have enough spatial resolution to accurately capture the wavefront,
(a) (b)
Longitudinal shift
in image
Transverse shift
in image
Figure 8. Chromatic aberration. (a) Longitudinal chromatic aberration. (b) Transverse chromatic
aberration. (The colour version of this figure is included in the online version of the journal.)
Journal of Modern Optics 3435
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

as the corrector cannot correct what the sensor does not measure. Other issues for
consideration are the dynamic range, which describes the largest wavefront that can be
measured and the sensitivity. Often increasing the dynamic range can result in a decrease
in sensitivity and so there is generally a trade-off between these two parameters.
Another issue for consideration is the choice of light source used for the wavefront
sensing light. As AO manipulates the monochromatic aberrations, a narrow bandwidth
source such as a laser or superluminescent diode (SLD) is used. The preferred wavelength
is in the near infrared region as this has several advantages over wavelengths in the visible
region of the spectrum. The retina is less sensitive to light in this region and so it is more
comfortable for the subject. As infrared is less visible, it is also an advantage when carrying
out psychophysical experiments as it interferes less with the detection of the stimulus. It is
important that enough light is reflected back from the retina so that accurate
measurements of the wavefront can be obtained that are not dominated by noise from
the detector for example. However, there are limits as to how much light can safely enter
the eye. Near infrared is less damaging to the retina which means that more light can be
safely put into the eye and more light is available for wavefront sensing. Also the retina
reflects more light in this region and there is less scattering from the anterior optics.
Typical powers used are on the order of tens of microwatts.
There are a variety of methods that are capable of measuring the eyes wavefront. To
date, however, only two types of sensor have been used for ophthalmic AO systems: the
ShackHartmann sensor and the pyramid sensor.
3.1.1. ShackHartmann sensor
Virtually all AO systems for the human eye in existence today use the ShackHartmann
sensor as the wavefront sensing device. It is available commercially from several
ophthalmic companies such as Wavefront Sciences. The demonstration of its feasibility for
(2) Corrector
Laser
P
P
P
P = Pupil plane
(1) Wavefront sensor
(3) Control computer
Imaging camera/
stimulus
CCD
Eye
Figure 9. General AO system showing how the three main components: the sensor, corrector and
controller, are coupled together. (The colour version of this figure is included in the online version of
the journal.)
3436 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

use in the human eye in 1994, by Liang and colleagues, is considered the cornerstone in the
development of AO for vision science [12]. It was the first simple objective technique
capable of making rapid and precise measurements of the eyes higher-order aberrations.
The sensor consists of an array of lenslets placed in a plane conjugate to the eyes pupil and
a detector at the focal plane of the array. The lenslets are normally square and so
contiguous to maximise efficiency, and the detector is normally a CCD camera. The array
effectively samples the slope of the wavefront at discrete intervals across the entire pupil of
the eye as shown in Figure 10. If the wavefront is plane, as it would be in the case of an
optically perfect eye, a regular array of spots will be formed on the camera. If the
wavefront is aberrated, each spot will be displaced according to the slope of the wavefront
across its corresponding lenslet. By measuring the x and y shift in position of each spot
relative to the positions for an aberration free eye, the wavefront across the pupil can be
determined. The location of the spot is determined by its centroid.
The slope of the wavefront in the x direction across a given lenslet is given by
oW
ox

W
x
a

x
f
, 13
f
L.A.
(a)
Aberration-free eye
Aberrated eye
(b)
Pupil plane CCD plane
Figure 10. Principle of the ShackHartmann sensor. A lenslet array at a plane conjugate to the
pupil measures the wavefront slope at discrete locations. (a) For an aberration free eye a regular
array of spots will be formed at the focal plane. (b) An irregular array will be formed for an
aberrated eye. (The colour version of this figure is included in the online version of the journal.)
Journal of Modern Optics 3437
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

where a is the diameter of the lenslet and f is the focal length. This is illustrated in
Figure 11. When locating the spots using a conventional algorithm, each spot is searched
for over a virtual aperture directly behind its corresponding lenslet and equal to the size of
the lenslet. Hence, the dynamic range of the sensor is limited by the maximum amount
a spot is able to move before crossing over into the next so-called search-block. This is
equal to half the diameter of a lenslet. From Equation (13)
W
max

a
2
2f
. 14
The sensitivity is the minimum slope that can be measured and is given by
W
min

x
min
a
f
. 15
The minimum spot displacement x
min
is determined by a variety of factors such as light
level and the specifics of the centroiding algorithm. Hence, for a given lenslet diameter,
decreasing the focal length increases the dynamic range but decreases the measurement
sensitivity. For a given focal length, decreasing the lenslet diameter decreases the dynamic
range but increases the measurement sensitivity. Hence, dynamic range and sensitivity are
inversely proportional to each other. Typical focal lengths available range from around
a few millimeters to 30 mm, and diameters are normally 100600 mm [41].
During the initial wavefront sensing measurement, i.e. before the AO system starts
correcting, a large dynamic range is of importance so that the wavefront can adequately be
captured. Once the AO system is in operation, the spots will be pulled close to a regular
grid and sensitivity then becomes the most important issue. Hence, ideally an adaptable
lenslet array is required. Seifert and colleagues have demonstrated such an array using
a liquid crystal display device which can generate Fresnel lenses of varying focal
length [42]. This has yet to be implemented in an AO system, however.
The spatial resolution of the sensor is determined by how many lenslets are sampling
the pupil. It has been estimated that the number of lenslets should be equal to the number
f
a
Wx
x
Figure 11. Relationship between the wavefront across one lenslet and the shift in the position of
the ShackHartmann spot. (The colour version of this figure is included in the online version of
the journal.)
3438 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

of Zernikes that are to be constructed [41]. Liang and colleagues have proposed that in
order to reach the diffraction limit for a 7.3 mm dilated pupil, Zernike modes up to and
including at least 8th radial order should be corrected for [19]. To accurately measure this
would therefore require 42 lenslets in the pupil. The speed of the ShackHartmann sensor
is primarily determined by the speed of the camera. Many high speed cameras are
available. However, if there is much light loss through the system then this will place
a limit on the minimum exposure which will in turn limit speed.
When using a highly coherent source, speckle arises from the interference of the light
after scattering from the rough surface of the retina. Hofer and colleagues have shown that
this causes a multiplicative increase in the power spectrum of the rms wavefront error [28].
They demonstrated a way to reduce this by obtaining a spatial average of the retina using
a high-speed scanner placed in a plane conjugate to the eyes pupil. This is illustrated in
Figure 12. On the return path from the eye the light is descanned such that the position of
the ShackHartmann spots are unaffected. The scan angle used is less than 1

to avoid the
effects of anisoplanatism.
When carrying out psychophysical experiments using AO it is necessary for the
stimulus light to bypass the scanner. This can be achieved using two cold mirrors which
reflect visible light but pass infrared light [43]. These are arranged at 45

to the beam as
shown in Figure 13. Figure 14 demonstrates the benefit of the scanning method to the
ShackHartmann spots. To avoid the non-common path error, an alternative method is to
use a rotating diffuser to reduce the beam coherence before it enters the system [44]. A less
coherent light source such as an SLD further reduces speckle.
The beam entering the eye is usually less than 1 mm in diameter. This is to provide
a diffraction-limited spot on the retina, which means the quality of the ShackHartmann
spots will be good and in principle they can be located with better accuracy. Normally the
beam enters the eye slightly off-axis so an aperture placed in a conjugate retinal plane can
be used to prevent corneal reflections from reaching the sensor [28].
3.1.2. Pyramid sensor
The pyramid sensor is a relatively new wavefront sensor suggested by Ragazzoni in 1996
for use in astronomy [45]. It also measures the slope of the wavefront and is proposed to
have a higher sensitivity than the ShackHartmann sensor. The operation of the pyramid
sensor is based on the principle of the Focault knife-edge test, which is illustrated in
Figure 15. Instead of knife-edges the pyramid sensor consists of a four-faceted glass
pyramid. In this case no rays are blocked and so it is more efficient. Four images of the
Pupil plane Scanner
Retinal plane
Figure 12. Principle of using a scanner to reduce speckle by obtaining a spatial average of the retina.
(The colour version of this figure is included in the online version of the journal.)
Journal of Modern Optics 3439
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

pupil are simultaneously formed, each effectively corresponding to a knife-edge in
a quarter of the pupil plane. The beam rotates around the tip of the pyramid. This is
achieved by either moving the pyramid back and forth along the axis, or the beam is
rotated using a steering mirror. The gain, i.e the change in the measured signal for a given
wavefront, depends upon the amplitude of rotations and so it can easily be changed
on-line. This is the major advantage of the pyramid sensor as compared to the Shack
Hartmann sensor.
It was first applied to the human eye in 2002 by Iglesias et al. [46]. They implemented
it using an extended source, which avoids having to oscillate the beam around the tip of
Scanner
Cold mirror
Cold mirror
Plane mirror
Infrared
To Eye
Visible
Figure 13. Method to bypass the scanner for the stimulus light. Two cold mirrors are used to
separate the visible light from the stimulus and the infrared light for the wavefront sensing.
(The colour version of this figure is included in the online version of the journal.)
(a)
(b)
Figure 14. The effect of scanning on the ShackHartmann spots. (a) The scanner is off.
Speckle is very evident in the spots. (b) The scanner is turned on and speckle has been significantly
reduced.
3440 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

the pyramid. The feasibility of the sensor for an AO system for the eye was demonstrated
in 2006 [47]. The authors suggest that this type of sensor can improve the performance of
AO systems for the eye.
3.2. Corrector
Once the wavefront has been measured it can then be corrected. The two main types of
correction devices are deformable mirrors and liquid crystal spatial light modulators
(LC-SLMs). Both types of device work on the principle of phase conjugation. An optical
beam can be described by its electric field
E Aexpi , 16
where A is the amplitude and is the phase. The corrector imparts a reversed phase onto
the beam to compensate for the phase distortions. This is mathematically equivalent to
reversing the sign in the exponential part of E and so taking the complex conjugate. Hence,
the term phase conjugation. The phase is given by

2p OPL
z
,
17
where z is the wavelength and OPL is the optical path length given by
OPL n z. 18
Consequently the phase can be modulated by manipulating either the refractive index n as
in the case of LC-SLMs, or by changing the physical distance travelled z, as in the case of
deformable mirrors. This is illustrated in Figure 16.
When selecting a corrective element, several factors are taken into account. As with the
sensor, speed, spatial resolution and dynamic range are of concern. Spatial resolution is
primarily governed by the number of actuators, or in the case of LC-SLMs the number
(a)
(b)
1
2
1 2
Figure 15. Focault knife-edge test. (a) For an aberration free system moving a knife edge from either
side perpendicular to the optical axis will result in the pupil image becoming dark. (b) For a system
with defocus the image of the pupil will be different depending on whether the knife-edge is above or
below the axis. (The colour version of this figure is included in the online version of the journal.)
Journal of Modern Optics 3441
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

of pixels. In general the higher the number of actuators/pixels the more spatially intricate
the wavefront that can be corrected for. Hence, in principle, the larger the number of
Zernike modes the device can correct for, provided the device also has a large enough
dynamic range. Dynamic range is determined by the stroke of the corrective device. In the
case of deformable mirrors for example, this is the magnitude of the maximum surface
deflection. Even if the device has a large number of actuators or pixels, if the stroke of the
device is too small much of it will be used up in correcting the lower order modes. Further
requirements are that the device is relatively small, being comparable to the size of the
pupil, so that it can be incorporated into compact systems. Many researchers consider the
corrective element to be the main component that limits the performance of an AO system.
The following section discusses the properties of the main types of correctors that have
been implemented for AO systems in the eye.
3.2.1. Deformable mirrors
Deformable mirrors are the most common correction devices. They essentially consist of
a mirrored surface whose shape is deformed by actuators. In order to correct for an
aberrated wavefront the mirror deforms into the same shape as the wavefront but with half
the amplitude, as the extra physical path is introduced into both the incident and reflected
path. This is illustrated in Figure 17. Speed is generally not an issue with deformable
mirrors as they can respond with speeds several orders of magnitude faster than the
aberration dynamics.
Deformable mirrors can be broadly categorised into two classes: segmented surface
and continuous surface. Segmented deformable mirrors consist of individual plane mirrors
each attached to a separate actuator as shown in Figure 18. Some devices are piston only
where each segment can only be moved perpendicular to the mirror plane, others can also
be tipped and tilted and so require three actuators per segment. The disadvantages of
segmented mirrors are that there are gaps between the segments. This leads to light loss
and diffraction effects. The percentage of the surface area of the corrector covered by
mirrors is known as the fill factor, which can approach 100%. The segments can be square
or hexagonal. As each segment can be adjusted independently, these mirrors are well suited
Aberrated wave
Plane wave
Plane wave
(a)
(b)
Figure 16. Principle of wavefront correction by manipulating the optical path length using two
methods. (a) Changing the physical path length using a deformable mirror. (b) Changing the path
length by altering the refractive index using a liquid crystal device. (The colour version of this figure
is included in the online version of the journal.)
3442 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

to correcting modes with higher spatial frequencies. They are not as well suited to lower-
order modes where a smooth approximation to the wavefront is required. However, this
is not such an issue for the piston tip tilt mirrors. Currently segmented mirrors can
be extremely small and have a high density of actuators. Boston Micromachines
Corporation for example has a segmented deformable mirror with 140 actuators in
a 4.9 mm diameter aperture. The segments are actuated based on the principle of
electrostatic attraction which means the mirror exhibits no hysteresis. The stroke of this
device is 5.5 mm.
The first deformable mirrors used in vision science were of the continuous surface type.
Typically with these types of mirrors, when an actuator is moved the surface deflection is
not restricted to the area directly above that given actuator. This phenomenon is known as
coupling. Hence, these mirrors are generally more suited to correcting modes of a lower
order. The four main types of continuous surface mirror are shown in Figure 19.
Mirrors of the type shown in Figure 19(a) normally consist of actuators that expand
when a voltage is applied. Examples of this type of mirror are the 97-actuator mirror from
Xinetics Inc. used by the group at Rochester. This has a stroke of around 4 mm. Another
common option is the 37-actuator piezoelectric deformable mirror produced by Flexible
Optical BV. Although having less actuators, it has a smaller diameter of 3 cm and a larger
stroke of 6 mm.
Incoming aberrated wave Reflected wave
a
a/2
Figure 17. Principle of correction achieved via a deformable mirror. The incoming aberrated
wavefront is phase retarded by an amount a at its centre. The advanced part of the wavefront is
made to travel an extra distance a by the deformable mirror. And so upon reflection the wavefront
becomes plane and effectively aberration free. (The colour version of this figure is included in the
online version of the journal.)
(a) (b)
Figure 18. Segmented deformable mirrors. (a) Piston only. (b) Piston and tip and tilt. (The colour
version of this figure is included in the online version of the journal.)
Journal of Modern Optics 3443
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

Bimorph mirrors consist of two or more layers of material bonded together as shown in
Figure 19(b). They work on a similar principle to a bimetal strip. When a voltage is applied
to an electrode the piezoelectric material expands and the curvature of the mirror surface is
affected. They usually have a large stroke of greater than 10 mm, but are better suited to the
lower order modes than the mirrors mentioned above. Typical devices that have been used
for vision science include the mirror from AOptix, which has 35 actuators and a diameter
of 15 mm.
Figure 19(c) is an example of a membrane mirror manufactured by OKO
Technologies. It consists of a thin metallic membrane suspended over an array of
actuators. Membrane mirrors work on the principle of electrostatic attraction and hence
exhibit no hysteresis. These types of mirrors when used for vision science applications
typically have 37-actuators and a pupil diameter of 15 mm. The maximum peak-to-valley
is around 9 mm. As the mirror is clamped at its edges it is desirable to use only 65% of the
surface [48]. The membrane mirror is an example of a MEMS (microelectromechanical
systems) mirror. These type of mirrors are fabricated using well established techniques
used in the integrated circuit industry making them lower cost. It must be noted that not
all MEMS mirrors are edge-clamped.
Figure 19(d) is one of the most recent additions to mirrors for vision science and is
currently only available from Imagine Eyes. The mirror has a considerable stroke of
50 mm peak-to-valley. It has 52 actuators over a 15 mm diameter pupil. Voltages are
applied to coils which produce a magnetic field. This attracts or repels magnets attached to
the mirror surface depending on the sign of the voltage. Owing to its size, large stroke and
relatively low cost, this type of mirror shows great promise.
All of the above mirrors have their trade-offs and all have been successfully integrated
into AO systems for vision science. Several investigators have modeled the performance of
the various mirror types for use in the eye (see for example [49,50]). The predicted stroke
and number of actuators required to reach diffraction-limited performance depends on the
particular mirror. Strokes as high as 53 mm and as many as 90 actuators across the pupil
have been deemed necessary [51].
3.2.2. Liquid crystal spatial light modulators
The effect of LC-SLMs on the wavefront is similar to that of piston segmented deformable
mirrors. As they are based on existing LCD technology they are generally of lower cost
(a)
(d )
(b)
V
V
(c)
V I
Figure 19. Types of continuous surface mirrors. (a) Mirror actuated by actuators that expand or
contract. (b) Bimorph mirror. (c) Membrane mirror. (d) Magnetic deformable mirror. (The colour
version of this figure is included in the online version of the journal.)
3444 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

than deformable mirrors. The refractive index can be changed in one of two ways. Either
optically, which involves imaging an intensity pattern onto the device, or electronically via
the use of electrodes. Devices are available in either transparent or reflective mode. They
operate based on neumatic liquid crystals. An example of a reflective device is shown in
Figure 20. When a voltage is applied to an electrode liquid crystal molecules change their
alignment which alters the refractive index and so modulates the phase of the light.
Several investigations have been carried out to determine the performance of various
LC-SLMs in the eye, see for example [5254]. Their stroke is typically around one
wavelength but this can be extended using modulo-2 phase wrapping. Their spatial
resolution is extremely high for example XGA devices have 1024 768 pixels and the
liquid crystal is relatively small. A disadvantage of LC-SLMs is the need for linearly
polarised light as the liquid crystal molecules only phase modulate light polarised along
their axis. When using a highly polarised light source much is lost as much of the light
reflected from the retina is depolarised. Fortunately, Marcos and colleagues have found
that the state of polarisation has mimimal effect on the ShackHartmann measurements
[55]. Another issue of concern has been their relatively low speed. However, the new device
from Hamamatsu quotes a frame rate of 60 Hz.
3.2.3. Strategies for enhancing correction capabilities
Once a corrective device has been selected there are several strategies that can be employed
to enhance the aberration correction ability of the system. The largest aberrations in the eye
are defocus and astigmatism. It is desirable to correct these aberrations using some other
Liquid crystal
Cover glass
Reflective electrodes
Figure 20. Principle of operation of a neumatic liquid crystal spatial light modulator. Applying
a voltage to the electrodes causes the liquid crystal molecules to change their alignment. The result is
a change in refractive index. (The colour version of this figure is included in the online version of the
journal.)
Journal of Modern Optics 3445
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

method rather than with a deformable mirror or LC-SLM in order that the correction
capabilities of these devices are reserved for the higher-order aberrations. Commonly the
subjects wear their spectacles or trial lenses are inserted into the system. An alternative
method to correct defocus is to use a Badal optometer. The Badal optometer consists of
a lens placed at a focal length away from the eye. A linear relationship exists between the
position of object and the refractive power induced at the eye. It is implemented in an AO
system as shown in Figure 21. The position of the focus is changed by two mirrors and
consequently the image of the pupil plane remains fixed. A continuously variable
cylindrical correction can be added to this by using two rotating cylinders in contact placed
at the pupil plane. Such a system can be calibrated with the wavefront sensor allowing for
a fully automated correction [6]. Another possible option to correct both defocus and
astigmatism, and perhaps more simple than the Badal optometer and rotating cylinders, is
the Alvarez lens pair [56]. It consists of two elements with complex but complementary
surface shapes. Defocus and astigmatism is varied linearly by sliding the elements relative to
each other in the x and y direction and rotating them. These have not as yet been utilised in
an AO system, however. It must be noted that these aforementioned procedures only
impart a static correction of defocus and astigmatism. The dynamic component of these
aberrations are left to the deformable mirror or SLM to correct.
To enhance both low and higher-order aberration correction another option is to
cascade two deformable mirrors [57,58]. Chen and colleagues use a Bimorph mirror to
remove as many of the low-order modes as possible and use a segmented mirror to remove
the remaining higher-order modes [57].
One way to amplify the stroke of a deformable mirror is to have the light strike the
mirror twice as shown in Figure 22. This effectively doubles the stroke of the mirror and
has been demonstrated by Webb and colleagues [59]. It has been employed in a vision
science system for accommodation studies by Hampson and colleagues [44]. It is a very
cost effective way to greatly increase stroke, but spatial resolution is unaffected as this is
limited by the number of actuators.
3.3. Controller
The controller is essentially the computer which provides the link between the sensor and
corrector. From the sensing data the controller determines the required signals to be sent
Focus spot
Hyperopia
Badal lens
L
2
L
1
Pupil plane
Figure 21. Implementation of a Badal optometer in an AO system. Focus is adjusted using two
plane mirrors which results in the pupil plane remaining fixed. (The colour version of this figure is
included in the online version of the journal.)
3446 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

to the actuators at that point in time. As ShackHartmann sensors and deformable
mirrors are the most widely used sensing and correcting elements, the focus of this section
is on the control of these devices.
3.3.1. The control matrix
The first step in the control process is to determine the appropriate signals to send to the
corrective device to produce a given wavefront. This is achieved via multiplication of the
sensing signals by the so-called control matrix. This matrix is derived from the influence
functions matrix IF. To obtain IF, a known voltage is applied to each actuator in turn and
the effect on the wavefront sensor is measured. Each column represents the change in the
wavefront per unit voltage applied to a given actuator. The surface of the mirror (as seen
by the sensor) can be considered to be represented by a linear superposition of these
functions. Hence, in matrix form
S IF V, 19
where V is a vector containing the actuator voltages. The nature of the elements of the
vector S describing the mirror surface depends upon how the influence functions have been
represented. Each influence function can either be represented by the raw Shack
Hartmann slopes, or the wavefront can be reconstructed and they can be represented by
a given number of Zernike polynomial coefficients. Hence, V is a vector of slopes or
Zernike coefficients. The choice depends upon the specific application. When aiming to
fully correct the wavefront, the raw slopes are preferred as this avoids any inaccuracies
owing to fitting errors. When performing psychophysical experiments where it may be
desirable to manipulate individual Zernike coefficients the second approach is more
desirable. From Equation (19) the voltages required to produce a given wavefront are
given by
V CM S, 20
where CM is the control matrix and is the pseudoinverse of the influence functions matrix.
The inverse is obtained using a technique known as singular value decomposition (SVD)
[60]. SVD is a very powerful technique as it allows one to determine the set of all possible
surfaces or so-called system modes that the mirror can produce. The number of modes is
1st Pass
2nd Pass
DM
PM
PM
DM
Figure 22. Amplification of the stroke of a deformable mirror by allowing the light to pass onto the
deformable mirror (DM) twice. (The colour version of this figure is included in the online version of
the journal.)
Journal of Modern Optics 3447
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

equal to the number of actuators. Depending on the particular corrective device and
system design, it may be advantageous to reject some of these modes in the formation of
the control matrix. For example some modes may require large actuator voltages to
produce them, which will cause the voltages to reach their limit. This is termed actuator
clipping.
3.3.2. Temporal control
An AO system can be controlled in either an open-loop or closed-loop manner. In
an open-loop system the wavefront sensor is placed before the corrective device and the
data from the sensor is used to directly determine the actuator control signals. This is
referred to as feedforward control. In closed-loop systems the corrective element is
placed before the sensor as illustrated in Figure 9 showing a general AO system.
The input to the control loop is the error signal, i.e. the residual wavefront after the
mirror surface has been updated. The advantages of open-loop control are simplicity
and stability. The disadvantages are that as there is no feedback the calibration must be
very accurate. All AO systems for the eye are operated in closed-loop. The so-called
control law used is the integral controller, where the control voltages V at a time t
i1
are given by
Vt
i1
g CM S
meas
S
req
Vt
i
, 21
where g is the gain, CM is the control matrix, S
meas
is the measured wavefront and S
req
is
the required wavefront. When aiming to fully correct the wavefront for example, each
element of S
req
is zero and so the control law is simply
Vt
i1
g CM S
meas
Vt
i
. 22
There are inevitably small time delays in the system which means that by the time the
signals reach the corrective device and the correction is applied, the wavefront aberrations
will have changed. The higher the gain g the faster the aberrations can be corrected. But
this can result in large oscillations and the system can become unstable. The temporal
correction capabilities is described by the bandwidth. In practice this is determined from
the ratio of the power spectrum of the eyes aberrations with the AO system running in
closed-loop to the power of the aberrations with the corrective element static.
The resulting plot is known as the disturbance power rejection curve (DPRC).
Although there are several definitions of bandwidth, one used often in vision science is
the frequency at which the two power spectra are equal, i.e. where the DPRC crosses 1.
A schematic of typical DPRCs is shown in Figure 23. The magnitude of frequency
components less than the bandwidth are reduced, whereas those beyond are enhanced.
A theoretical analysis of the temporal behaviour of a typical vision science AO system can
be found at [17].
Owing to time delays the wavefront needs to be sampled at a rate of 1020 times the
required bandwidth. Typical systems sensing at around 25 Hz normally use a gain of
around 0.3 and have a bandwidth of around 1 Hz. Future work on modelling of the
aberration dynamics will help to optimise the loop as more sophisticated control laws can
potentially be developed to increase correction speed.
3448 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

4. Adaptive optics and retinal imaging
Perhaps the most widely used application of AO in the human eye is retinal imaging. Much
of the work has focussed on imaging the cone photoreceptors. The ability to image these
individual cells in vivo provides the unique opportunity to non-invasively monitor the
functioning of the retina and to detect disease at an earlier stage. This also means that the
progression of disease can be monitored with great accuracy, allowing for the opportunity
to determine more effectively the efficacy of therapies. The first direct images of the
photoreceptor mosaic in vivo were obtained by Miller and colleagues using a high-
magnification fundus camera [11]. However, this was only achieved in subjects with good
optical quality. With the advent of AO, it is now possible to routinely image cone
photoreceptors, and other cells in vivo.
4.1. Imaging modalities
Any system imaging the eye in vivo has its transverse resolution limited by the eyes
aberrations. So far AO has been combined with three types of retinal imaging instruments
that are widely used in clinical practice. Aside from the work of Dreher and colleagues
in the late 1980s, the first systems were AO assisted conventional ophthalmoscopes, with
the first system developed in 1997 by Liang et al. [15]. Although providing images of
individual photoreceptors, this type of imaging modality is not suitable for imaging at
various depths in the retina. As a result AO was later combined with the confocal scanning
laser ophthalmoscope (cSLO) in 2002 by Roorda and colleagues [61]. A year later AO was
combined with ocular coherence tomography (OCT) [62]. This interferometric technique
has superior depth resolution over the cSLO and has allowed for diffraction-limited
volumetric imaging of the retina [58]. The specifics of the performance of the AO system
are generally similar across instruments. The closed-loop bandwidths are normally around
1
Frequency (Hz)
Closed-loop B.W.
P
o
w
e
r

w
i
t
h

A
O

o
n
P
o
w
e
r

w
i
t
h

A
O

o
f
f
High gain
Low gain
Figure 23. Typical disturbance power rejection curves showing the effect of gain of the controller on
the closed-loop bandwidth. Increasing the gain increase the bandwidth, but has an increasingly
adverse effect on the fluctuations at higher frequencies. (The colour version of this figure is included
in the online version of the journal.)
Journal of Modern Optics 3449
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

1 Hz and it is not uncommon to achieve residual rms values as low as 0.1 mm across
a dilated pupil. What primarily separates the techniques is their axial resolution.
Although there are ongoing efforts to make imaging systems equipped with AO to
become commercially available, they are still confined to the laboratory at present.
The following section illustrates the principle of each of these AO assisted imaging
modalities and the current developments to date.
4.1.1. Flood illumination ophthalmoscope
Combining AO with a flood-illuminated ophthalmoscope is the most straight forward
imaging modality. A patch of retina is illuminated by a low coherence light source and
a CCD camera in a conjugate retinal plane captures the image. The patch is around 1

to
avoid the effects of anisoplanatism. The image is captured in around 4 ms to reduce image
blur owing to small eye movements. A schematic of a typical set up is shown in Figure 24.
It is desirable for the imaging light source to have low spatial coherence to reduce
speckle, and a narrow spectral bandwidth to reduce the impact of chromatic aberration. If
the bandwidth is too narrow, however, temporal coherence will lead to an increase in
speckle. Several types of imaging light source have been employed. The first system,
constructed at Rochester, uses a Krypton flash-lamp [15]. As this is a broad bandwidth
source filters are used with a typical bandwidth of 525 nm to select a particular
wavelength range. Once AO has reduced the aberrations of the eye to a suitable level
governed by a suitably low residual rms wavefront error or given number of mirror
Laser
(wavefront sensing)
Retinal image
Flash lamp
(retinal imaging)
Cold mirror
Deformable
mirror
CCD
CCD
Figure 24. Typical setup of an AO assisted flood illumination ophthalmoscope. The eye is
illuminated by a flashlamp. Blur in the image owing to the eyes wavefront aberrations is corrected
for using a deformable mirror before being captured by a CCD camera. (The colour version of this
figure is included in the online version of the journal.)
3450 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

iterations, the flash-lamp is triggered and an image captured. As the lamp takes 510 s to
charge continuous imaging is not possible. Another option for the light source is an SLD
passed through a multimode fibre to reduce its coherence, or a multimode laser diode.
Using these types of sources Rha and colleagues have demonstrated the ability to
continuously capture images of the retina with rates of up to 60 Hz [63]. This real-time
imaging ability allowed them to monitor rapid fluctuations in the reflectance of single
cones. Several other groups also use AO assisted flood illumination ophthalmoscopes, for
example [64,65].
A consideration in retinal imaging is what wavelength to use. From Equation (1) it can
be seen that the width of the PSF decreases with decreasing wavelength. Hence, in principle
resolution is improved. However, there are several issues to consider such as safety and the
availability of a suitable light source. Typical wavelengths used in AO assisted flood
illumination ophthalmoscopy are in the visible region. It has been found that the contrast of
retinal images shows only slight variation in the wavelength range 550750 nm [66]. As the
wavefront source is normally in the infrared region, the chromatic difference in focus is
normally accounted for by translating the CCD camera along the optical axis.
4.1.2. Scanning laser ophthalmoscope
AO can also be combined with a confocal scanning laser ophthalmoscope (cSLO) as
illustrated in Figure 25.
The main advantage of this technique over conventional flood-illumination is its axial
sectioning capabilities. In a cSLO the laser is scanned across the retina and the image is built
up point-by-point. This allows the use of more sensitive point detectors such as
photomultipliers or avalanche photodiodes. Before reaching the detector the light passes
through a small aperture conjugate to the retinal plane being imaged. Owing to this so-called
Laser
(wave sensing
& retinal imaging)
CCD
X-Y
scanning
optics
PMT
Confocal pinhole
& detector
Figure 25. Principle of the AO assisted confocal laser scanning ophthalmoscope. The wavefront
sensing source is scanned across the retina and is used to build an image point by point. (The colour
version of this figure is included in the online version of the journal.)
Journal of Modern Optics 3451
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

confocal pin-hole, light scatter from other layers of the retina for example are prevented
from reaching the image as illustrated in Figure 26. Hence, image contrast is increased.
Unlike in the flood illumination method, the same light source is used for wavefront
sensing and imaging. Although the point source is scanned on the retina, the outgoing light
passes back via the same path and so is descanned as in the speckle reduction method
illustrated in Figure 12. It is not possible to use the same imaging light source for sensing in
the flood illumination method as the patch on the retina is too large. The ability to use the
same light source has two advantages. Firstly there is no need to account for aberration
differences between the imaging wavelength and wavefront sensing wavelength. Secondly
the aberrations are measured over the entire imaged area. The light first passes by the
deformable mirror before entering the eye. Hence, the aberrations are corrected on the way
into the eye and on the way out, which increases resolution. This also means that changing
the depth of the beam in the retina for optical sectioning can be done by the corrective
device.
The general AO assisted cSLO system illustrated in Figure 25 shows the pupil plane
being relayed to the sensor and corrector using lenses. Often, however, curved mirrors are
the preferred over lenses. The advantage of mirrors is that they dont have back reflections.
This isnt such a major issue in systems where the wavefront sensing source enters the
system close to the eye as back-reflections from the optics can be readily removed using the
same method used to block corneal reflections. However, this is a major issue for the cSLO
where the beam is passing through many more optics and also being scanned. Another
advantage of mirrors is that they are insensitive to wavelength. As the mirrors need to be
used off-axis often optical design programs such as Zeemax are used in order to determine
their optimal position to minimise system aberrations.
Since the first demonstration of an AO assisted cSLO in 2002 a variety of
modifications have been made to such instruments. These include using two deformable
mirrors to increase the aberration correction ability of the system [57,67], fitting it with
multiple imaging wavelengths [68], and compensation of eye movements [69]. The typical
image acquisition speed is 2030 Hz. As axial resolution depends upon the PSF of the
ingoing and outgoing beams, AO not only increases the transverse resolution but also
the axial resolution. Romero-Borja and colleagues have reported an axial resolution as low
as 71 mm [70].
4.1.3. Optical coherence tomography
OCT is an imaging technique based on interferometry. It consists of a Michelson
interferometer in conjunction with a low-coherence light source. The technique provides
Confocal pinhole
Different retinal layers
Figure 26. Principle of the confocal pin-hole used for axial sectioning in the confocal scanning laser
ophthalmoscope. (The colour version of this figure is included in the online version of the journal.)
3452 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

a higher sensitivity and axial resolution than conventional flood illumination ophthalmo-
scopes and SLOs. The axial resolution is governed by the mean wavelength and bandwidth
of the imaging light source. The transverse resolution is determined by the performance of
the AO system. Hence, unlike AO assisted cSLOs, axial and transverse resolution are
controlled by independent mechanisms. Ultrahigh resolution (UHR) OCT uses a light
source with a broader bandwidth and is capable of an axial resolution of 3 mm. This makes
an AO assisted OCT machine a highly desirable tool for taking high resolution volumetric
images of the retina. The cost, however, is an increase in system complexity. Owing to the
large spectral bandwidth of light used, chromatic aberration is a particular problem for
these instruments. As a potential solution to this, Ferna ndez and colleagues have designed
a lens capable of correcting chromatic differences in the infrared region [71].
There are two broad categories of operation based on how the depth resolved image is
formed in an OCT system: time domain OCT and Fourier/Spectral domain OCT. These
two regimes are illustrated in Figure 27. In the case of time-domain OCT the reference
mirror is moved axially and a retinal image is acquired for each position of this mirror.
Hence, the depth images are collected over time. For Fourier domain OCT the reference
mirror remains fixed and a spectrometer is used in conjunction with a linear array.
The lateral (xy) portion of the image can be obtained by scanning the retina as in the case
of the SLO, or simply illuminating a patch on the retina and taking a full image as in the
case of flood illumination ophthalmoscopy. The two main ways of incorporating AO into
an OCT system is either in the channel containing the eye or the detector channel. Each of
these implementations has advantages and disadvantages. For an overview see [72].
AO was first combined with OCT in 2003 by Miller and colleagues in 2003 [62].
The OCT component operated in time-domain mode and had an axial resolution of 14 mm.
They were able to increase the transverse resolution using AO to 35 mm. A year later
Hermann and colleagues demonstrated the first UHR OCT-AO system and were able to
achieve an axial resolution of 3 mm [73]. Again this was a time domain instrument which
was more susceptible to image blur due to eye motion than with the faster Fourier domain
approach. In 2005 Zhang et al. demonstrated the first AO assisted Fourier domain
Low coherence source
Reference mirror
Point
detector
Low coherence source
Reference mirror
Detector array
Spectrometer
(Fixed position)
(a)
(b)
Figure 27. Two operating regimes of ocular coherence tomography. They differ in the method they
use to image at each depth. (a) Time domain OCT. An image is acquired for each position of the
reference mirror. (b) Frequency domain OCT. The reference mirror remains fixed and images at
various depths are acquired simultaneously using a spectrometer and linear array. (The colour
version of this figure is included in the online version of the journal.)
Journal of Modern Optics 3453
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

OCT system [74]. They were able to image the inner and outer segments of individual
photoreceptors. A detailed review of the field of OCT-AO systems can be found at [75].
The current state of the art in terms of AO assisted OCT imaging is the system
demonstrated by Zawadzki and colleagues [58]. This system operates in the Fourier
domain and contains an ultra broadband imaging source to increase axial resolution.
An achromatising lens is used to reduce the eyes chromatic aberration and two
deformable mirrors are used in cascade to increase transverse resolution. Diffraction-
limited resolution was achieved in the eyes tested. The instrument has a volumetric
resolution of 3.5 3.5 3.5 mm
3
. Merino and colleagues have demonstrated an AO
assisted SLO-OCT hybrid [76].
4.1.4. Multiconjugate AO
The goal of multiconjugate AO (MCAO) is to obtain high-resolution retinal images over
a wide field. This will make retinal imaging less time consuming as it will avoid the need to
image at multiple locations and montage the images. The idea of MCAO is a relatively new
one and was first proposed by Beckers for astronomical applications [77]. In all
conventional AO systems the correction device or devices are placed in a plane conjugate
to the pupil, where it is assumed the wavefront becomes aberrated. In MCAO each
correction device is conjugate to a separate plane in the eye. The principle is illustrated in
Figure 28. Suppose there are two locations within the eye which impart aberrations on the
wavefront and the correction device is placed in a plane conjugate to the pupil. When
measuring the aberrations on-axis, the corrective device can readily correct these
aberrations by assuming the correct shape (Figure 28(a)). If through the same eye the
aberrations resulting from an off-axis point are measured, the corrective device needs to
attain a different shape from that for the on-axis case (Figure 28(b)). Otherwise a perfect
correction will not be obtained. Hence, the solution to obtain perfect correction over
a wide field of view is to have a correction device conjugated to each aberrating plane
(Figure 28(c)). MCAO inevitably requires the eye to be illuminated by several wavefront
sensing beams.
Based on a computer model, Bedggood and colleagues have predicted that MCAO
using five deformable mirrors should be able to increase the area of the isoplanatic patch
by a factor of 6 [78]. Using two deformable mirrors, one conjugate to the pupil plane, and
one conjugate to 3 mm in front of the eye, Thuang and colleagues have measured a seven-
fold increase in the width of the isoplanatic patch of an artificial eye [79]. MCAO on a real
human eye is yet to be demonstrated.
Other less expensive ways of increasing the isoplanatic patch have been investigated
such as using one deformable mirror to correct the average wavefront based on the
measurement from two retinal locations. However, Dubinin and colleagues for example
predicted only a factor of 1.251.7 increase in the size of the isoplanatic patch for their two
subjects [80].
4.2. Image processing
Once retinal images have been obtained there is often some post-processing involved. This
includes deconvolution procedures to remove blur from residual aberrations [81,82],
3454 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

photomontaging images from different retinal locations to obtain a wide field-of-view
image [83,84], and algorithms to automatically locate cones [85]. It is also common
practice to average images to increase contrast.
4.3. Imaging applications
As mentioned previously, much of the work involving retinal imaging with AO has
focussed on various aspects of imaging the cone mosaic. The cones can readily be imaged
as they act like waveguides. Hence, they reflect incident light directly back towards the
pupil and appear as bright spots in the images. Owing to their waveguide nature, their
brightness depends upon the angle of the incident light relative to their axis. They typically
point towards the centre of the pupil to minimise the impact on vision of peripheral rays.
This is known as the StilesCrawford effect. Using AO, Roorda and colleagues were able
Aberrating layer
Pupil plane
(a)
On-axis point
(b)
Off-axis point
(c)
DM conjugate
to each aberrating layer
Single DM conjugate
to pupil plane
Single DM conjugate
to pupil plane
Wide field correction
Correction of axial
aberrations
Correction of off-axis
aberrations
Wide field
Figure 28. Principle of multiconjugate AO. Using a single deformable mirror conjugate to the pupil
plane will require a different correction depending on the retinal location (a) and (b). For a wide-field
correction a mirror conjugate to each aberrating plane is needed. (The colour version of this figure is
included in the online version of the journal.)
Journal of Modern Optics 3455
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

to measure for the first time the angular tuning of individual cones [86]. This was achieved
by measuring the change in their reflectance with angle of the incident illumination beam.
They found minimal disarray amongst cones.
Pallikaris and colleagues have found that the reflectance of individual cones varies over
the course of minutes [87]. They propose it may be as a result of disc shedding. Changes in
reflectance has since been used as a tool for functional imaging [88,89]. This has important
consequences in detecting disease. Locations where there are cones that have been
damaged can easily be recognised as dark patches in the image.
4.3.1. Arrangement of the cone classes
Colour vision is mediated by three cone classes: L, M and S, responsible for the detection
of long, medium and short wavelengths across the visible spectrum, respectively.
Combining AO and retinal densitometry led to the first in vivo images of the arrangement
of the three cone classes [90]. In this technique a cone class is bleached prior to taking
an image. By comparing the reflectance of cones in different images, the locations of each
class can be identified. Results from two subjects used in the study are shown in Figure 29.
Being able to class cones in vivo allows investigators to correlate perceptual colour
vision with retinal arrangement and the density of each cone class. It has been shown that
the assignment of L and M cones is not entirely random, and that although the ratio of
L:M cones varies considerably across subjects they still have normal colour vision [91,92].
4.3.2. The cause of colour blindness
Individuals who are redgreen colour-blind are so as a result of the functional loss of one
cone class. It was commonly believed that such individuals had a normal fully functional
cone mosaic in which there were two instead of three photopigments. As a result of
imaging with AO, it has been shown that this is not always the case. Carroll and colleagues
found that one of their two colour-blind subjects had patches in the retina devoid of
functional cones [93]. Hence, AO was able to elucidate a new cause of colour blindness.
(Figure 30(a)) shows the image of the cone mosaic of the individual who although missing
the gene for the L photopigment, has a complete cone mosaic. The other subject had
a mutation of the M pigment gene, resulting in areas with no functioning cone
(a) (b)
5 arc mins
Figure 29. Image of the arrangement of the cone classes of two subjects. Adapted and reprinted by
permission from Macmillan Publishers Ltd [90]. (The colour version of this figure is included in the
online version of the journal.)
3456 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

class (Figure 30(b)). This study also highlights the use of linking AO retinal imaging with
genetics. AO has also been used to image subjects with S-cone dystrophy [94].
4.3.3. Disease
The ability to resolve individual cones makes AO a more reliable disease detection method.
If the disease is in the early stages, psychophysical procedures measuring visual
performance may not be sensitive enough to detect disease onset. AO has routinely
imaged a range of retinal diseases that adversely affect cone photoreceptors such as cone/
rod dystrophy and retinitis pigmentosa. Examples of such studies include [9597].
Figure 31 shows results obtained by Wolfing and colleagues where they imaged a patient
with rod-cone dystrophy [96]. Figure 31(a) shows a retinal image of the patient taken at 4

nasal to fixation. Compared to an aged-matched normal subject at the same location, there
are dark patches and highly-reflection scar tissue.
AO has also been used to image the flow of single blood cells through the vasculature
of the retina and determined their velocity [98,99]. This will have specific application in
early detection of diseases that cause changes in the retinal vasculature such as diabetic
retinopathy. Rod photoreceptors have also been imaged [100]. Again this will prove
valuable in the early detection of disease.
(a) (b)
25 m
Figure 31. Comparison of images of the cone mosaic from a subject with rod-cone dystrophy (a),
and a normal subject (b). Measurements were obtained 4

nasal to fixation. This figure was


published in [96]. Copyright Elsevier.
(a) (b)
20 m
Figure 30. Retinal images of two redgreen colour-blind-subjects. (a) The subject has a normal-
appearing full cone mosaic. (b) The subject has areas devoid of functioning cones.
Journal of Modern Optics 3457
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

4.3.4. Locus of fixation of the foveal cone mosaic
Another advantage afforded by high resolution imaging of the retina is determining its
location. Putnam and colleagues were able to record the retinal position of a stimulus to
within an accuracy of at least one fifth of a cone diameter [101]. Averaging across three
subjects they found that the locus of fixation was displaced from the location of highest
cone density by around 50 mm. Being able to accurately determine the position of stimuli
has value when performing psychophysical experiments for example, where one may wish
to illuminate individual cones. Being able to accurately determine the location of the retina
is also the first step involved in compensation of the effect of eye movements on image blur
in high-resolution retinal imaging.
5. Adaptive optics and psychophysics
Visual psychophysics is the study of the relationship between the physical properties of the
external visual world and its internal, perceptual representation. The first stage in visual
perception is the conversion of a blurred retinal image into electrical signals by the retinas
photoreceptors. Combining AO with imaging and psychophysical techniques allows the
measurement of how visual perception, retinal image blur and underlying retinal
architecture interact within the human nervous system.
A typical AO system for psychophysics is very similar to a conventional AO assisted
flood illumination ophthalmoscope. The difference is that instead of a camera capturing
the retinal image from the eye via the corrective device a stimulus is projected onto the
retina via the deformable mirror. As infrared is used for the wavefront sensing but the
stimulus is in the visible region of the spectrum, there will be a focussing error owing to
chromatic aberration. This can be offset by adjusting the position of the stimulus until it
subjectively appears to be in the same focal plane as the focussed spot from the infrared
source [28]. There are several devices commonly used to generate the stimulus. For
conventional psychophysical experiments not involving AO, the stimulus is normally
displayed on a computer monitor. These are difficult to incorporate into AO systems,
however, owing to their size. Furthermore, once the image reaches the retina the
luminance is often too low due to difficulties in collecting light from the monitor and light
loss through the optics of the system. To overcome these difficulties the common stimulus
generators of choice are either projectors or high luminance liquid crystal microdisplays.
When purchasing one of these devices there are several factors that need to be taken into
consideration such as spatial resolution and range of chromaticities. An overview of this
can be found in [102]. Poonja and colleagues have demonstrated it is possible to present
a stimulus onto the retina using an AO assisted SLO [103].
5.1. Applications
AO has been used in a variety of experiments involving visual perception. The main goal of
AO in imaging applications is to reduce the wavefront to zero. In terms of psychophysical
applications this helps investigators to determine the benefit to vision of a perfect optical
correction. AO is not only capable of correcting the wavefront but can also be used to
induce specific wavefronts allowing one to determine, for example, the effect of individual
aberrations.
3458 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

5.1.1. The limits of supernormal vision
The first investigations carried out were to determine how much benefit to vision could be
obtained by correcting the monochromatic higher-order aberrations of eyes with normal
optical quality. Liang and colleagues were the first to demonstrate the visual benefit of
correcting the higher-order aberrations of the eye using AO when they were able to
considerably increase the contrast sensitivity of the two subjects tested [15]. For example,
contrast sensitivity was improved by a factor of 6 when viewing a grating with a spatial
frequency of 27.5 c deg
1
. This, however, was performed in monochromatic light and so
this study was later extended to viewing under white light conditions [104]. Although the
benefit of correcting the higher-order monochromatic aberrations was reduced in white
light, contrast sensitivity and acuity still showed improvement. These results encouraged
the development of customised correction of higher-order aberrations using contact lenses
and refractive surgery. Note that correction of these aberrations with spectacle lenses is
severely limited owing to rotation of the eye relative to the lens.
If all the aberrations of the eye are corrected for, the ultimate limit on
visual performance is set by the sampling of the photoreceptor mosaic. The spacing of
cones at the fovea, which is the area of the retina providing highest spatial resolution, is
around 2 mm. In order to correctly resolve a grating for example, the separation of the
bars must be at least 4 mm (corresponding to around 60 c deg
1
). Considering Figure 4,
for a typical 3 mm pupil, which is typical for everyday viewing conditions under
bright light, the resolution of the optics of the eye and that of the retina are
matched. Correcting the aberrations over a larger pupil can potentially give rise to
aliasing in which the imaging is incorrectly perceived. A simple example of this is shown
in Figure 32.
As light level decreases the pupil size will inevitably increase. However, spatial vision is
no longer initiated by cone photoreceptors and so aliasing has the potential to occur at
lower spatial frequencies. A systematic study on the benefit of AO for a range of
luminances and a range of typical pupil sizes has demonstrated that the benefit is limited
[105]. This confirms a commonly held belief that the optics and neural system are well
matched. The largest benefit will be seen in those eyes with worse than normal optical
quality such as those suffering from keratoconus in which the cornea thins and becomes
more conical in shape.
+
+
+
+
+
+
+
+
-
-
-
-
-
-
-
-
-
-
-
-
- -
-
-
-
+
+
+
+
+
+
+
+
-
-
-
-
-
- -
(a) (b)
Figure 32. A simple example of aliasing. (a) The pattern is repeating itself every unit in the vertical
direction and there are no photoreceptors on each strip in this direction. (b) Aliased perception.
The photoreceptors are shown to be rectangularly packed as opposed to hexagonally packed for
simplicity.
Journal of Modern Optics 3459
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

Lundstro m and colleagues have used AO to directly assess the impact of both low and
higher-order aberrations on peripheral resolution acuity [106]. They found negligible
improvement. This was most likely owing to the well-known decrease in sampling as one
moves further from the fovea.
5.1.2. Visual simulator
AO systems used to generate specific wavefronts are commonly termed visual simulators
[107]. Using a commercial simulator (crx1, Imagine Eyes), Rocha and colleagues found
that symmetrical aberrations such as defocus and spherical aberration, are the most
detrimental to visual performance [108]. As mentioned previously, when averaging the
level of higher-order aberrations across the population, spherical aberration is non-
negligible. Hence, it is a target aberration for customised correction using contact lenses
for example. Using an AO visual simulator to induce or correct spherical aberration, it has
been confirmed that correcting this aberration gives the best visual performance [109,110].
Another application is the simulation of outcomes of refractive surgery [111]. This will
give the patient an opportunity to see how their vision will be affected once the surgery has
taken place. This is limited, however, by how well the outcome can be predicted. AO visual
simulators can also be used as a phoropters chart [112]. The idea is that the patient views
a visual acuity chart whilst their aberrations are measured and then corrected by AO. Then
an optimal spectacle prescription is calculated based on an image quality metric
incorporating the effect of all of the aberrations present in the eye, not just defocus and
astigmatism. AO can also be utilised to test such metrics. Manzanera and colleagues have
demonstrated a visual simulator to aid in the design of lenses for presbyopic subjects, i.e
those with reduced accommodation owing to age [113].
Artal and colleagues used an AO system to present a rotated version of their wavefront
in five subjects. They found for every subject that it was sharper with their natural
unrotated aberrations. Hence, demonstrating that the neural system adapts to its own
aberrations [114]. Chen and colleagues provided further adaptation of the eyes adaptation
to its wavefront where they found that the best subjective correction was when some
higher-order aberrations remained present [115]. This has important application in the
field of customised vision correction as it shows that full correction of the eyes wavefront
may not provide the best visual outcome.
5.1.3. Microstimulation
AO allows the ability to image a point source onto the retina with a diameter less than that
of the size of a fovea cone. Hofer and colleagues have demonstrated that cones from
the same class can elicit a different response [116]. Makous and colleagues have
taken advantage of the small point sources to perform microperimetry to detect
microscotomas [117].
6. Other applications of adaptive optics in vision science
Another aspect of the visual system under investigation using AO is the effect of higher-
order aberrations on the control of accommodation. Hence, AO provides the opportunity
3460 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

to investigate the eyes AO system. In a typical AO system using a ShackHartmann
sensor, the controller uses the slopes to determine how much to move the actuators.
The accommodation system uses a variety of so-called cues to determine the correct power
of the lens. These include chromatic aberration, size changes, binocular vision cues. For
a review see [118]. However, some subjects can still accommodate adequately when
viewing a target monocularly in monochromatic light where these cues are absent, leading
many investigators to recognise the importance of monochromatic aberrations.
The typical set-up for a system investigating accommodation is the same as that for
psychophysical experiments. Normally AO systems carrying out accommodation
experiments use a Badal optometer. This serves two purposes: one is to set the
accommodation level at which one wants to carry out the experiment and the other is to
provide the step for dynamic accommodation experiments. This step can be produced with
the deformable mirror, however, this is limited because of stroke. Normally accommoda-
tion is measured by the Zernike defocus term. The difficulty faced by AO systems is that
the measurements from the eye pass by the deformable mirror before reaching the sensor.
Hampson and colleagues simultaneously measure the eyes aberrations independent of the
deformable mirror as shown in Figure 33.
Even-order monochromatic aberrations can cause the appearance of the object to be
different depending on whether the object is focussed in front or behind the retina. Such
aberrations can potentially provide the accommodation system with a directional cue. An
illustration of how the PSF changes with defocus is shown in Figure 34 [119]. Wilson and
colleagues have demonstrated that differences in the PSF can be seen subjectively [120].
Studies measuring the dynamic accommodation response whilst aberrations are manipu-
lated have shown that some subjects use higher-order aberrations [121,122]. It has also been
found that the higher-order aberrations can affect the control of steady-state accommoda-
tion [44]. Currently investigators do not know with certainty why some subjects use higher-
order monochromatic aberrations and others do not. These results have important
applications for customised optical corrections as correcting all of the eyes aberrations
could potentially adversely affect the accommodation system for some individuals.
AO is also being used in the study of myopia. Investigations have included determining
the difference between myopic and emmetropic eyes in terms of both their photoreceptor
packing density [123,124] and the benefit of correcting higher-order aberrations [125].
C
C
D
ShackHartmann
C
C
D
ShackHartmann
Eye
Deformable
mirror
Eye
Deformable
mirror
Figure 33. Illustration of a method to measure the eyes aberrations independently in a closed loop
adaptive optics system. (The colour version of this figure is included in the online version of the
journal.)
Journal of Modern Optics 3461
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

Other AO-type applications being explored include the use of deformable mirrors to
generate depth cues to simulate accommodation in virtual displays [126], and the use of
remotely controlled intra-ocular lenses [127,128].
7. The future of AO in the eye
The introduction of the technique of AO to the eye has proven to be an invaluable tool in
investigating various aspects of the visual system. The field has grown at a rapid rate since
its first demonstration in the late 1980s, with system performance ever increasing and our
knowledge of the visual system growing. There is still, however, much more to be gained
in this field.
7.1. Imaging
Currently, diffraction sets the upper resolution limit in AO systems. AO imaging systems
are approaching the diffraction limit, with some systems reaching it. Combining AO with
other imaging techniques such as structured illumination will potentially allow the
diffraction limit to be surpassed. Structured illumination has been used in microscopy to
double the diffraction limit [129]. This involves illuminating the object with a high spatial
frequency grating pattern of known characteristics and processing the aliased image.
Defocus Astigmatism Sperical ab. Coma
Real eye
Trefoil
1
0.5
0
+0.5
+1
Figure 34. Variations in the PSF for various aberrations at different levels of defocus. Simulations
were carried out using a 5mm pupil with 1mm of each aberration added. The last column shows that
of a real eye. Reprinted from [118] with permission from Elsevier.
3462 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

Another step in imaging is the development of MCAO systems, allowing high resolution
wide-field retinal images to be obtained. Theoretical modelling and demonstrations on
model eyes have produced encouraging results and efforts are being made to apply the
technique to a real human eye. This will make imaging systems more efficient as there is less
need for the photomontaging of multiple images. This will encourage clinical development.
The most widely imaged cells are the photoreceptors. As mentioned previously, these
prove relatively easy to image owing to their waveguiding properties. A variety of diseases
can be detected using these cells. Being able to image other retinal cells will expand this
capability. Glaucoma, for example affects ganglion cells. These cells are vital in conveying
the image to the brain. Unfortunately, they cannot be imaged using AO alone as they
do not reflect enough light back. Combining AO with fluorescing markers has the
potential to overcome this [130].
7.1.1. Psychophysics
Although a binocular ShackHartmann sensor has been demonstrated [131], to date there
are no binocular AO systems. This is the next logical step in instrumentation development
for assessing visual performance. Owing to binocular summation, if one wants to truly
stimulate outcome of refractive surgery for example, a binocular system is needed. Studies
have shown that impact of monochromatic aberrations on binocular is dependent upon
both the type of Zernike aberration [132] and the interocular differences [133].
Furthermore, it is well known that accommodation depends upon input from the two
eyes. A binocular AO system for accommodation studies is under construction at the
University of Bradford.
7.1.2. Other areas
With the advent of polychromatic ShackHartmann sensors, it is well established that the
monochromatic aberrations vary with wavelength. There are currently no polychromatic
AO systems capable of manipulating the aberration fluctuations across a range of
wavelengths. Interestingly, McLellan and colleagues have suggested that monochromatic
aberrations reduce the impact of blur owing to chromatic aberration [134]. Investigators
currently cannot explain why some subjects cannot accommodate in monochromatic light,
and why some of those who can, cannot accommodate when monochromatic aberrations
are corrected. Having a dynamic polychromatic AO system would help to answer these
questions. This would also be of benefit to imaging systems where low coherence light
sources are used such as in OCT. Systems are likely to be expensive as they will require
multiple correction devices.
AO may be beneficial to surgical procedures targeting specific retinal areas. This high
resolution imaging along with the development of stabilisation techniques will allow for
a more precise controlled laser beam delivery. One area of application could be treating
microaneurysms in diabetic retinopathy.
Currently the goal of MCAO is focussed on obtaining retinal images over a large
field of view. Being able to manipulate the blur of the stimulus over a wide field will
have benefits in assessing the impact of wide field correction on visual acuity and
accommodation.
Journal of Modern Optics 3463
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

References
[1] von Helmholtz, H. Popular Scientific Lectures; Dover: New York, 1962.
[2] Gros, D.A.; West, R.W. Introduction to the Optics of the Eye; Butterworth Heinemann: Oxford,
2001.
[3] Cox, I. Customised Vision Correction Devices. In Adaptive Optics for Vision Science; Porter, J.,
Queener, H., Lin, J., Thorn, K., Awwal, A.A.S., Eds.; Wiley and Sons, Inc: New York, 2006;
pp 291310.
[4] MacRae, S.M. Customized Corneal Ablation. In Adaptive Optics for Vision Science; Porter, J.,
Queener, H., Lin, J., Thorn, K., Awwal, A.A.S., Eds.; Wiley and Sons, Inc: New York, 2006;
pp 311330.
[5] Zhu, M.; Collins, M.J.; Iskander, R.D. Ophthal. Physiol. Opt. 2004, 24, 562571.
[6] Hampson, K.M.; Munro, I.; Paterson, C.; Dainty, C. J. Opt. Soc. Am. A 2005, 22, 12411250.
[7] Babcock, H.W. Pub. Astr. Soc. Pac. 1953, 65, 229236.
[8] Hardy, J.W.; Lefebvre, J.E.; Koliopoulos, C.L. J. Opt. Soc. Am. A 1977, 67, 360369.
[9] Dreher, A.W.; Bille, J.F.; Weinreb, R.N. Appl. Opt. 1989, 28, 804808.
[10] Artal, P.; Navarro, R. Opt. Lett. 1989, 14, 10981100.
[11] Miller, D.T.; Williams, D.R.; Morris, G.M.; Liang, J. Vis. Res. 1996, 36, 10671079.
[12] Liang, J.; Grimm, B.; Goelz, S.; Bille, J.F. J. Opt. Soc. Am. A 1994, 11, 19491957.
[13] Smirnov, M.S. Biophys. J. 1962, 7, 766795.
[14] Charman, W.N. Contact Lens Ant. Eye 2005, 28, 7592.
[15] Liang, J.; Williams, D.R.; Miller, D.T. J. Opt. Soc. Am. A 1997, 14, 28842892.
[16] Ferna ndez, E.J.; Iglesias, I.; Artal, P. Opt. Lett. 2001, 26, 746748.
[17] Hofer, H.; Chen, L.; Yoon, G.Y.; Singer, B.; Yamauchi, Y.; Williams, D.R. Opt. Exp. 2001, 8,
631643.
[18] Curcio, C.A.; Sloan, K.R.; Kalina, R.E.; Hendrickson, A.E. J. Comp. Neurol. 1990, 292,
497523.
[19] Liang, J.; Williams, D.R. J. Opt. Soc. Am. A 1997, 14, 28732883.
[20] Porter, J.; Guirao, A.; Cox, I.G.; Williams, D.R. J. Opt. Soc. Am. A 2001, 18, 17931803.
[21] Thibos, L.N.; Applegate, R.A.; Schwiegerling, J.T.; Webb, R. J. Refract. Surg. 2002, 18, 652660.
[22] Born, M.; Wolf, E. Principles of Optics; Cambridge University Press: Cambridge, UK, 1989.
[23] Applegate, R.A.; Donnelly, W.J.; Marsack, J.D.; Koenig, D.E. J. Opt. Soc. Am. A 2007, 24,
578587.
[24] Cheng, H.; Barnett, J.K.; Vilupuru, A.S.; Marsack, J.D.; Kasthurirangan, S.; Applegate, R.A.;
Roorda, A. J. Vis. 2004, 4, 272280.
[25] Charman, W.N. Ophthal. Pysiol. Opt. 2005, 25, 285301.
[26] Castejo n-Mocho n, J.F.; Lo pez-Gil, N.; Benito, A.; Artal, P. Vis. Res. 2002, 42, 16111617.
[27] Plainis, S.; Pallikaris, I.G. J. Mod. Opt. 2008, 55, 759772.
[28] Hofer, H.; Artal, P.; Singer, B.; Arago n, J.L.; Williams, D.R. J. Opt. Soc. Am. A2001, 18, 497506.
[29] Diaz-Santana, L.; Torti, C.; Munro, I.; Gasson, P.; Dainty, P. Opt. Exp. 2003, 11, 25972605.
[30] Gruppetta, S.; Lacombe, F.; Puget, P. Opt. Exp. 2005, 13, 76317636.
[31] Spauschus, A.; Marsden, J.; Halliday, D.M.; Rosenberg, J.R.; Brown, P. Exp. Brain Res. 1999,
126, 556562.
[32] Stark, L.; Campbell, F.W.; Atwood, J. Nature. 1958, 182, 857858.
[33] Cagigal, M.P.; Canales, V.F.; Castejo n-Mocho n, J.F.; Prieto, P.M.; Lo pez-Gil, N.; Artal, P.
Opt. Lett. 2002, 27, 3739.
[34] Ferna ndez, E.J.; Artal, P. Appl. Opt. 2007, 46, 69716977.
[35] Jain, P.; Schwiegerling, J. J. Mod. Opt. 2008, 55, 737748.
[36] Manzanera, S.; Canovas, C.; Prieto, P.M.; Artal, P. Opt. Exp. 2008, 16, 77487755.
[37] Llorente, L.; Diaz-Santana, L.; Lara-Saucedo, D.; Marcos, S. Optom. Vis. Sci. 2003, 80, 2635.
3464 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

[38] Williams, D.R.; Artal, P.; Navarro, R.; McMahon, M.J.; Brainard, D.H. Vis. Res. 1996, 36,
11031114.
[39] Atchison, D.A.; Scott, D.H. J. Opt. Soc. Am. A 2002, 19, 21802184.
[40] Bedggood, P.; Daaboul, M.; Ashman, R.; Smith, G.; Mehta, A. J. Biomed. Opt. 2008, 13,
024008.
[41] Yoon, G. Wavefront Sensing and Diagnostic Uses. Adaptive Optics for Vision Science; Wiley
and Sons, Inc: New York, 2006; pp 6384.
[42] Seifert, L.; Liesener, J.; Tiziani, H. Opt. Comm. 2003, 216, 313319.
[43] Hampson, K.M.; Paterson, C.; Dainty, C.; Mallen, E.A.H. J. Opt. Soc. Am. A 2006, 23,
10821088.
[44] Hampson, K.M.; Chin, S.S.; Mallen, E.A.H. Does the Accommodative Mechanism of the
Human Eye Calibrate itself Using Aberration Dynamics? In Adaptive Optics for Industry and
Medicine, Proceedings of the Sixth International Workshop, Dainty, C., Ed.; Imperial College
Press: London, 2007; pp 287292.
[45] Ragazzoni, R. J. Mod. Opt. 1996, 43, 289293.
[46] Iglesias, I.; Ragazzoni, R.; Julien, Y.; Artal, P. Opt. Exp. 2002, 10, 419428.
[47] Chamot, S.R.; Dainty, C.; Esposito, S. Opt. Exp. 2006, 14, 518526.
[48] Ferna ndez, E.J.; Artal, P. Opt. Exp. 2003, 11, 10561069.
[49] Dalimier, E.; Dainty, C. Opt. Exp. 2005, 13, 42754285.
[50] Doble, N.; Miller, D.T.; Yoon, G.; Williams, D.R. Appl. Opt. 2007, 46, 45014514.
[51] Doble, N.; Miller, D.T. Wavefront Correctors for Vision Science. In Adaptive Optics for Vision
Science; Wiley and Sons, Inc: New York, 2006; pp 83117.
[52] Prieto, P.M.; Ferna ndez, E.J.; Manzanera, S.; Artal, P. Opt. Exp. 2004, 12, 40594071.
[53] Quan, W.; Wang, Z.Q.; Mu, G.G.; Ning, L. Int. J. Light and Elec. Opt. 2003, 114, 467471.
[54] Vargas-Mart n, F.; Prieto, P.M.; Artal, P. J. Opt. Soc. Am. A 1998, 15, 25522562.
[55] Marcos, S.; Diaz-Santana, L.; LLorente, L.; Dainty, C. J. Opt. Soc. Am. A 2002, 19,
10631072.
[56] Alvarez, A.W. Two-element Variable-power Spherical Lens. US Patent 3,305,294, 21 February
1967.
[57] Chen, D.C.; Jones, S.M.; Silva, D.A.; Olivier, S.S. J. Opt. Soc. Am. A 2007, 24, 13051312.
[58] Zawadzki, R.J.; Cense, B.; Zhang, Y.; Choi, S.S.; Miller, D.T.; Werner, J.S. Opt. Exp. 2008, 16,
81268143.
[59] Webb, R.H.; Albanese, M.J.; Zhou, Y.; Bifano, T.; Burns, S.A. Appl. Opt. 2004, 43, 53305333.
[60] Press, W.H.; Teukolsky, S.A.; Vetterling, W.T.; Flannery, B.P. Numerical Recipes in C;
Cambridge University Press: Cambridge, UK, 1992.
[61] Roorda, A.; Romero-Boorja, F.; Donnelley, W.J.; Queener, H. Opt. Exp. 2002, 10, 405412.
[62] Miller, D.; Qu, J.; Jonnal, R.S.; Thorn, K.E. Proc. SPIE 2003, 4956, 6572.
[63] Rha, J.; Jonnal, R.S.; Thorn, K.E.; Qu, J.; Zhang, Y.; Miller, D.T. Opt. Exp. 2006, 14,
45524569.
[64] Gargasson, J.F.L.; Glanc, M.; Le na, P. Opt. Acous. Imag. Biol. Media. 2001, 15, 11311138.
[65] Larichev, A.; Ivanov, P.; Iroshnikov, N.; Shmalgauzen, V.I.; Otten, L.J. Quantum Electron.
2002, 32, 902908.
[66] Choi, S.S.; Doble, N.; Lin, J.; Christou, J.; Williams, D.R. J. Opt. Soc. Am. A 2005, 22,
25982605.
[67] Hammer, D.X.; Ferguson, R.D.; Bigelow, C.E.; Iftimia, N.V.; Ustun, T.E.; Burns, S.A.
Opt. Exp. 2006, 14, 33543367.
[68] Grieve, K.; Tiruveedhula, P.; Zhang, Y.; Roorda, A. Opt. Exp. 2006, 14, 1223012242.
[69] Burns, S.A.; Tumbar, R.; Elsner, A.E.; Ferguson, D.; Hammer, D.X. J. Opt. Soc. Am. A 2007,
24, 13131326.
[70] Romero-Borja, F.; Venkateswaran, K.; Roorda, A.; Herbert, T. Appl. Opt. 2005, 44, 40324040.
Journal of Modern Optics 3465
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

[71] Ferna ndez, E.J.; Unterhuber, A.; Povaz ay, B.; Hermann, B.; Artal, P.; Drexler, W. Opt. Exp.
2006, 14, 62136225.
[72] Roorda, A.; Miller, D.T.; Christou, J. Strategies for High-resolution Retinal Imaging.
In Adaptive Optics for Vision Science; Porter, J., Queener, H., Lin, J., Thorn, K., Awwal,
A.A.S., Eds.; Wiley and Sons, Inc: New York, 2006; pp 264266.
[73] Hermann, B.; Ferna ndez, E.J.; Unterhuber, A.; Sattmann, H.; Fercher, A.F.; Drexler, W.;
Prieto, P.M.; Artal, P. Opt. Lett. 2004, 29, 21422144.
[74] Zhang, Y.; Rha, J.; Jonnal, R.S.; Miller, D.T. Opt. Exp. 2005, 13, 47924811.
[75] Pircher, M.; Zawadzki, R.J. Expert Re. Ophthalmol. 2007, 2, 10191035.
[76] Merino, D.; Dainty, C.; Bradu, A.; Podoleanu, A.G. Opt. Exp. 2006, 14, 33453353.
[77] Beckers, J.M. Increasing the Size of the Isoplanatic Patch Size with Multiconjugate Adaptive
Optics. In Proceedings of European Southern Observatory Conference and Workshop on Very
Large Telescopes and their Instrumentation, ESO Conference and Workshop; Vol. 30; Ulrich,
M.-H., Ed.; European Southern Observatory: Garching, Germany, 1998.
[78] Bedggood, P.A.; Ashman, R.; Smith, G.; Metha, A.B. Opt. Exp. 2006, 14, 80198030.
[79] Thaung, J.; Owner-Petersen, M.; Popovic, Z. Dual-conjugate Adaptive Optics Instrument for
Wide-field Retinal Imaging. In Adaptive Optics for Industry and Medicine, Proceedings of the
Sixth International Workshop, Dainty, C., Ed.; Imperial College Press: London, 2007.
[80] Dubinin, A.V.; Cherezova, T.Y.; Kudryashov, A.V. High-resolution Field-of-view Widening
in Human Eye Retina Imaging. In Adaptive Optics for Industry and Medicine, Proceedings of
the Sixth International Workshop, Dainty, C., Ed.; Imperial College Press: London, 2007.
[81] Arines, J.; Bara , S. Opt. Exp. 2003, 11, 761766.
[82] Christou, J.C.; Roorda, A.; Williams, D.R. J. Opt. Soc. Am. A 2004, 21, 13931401.
[83] Glanc, M.; Gendron, E.; Lacombe, F.; Lafaille, D.; Le Gargasson, J.F.; Le na, P. Opt. Comm.
2004, 230, 225238.
[84] Xue, B.; Choi, S.S.; Doble, N.; Werner, J.S. J. Opt. Soc. Am. A 2007, 24, 13641372.
[85] Li, K.Y.; Roorda, A. J. Opt. Soc. Am. A 2007, 24, 13581363.
[86] Roorda, A.; Williams, D.R. J. Vis. 2002, 2, 404412.
[87] Pallikaris, A.; Williams, D.R.; Hofer, H. IOVS 2003, 44, 45804592.
[88] Jonnal, R.S.; Rha, J.; Zhang, Y.; Cense, B.; Gao, W.; Miller, D.T. Opt. Exp. 2007, 15,
1614116160.
[89] Grieve, K.; Roorda, A. IOVS 2008, 49, 713719.
[90] Roorda, A.; Williams, D.R. Nature 1999, 397, 520522.
[91] Hofer, H.; Carroll, J.; Neitz, J.; Neitz, M.; Williams, D.R. J. Neuroscience 2005, 25, 96699679.
[92] Brainard, D.H.; Roorda, A.; Yamauchi, Y.; Calderone, J.B.; Metha, A.; Neitz, M.; Neitz, J.;
Williams, D.R.; Jacobs, G.H. J. Opt. Soc. Am. A 2000, 17, 607614.
[93] Carroll, J.; Neitz, M.; Hofer, H.; Neitz, J.; Williams, D.R. PNAS 2004, 101, 84618466.
[94] Baraas, R.C.; Carroll, J.; Gunther, K.L.; Chung, M.; Williams, D.R.; Foster, D.H.; Neitz, M.
J. Opt. Soc. Am. A 2007, 24, 14381447.
[95] Choi, S.S.; Doble, N.; Hardy, J.L.; Jones, S.M.; Keltner, J.L.; Olivier, S.S.; Werner, J.S. IOVS
2006, 47, 20802092.
[96] Wolfing, J.I.; Chung, M.; Carroll, J.; Roorda, A.; Williams, D.R. Ophthamology 2006, 113,
10141019.
[97] Duncan, J.L.; Zhang, Y.; Gandhi, J.; Nakanishi, C.; Othman, M.; Branham, K.E.H.; Swaroop,
A.; Roorda, A. IOVS 2007, 48, 32833291.
[98] Martin, J.A.; Roorda, A. Ophthamology 2005, 112, 22192224.
[99] Zhong, Z.; Petrig, B.L.; Qi, X.; Burns, S.A. Opt. Exp. 2008, 16, 1274612756.
[100] Choi, S.S.; Doble, N.; Christou, J. Invest. Ophtha. Vis. Sci. 2004, 45, E-abstract 2794.
[101] Putnam, N.M.; Hofer, H.J.; Dobley, N.; Chen, L.; Carroll, J.; Williams, D.R. J. Vis. 2005, 5,
632639.
3466 K.M. Hampson
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

[102] Hardy, J.L.; Delahunt, P.B.; Werner, J.S. Visual Psychophysics with Adaptive Optics.
In Adaptive Optics for Vision Science: Porter, J., Queener, H., Lin, J., Thorn, K., Awwal,
A.A.S., Eds.; Wiley and Sons, Inc: New York, 2006; pp 363394.
[103] Poonja, S.; Patel, S.; Henry, L.; Roorda, A. J. Refract. Surg. 2005, 21, S575S580.
[104] Yoon, G.Y.; Williams, D.R. J. Opt. Soc. Am. A 2002, 19, 266275.
[105] Dalimier, E.; Dainty, C.; Barbur, J.L. J. Mod. Opt. 2008, 55, 791803.
[106] Lundstro m, L.; Manzanera, S.; Prieto, P.M.; Ayala, D.B.; Gorceix, N.; Gustafsson, J.;
Unsbo, P.; Artal, P. Opt. Exp. 2007, 15, 1265412661.
[107] Ferna ndez, E.J.; Manzanera, S.; Piers, P.; Artal, P. J. Refract. Surg. 2002, 18, S634S638.
[108] Rocha, K.M.; Vabre, L.; Harms, F.; Chateau, N.; Krueger, R.R. J. Refract. Surg. 2007, 23,
953959.
[109] Piers, P.A.; Ferna ndez, E.J.; Manzanera, S.; Norrby, S.; Artal, P. IOVS 2004, 45, 46014610.
[110] Piers, P.A.; Manzanera, S.; Prieto, P.M.; Gorceix, N.; Artal, P. J. Cataract Refract. Surg. 2007,
33, 17211726.
[111] Bille, J.F. J. Refract. Surg. 2000, 16, S608S610.
[112] Awwal, A.; Bauman, B.; Gavel, D.; Olivier, S.S.; Jones, S.; Hardy, J.L.; Barnes, T.B.; Dainty,
C. Characterization and Operation of a Liquid Crystal Adaptive Optics Phoropter,
Proceedings SPIE The International Society for Optical Engineering, 2003, ISSU 5169,
104127.
[113] Manzanera, S.; Prieto, P.M.; Ayala, D.B.; Lindacher, J.M.; Artal, P. Opt. Exp. 2007, 15,
1617716188.
[114] Artal, P.; Chen, L.; Ferna ndez, E.J.; Singer, B.; Manzanera, S.; Williams, D.R. J. Vis. 2004, 4,
281287.
[115] Chen, L.; Artal, P.; Gutierrez, D.; Williams, D.R. J. Vis. 2007, 7, 19.
[116] Hofer, H.; Singer, B.; Williams, D.R. J. Vis. 2005, 5, 444454.
[117] Makous, W.; Carroll, J.; Wolfing, J.I.; Lin, J.; Christie, N.; Williams, D.R. IOVS 2006, 47,
41604167.
[118] Ciuffreda, K.J. Accommodation and its Anomalies. In Vision and Visual Dysfunction; Vol. 1,
Charman, N., Ed.; Macmillan: London, 1991; pp 231279.
[119] Lopez-Gil, N.; Rucker, F.J.; Stark, L.R.; Badar, M.; Borgovan, T.; Burke, S.; Kruger, P.B. Vis.
Res. 2007, 47, 755765.
[120] Wilson, B.J.; Decker, K.E.; Roorda, A. J. Opt. Soc. Am. A 2002, 19, 833839.
[121] Ferna ndez, E.J.; Artal, P. J. Opt. Soc. Am. A 2005, 22, 17321738.
[122] Chen, L.; Kruger, P.B.; Hofer, H.; Singer, B.; Williams, D.R. J. Opt. Soc. Am. A 2006, 23, 18.
[123] Kitaguchi, Y.; Bessho, K.; Yamaguchi, T.; Nakazawa, N.; Mihashi, T.; Fujikado, T. Jpn J.
Ophthalmol. 2007, 51, 456461.
[124] Chui, T.Y.P.; Song, H.; Burns, S.A. IOVS 2008, 49, 46794687.
[125] Rossi, E.A.; Weiser, P.; Tarrant, J.; Roorda, A. J. Vis. 2007, 7, 114.
[126] McQuaide, S.C.; Seibel, E.J.; Kelly, J.P.; Schowengerdt, B.T.; Furness, T.A. Displays 2003, 24,
6572.
[127] Simonov, A.N.; Vdovin, G.; Loktev, M. Opt. Exp. 2007, 15, 74687478.
[128] Vdovin, G.; Loktev, M.; Naumov, A. Opt. Exp. 2003, 11, 810817.
[129] Gustafsson, M.G.L. J. Microsc. 2000, 198, 8287.
[130] Gray, D.C.; Merigan, W.; Wolfing, J.I.; Gee, B.P.; Porter, J.; Dubra, A.; Twietmeyer, T.H.;
Ahmad, K.; Tumbar, R.; Reinholz, F.; Williams, D.R. Opt. Exp. 2006, 14, 714458.
[131] Hampson, K.M.; Chin, S.S.; Mallen, E.A.H. J. Mod. Opt. 2008, 55, 703716.
[132] Fam, H.B.; Lim, K.L. J. Refract. Surg. 2004, 5, S570575.
[133] Jime nez, J.R.; Castro, J.J.; Jime nez, R.; Hita, E. Optom. Vis. Sci. 2008, 85, 174179.
[134] McLellan, J.S.; Marcos, S.; Prieto, P.M.; Burns, S.A. Nature 2002, 417, 174176.
Journal of Modern Optics 3467
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
t
i

P
u
t
r
a

M
a
l
a
y
s
i
a
]

a
t

2
3
:
4
6

2
0

M
a
r
c
h

2
0
1
3

Vous aimerez peut-être aussi