Can perceptual learning be used to treat amblyopia beyond the critical period of visual development?

Abstract

Citation information: Astle AT, Webb BS & McGraw PV. Can perceptual learning be used to treat amblyopia beyond the critical period of visual development? Ophthalmic Physiol Opt 2011, 31, 564–573. doi: 10.1111/j.1475-1313.2011.00873.x

Abstract

Background:  Amblyopia presents early in childhood and affects approximately 3% of western populations. The monocular visual acuity loss is conventionally treated during the ‘critical periods’ of visual development by occluding or penalising the fellow eye to encourage use of the amblyopic eye. Despite the measurable success of this approach in many children, substantial numbers of people still suffer with amblyopia later in life because either they were never diagnosed in childhood, did not respond to the original treatment, the amblyopia was only partially remediated, or their acuity loss returned after cessation of treatment.

Purpose:  In this review, we consider whether the visual deficits of this largely overlooked amblyopic group are amenable to conventional and innovative therapeutic interventions later in life, well beyond the age at which treatment is thought to be effective.

Recent findings:  There is a considerable body of evidence that residual plasticity is present in the adult visual brain and this can be harnessed to improve function in adults with amblyopia. Perceptual training protocols have been developed to optimise visual gains in this clinical population. Results thus far are extremely encouraging; marked visual improvements have been demonstrated, the perceptual benefits transfer to new visual tasks and appear to be relatively enduring. The essential ingredients of perceptual training protocols are being incorporated into video game formats, facilitating home-based interventions.

Summary:  Many studies support perceptual training as a tool for improving vision in amblyopes beyond the critical period. Should this novel form of treatment stand up to the scrutiny of a randomised controlled trial, clinicians may need to re-evaluate their therapeutic approach to adults with amblyopia.

Amblyopia

Amblyopia is a developmental visual disorder characterised by reduced vision, usually in one eye, despite refractive correction and in the absence of overt ocular pathology.¹ It affects approximately 3% of the population^2–5 and accounts for the majority of children’s hospital eye appointments in the UK.⁶ It is caused by a disturbance of normal visual input during the critical period(s) of visual development.⁷ Amblyopia is most frequently associated with anisometropia and strabismus and less commonly with form deprivation.⁸ Whether anisometropia or strabismus is more frequently associated with amblyopia is less clear. For example, some studies have found strabismus to be more prevalent in amblyopes,⁸ others have found anisometropia to be comparatively more common,^2,9,10 while others have found roughly equal proportions of strabismus and anisometropia in amblyopic populations.^11,12 Amblyopia is conventionally treated by correcting any refractive error, surgical correction of any strabismus, a period of optical treatment,^13,14 followed by occlusion of the good eye with a patch or atropine penalisation. Despite successful treatment of many children using occlusion therapy, a substantial number of individuals still suffer from amblyopia later in life. Data from a study that examined over 8,000 subjects suggest that 80% of amblyopia persists beyond 16 years of age.¹⁵ This may be due to failure of detection, lack of response to the original treatment, acuity loss following cessation of treatment, or only partial remediation of the amblyopia. Indeed, a study in 2004 found that amblyopia is fully treated in only 30% of those treated with occlusion therapy.¹⁶

Neural plasticity and the critical period

Even though there is a significant population of amblyopes who may benefit from improved visual acuity, treatment is rarely offered to children beyond 9 years of age, as it was assumed that the visual system lacked the necessary neural plasticity (the ability of neural systems to undergo change) to improve function at this age. The evidence that points towards diminishing neural plasticity as childhood advances, is derived in-part from cases of deprivational amblyopia associated with cataracts.¹⁷ When cataracts develop after an ocular injury, it is possible to accurately determine the precise onset and duration of the visual disturbance. Once a cataract is removed, the severity of the resulting visual deficit is strongly dependent on the both the age of onset of the visual disturbance and how long it was present for. The earlier the onset and longer the period of deprivation, the worse the visual outcome. Short periods of deprivation early in life can have profound effects on vision. In contrast, normal levels of vision are present on removal of a cataract, even if present for lengthy periods of time, provided the onset of the cataract occurred after 9 years of age.¹⁷ In other words, amblyopia is not induced in those older than 9 years of age. The period of time during which both deprivation can affect visual development, and amblyopia can be induced, is referred to as the critical period of visual development.⁷

We now know that there is not a single critical period for the developing visual system, but that the first 9 years of childhood represents the envelope of a number of overlapping periods, each of which have different start and end points.^18,19 For example the critical period for monocular deprivation ends later than the critical period for deprivation of visual motion direction.²⁰ Some have suggested that plasticity is an intrinsic feature of the brain and that normal development consists of structural and functional changes that ultimately result in stabilisation.²¹ This process proffers the dual advantages of adaptability during the changing conditions of infancy and efficient neural structures in adulthood.²¹ However, there is also widespread evidence to suggest that the brain possesses significant neural plasticity in adulthood. Good examples include the extensive cortical remapping observed in adult amputees²² and stroke victims.²³ It is also illustrated by the enlarged sensorimotor cortical representation of the reading finger in blind subjects who become skilled at reading Braille.²⁴

The concept of a critical period for the disruption of vision (and associated development of amblyopia) has often led to the assumption that this corresponds to the period of time during which vision can be improved (and amblyopia treated).²⁵ However, there is a considerable body of evidence to suggest that residual plasticity is present in the adult visual brain and even though the nature of the plasticity present in adulthood may differ from that present in infancy,²⁶ it can be harnessed to improve visual function in adults with amblyopia.²¹ Below we outline evidence to support this contention. We draw on evidence from studies that have used both conventional and innovative therapeutic interventions to improve vision in amblyopes at ages where treatment was thought to be ineffective, well beyond the critical period of visual development.

Evidence to support treatment of amblyopia beyond the critical period: traditional treatment

Many studies support the notion that amblyopia can be treated beyond the critical period.^27–34 In 2005 a large scale randomised trial carried out by the Paediatric Eye Disease Investigator Group (PEDIG),³⁵ which included over 500 amblyopic subjects, showed that refractive correction alone can improve vision in 25% (50/203 subjects) of 7–12 year olds. Furthermore, a combination of refractive correction, occlusion therapy, atropine penalisation and near work improved vision in 53% (106/201 subjects) of those in the same age group. In an older group of children, aged 13–17 years, who had received no previous treatment, 47% (8/14 subjects) improved with a combination of refractive correction, occlusion therapy and near work. Therefore, there is very little difference in the success rates between 7 to 12 and 13 to 17 year olds who had not been previously treated. Another study that tested 102 amblyopes with anisometropia found no difference in the benefit of treatment from patching between groups aged 6–12 and 13–20 years.²⁸ These are important findings because they suggest that the period of time that amblyopia can be treated stretches to at least the late teens. Others have also found evidence for the ability of the amblyopic visual system to improve later in life following occlusion therapy. For example, the vast majority of amblyopes aged 9–14 years (94%)²⁹ and 10–16 years (81%)³⁰ improve by at least 0.2 logMAR after occlusion, and an average improvement of 0.46 logMAR has been found in 11–15 years olds with amblyopia.³¹ Simmers and colleagues³² found evidence for improvements on a range of visual measures in amblyopes receiving occlusion and no relationship between the degree of improvements found and age. In 1977, Birnbaum and colleagues³⁴ conducted a literature survey of twenty-three studies looking at the effects of a range of treatments (including occlusion and pleoptics) on amblyopic patients under the age of 7 and those over the age of 7. They found no statistical difference between the success of treatment between the two groups. Additionally, improvements in visual acuity in the amblyopic eye have been found following loss of vision in the good eye,^36,37 or loss of the good eye³⁸ in adults. All of these studies provide evidence of plasticity in the amblyopic visual system after the period during which amblyopes were thought to be sensitive to treatment. This suggests that a failure to offer treatment to individuals over 7 years of age could potentially deprive them of improved visual function.

Evidence to support treatment of amblyopia beyond the critical period: Perceptual learning

More recently, psychophysical studies have demonstrated plasticity in the adult visual system, attributed to a phenomenon referred to as perceptual learning. Perceptual learning describes permanent and consistent improvements in performance on sensory tasks as a result of experience or practice.³⁹ Perceptual learning has been demonstrated on a variety of tasks in adults with normal vision (see Fine & Jacobs⁴⁰ for a review). Perceptual learning does not simply refer to the improvements you may find as a result of becoming more familiar with a particular task procedure. This is illustrated by the fact that improvements in performance are often strongly coupled to trained visual attributes (e.g. direction of motion⁴¹) or tasks^41–44 in otherwise unchanged procedures or tasks. Improvements in performance are also found in individuals who are experienced and highly familiar with the testing procedures – which would not be expected if improvements were due to learning of task procedure or some other general cognitive strategy.⁴⁵ Additionally, the better levels of vision demonstrated after training on tasks such as Vernier acuity are not thought be due to improvements in accommodation or fixation accuracy since they are relatively insensitive to blur⁴⁶ and image motion^47,48. Instead, improvements are thought to be due to fundamental alterations in the cortical processing of information required to reach a sensory decision.⁴⁹ Physiological correlates to changes in performance found with perceptual learning have been demonstrated by electrophysiological studies, which have shown changes in response properties of early visual neurons in conjunction with improved behavioural sensitivity^43,50 and fMRI studies that have shown increased activity of V1 neurons following a period of training on a visual texture discrimination task⁵¹ and an orientation discrimination task.⁵² Whether improvements are due to a change to the representation or readout of visual cortex is hotly debated.⁵³

Perceptual learning has also been reported in adults with amblyopia (see Levi & Li²⁵ for a review). What is particularly interesting about the improvements found in amblyopes is that they appear to be more general than those found in normal subjects - as training also leads to improvements on tasks with stimuli or stimulus parameters which have not specifically been trained.^54–59 This is extremely important if it is to be considered as a potential way of improving the amblyopic vision, which includes a wide variety of deficits in visual processing.⁶⁰ We will now consider the essential ingredients of perceptual learning.

Perceptual learning is an active process. Those being trained are actively engaged in a visual task and participation involves making judgments based on characteristics of the stimuli. Many of the training tasks used involved repeat measurements of thresholds using systematic psychophysical procedures, which require active engagement of attention, and where feedback is provided on each trial.⁵⁴ This sets perceptual learning apart from some forms of treatment previously proposed to improve amblyopic visual performance (e.g. pleoptics). Amblyopes receiving occlusion therapy or atropine penalisation will be exposed to a range of visual stimulation during the course of everyday activities but much of this will be passive.⁶¹ More than 30 years ago the Cambridge visual stimulator was presented as a potential treatment tool for amblyopia. It involved exposure to rotating gratings of various spatial frequencies while the good eye was occluded, and resulted in improvements in vision of amblyopic subjects.⁶¹ However, when compared to the improvements found in a control group undergoing the same procedure but without exposure to the gratings, the improvements were not significantly different.⁶² Rather than passive exposure to gratings, improvements are likely to have been driven by occlusion and common near activities. Perceptual learning studies involve subjects undergoing intense periods of training under strictly controlled experimental conditions. This is in contrast to occlusion therapy where there is little control over the visual experience of the amblyopic eye outside the eye clinic, or indeed the amount of occlusion.⁶³

In addition to perceptual learning being an active process, perceptual learning studies present subjects with stimuli within a certain visibility range that may be important for improvements to occur. By repeatedly measuring performance close to its limits (i.e. detection or discrimination thresholds), subjects are exposed to and have to make judgments on stimuli that are close to their individual threshold. Again, this is different to the Cambridge visual stimulator and different from most of the visual experience encountered during the course of occlusion or penalisation. Using a psychophysical procedure such as an adaptive staircase (as is often used in perceptual learning) will mean that most stimuli are presented close to the limits of their performance. Stimuli that are just above threshold will be visible but difficult to see. As performance improves, stimulus intensity is reduced adaptively. These incremental shifts in stimulus intensity may be important since animal studies have shown that it is possible to adjust to shifts in visual experience provided that they are sufficiently small, but not possible if they are too large.⁶⁴ Task difficulty, or precision, is thought to have an important effect on the degree of transfer of performance improvements to other tasks, with studies showing more generalisation following training on easier tasks⁶⁵ or when the transfer task has lower precision demands.⁶⁶

Improvements in performance associated with perceptual learning follow a characteristic time course. Improvements commonly occur at an exponential rate, with greater improvements in performance found with longer durations of training until performance reaches a plateau.⁵⁶ The amount of training required to reach a steady level appears to be dependent on the initial size of the deficit, with greater time required to reach asymptotic performance (and larger improvements found) for amblyopes with poorer initial performance.⁴⁵ Once a plateau is reached, continued training may result in further improvements – as demonstrated by Li and colleagues^45,56 following training on a positional acuity task. We also found this for an amblyopic subject who trained on a letter based contrast sensitivity task.⁶⁷ The subject practiced the task for 25 sessions in total (see Figure 1). Data from sessions 1 to 12 and from sessions 13 to 25 were fitted with exponential functions demonstrating two distinct regions of improvements in performance. The finding that there are multiple cascades in performance improvements with perceptual learning, that there are individual variations in the response to training, and variable effects of training on different tasks means that both determining a dose-response function for perceptual learning and deciding on an optimum point at which to cease training is an important challenge for future research.⁴⁵ A recent study⁶⁸ directly compared the time course of perceptual learning on a contrast detection task to occlusion. They found that a 0.1 logMAR improvement in visual acuity required 154 h of occlusion. Those receiving perceptual learning improved by 0.25 logMAR on average. This would have required 385 h of occlusion, yet the improvement was achieved in around 1/13th of this time. Others have compared the improvements found in adult amblyopes with perceptual learning to the duration of occlusion that would be required to achieve an equivalent improvement with occlusion therapy in children.⁶⁹ For example, the improvement in visual acuity found in adult amblyopes following training on a contrast based task⁵⁵ would require 500 h of occlusion in children.⁷⁰

Figure 1.  Data from one adult amblyopic observer who practiced a letter based contrast sensitivity task. The subject trained for 25 sessions in total. Data for sessions 1–12 are fitted with an exponential function shown by the solid line. Data for sessions 13–25 are fitted with an exponential function shown by the dashed line.

Even though improvements in amblyopic visual performance have been found for a range of visual tasks, improvements are not equivalent for all tasks and we should consider training with stimuli configurations that lead to the greatest possible improvements in visual function, targeting the deficits associated with amblyopia. A factor analysis of 427 amblyopic subjects has shown that the amblyopic visual deficit can be well characterised in terms of visual acuity and contrast sensitivity performance.⁶⁰ We have demonstrated that training on contrast based tasks leads to greater degrees of improvement and transfer compared to training on acuity based tasks.⁶⁷ Most perceptual learning studies in adults with amblyopia have found the largest improvements in performance with contrast-based tasks,^55,57,68,71 with improvements of up to 70% reported.⁵⁷ Moreover, training on tasks using stimuli which are broadband (in spatial frequency and orientation) and crowded appear to result in greater amounts of learning compared to tasks using narrowband, uncrowded stimuli.⁶⁷ Amblyopic subjects are characteristically deficient in stereoacuity and crowding, functions which can be improved with appropriate training (Hussain Z, Webb BS, Astle AT & McGraw PV, unpublished data).⁷²

Are improvements long lasting? Regression of visual acuity is a critical issue, as even a small reduction in visual acuity could potentially have important implications on visual standards and therefore on the ability to undertake certain occupations.⁷³ Numerous studies have investigated the long term outcome of amblyopia treatment. One study has shown that 30% of those successfully treated with traditional methods regress to pre-treatment acuity levels.⁷⁴ The proportion of amblyopes whose visual acuity regresses after treatment will be dependent to some extent on the clinical characteristics of the population sampled. For example, one study showed a regression of visual acuity in 42% of amblyopes whose visual acuity was better than 6/30 prior to treatment and in 63% of those with acuity worse than 6/30.⁷⁵ A recent randomised controlled trial has provided further evidence for visual improvements in amblyopes in older age groups following intermittent photic stimulation, which involves near work and exposure to a flickering red background.⁷⁶ However, these improvements did not endure. In contrast, the effects of perceptual learning in amblyopic subjects so far appear to be relatively permanent and long-lasting. For example, improvements in contrast sensitivity have been demonstrated to be almost fully retained after 12 months⁵⁵ and 18 months⁷¹ in separate studies, supporting the notion that the targeted approach of perceptual learning may be more enduring than previous treatments.^{54,55,68,71,77} One could argue that the improvements found with perceptual learning represent a re-uptake of the improvements found with occlusion therapy and later regress after cessation of occlusion. It is difficult to determine whether or not this is the case as accurate records of post- treatment visual acuity are rarely available for adult amblyopes and thus quantifying the effect of visual acuity regression after treatment is problematic. However, improvements in the visual performance of amblyopes older than 8 years of age, who have not previously been patched, have been found with perceptual learning.⁷⁸

Is the effect of treatment influenced by age? It appears that the improvements found with perceptual learning occur irrespective of age. We trained amblyopic subjects on a variety of acuity and contrast based visual tasks⁶⁷ and found no significant correlation between the amount of improvement and the age of the subjects (r₂₇ = 0.26, p = 0.17; see Figure 2). Additionally a recent review of perceptual learning in amblyopia analysed the findings of 15 studies and also found no significant relationship between improvement and age, for subjects aged 10–40 years.⁶⁹ A number of studies have demonstrated improvements in younger children with amblyopia following periods of perceptual learning (e.g.^56,78,79), even in subjects who showed poor compliance to treatment with occlusion or did not respond to occlusion.⁷⁰ Therefore, perceptual learning could potentially benefit a large proportion of amblyopes well in to adulthood and could supplement occlusion therapy in children.

Figure 2.  Change in performance following training on a variety of acuity and contrast based tasks, expressed as a ratio of start performance (PPR) relative to subject age. A lower PPR corresponds to a greater improvement in performance. The horizontal dashed line PPR = 1 corresponds to no improvement. Lines lying below this line denote individuals who improved in performance. There is no significant relationship between improvement and age of subjects (r₂₇ = 0.26, p = 0.17).

Are improvements found in perceptual learning studies purely due to occlusion received during training? Since many perceptual learning studies involve subjects wearing a patch over their good eye while carrying out training, it raises the question as to whether improvements found in these studies can be attributed to occlusion alone. A number of perceptual learning studies have addressed this issue. One study found a relatively large (62%) improvement in performance following training on a contrast detection task.⁵⁵ However, no improvements were found in control groups – despite them also receiving occlusion. A more recent study, which trained amblyopic subjects on an array of tasks, showed that improvements were greatest when a letter based contrast task was trained and that no improvement in performance was found when subjects trained on a grating acuity task – despite each group being patched for the same duration. Subjects who showed no improvement following training on the grating acuity task went on to improve significantly when trained on the letter contrast task.⁶⁷ Therefore it is the change in task rather than occlusion that lead to improved performance, as both tasks had occlusion in common. Others have shown that the improvements found on a position acuity task are greater when subjects receive training in addition to occlusion compared to occlusion alone.⁵⁶ All this suggests that visually targeted training is the important factor, rather than occlusion per se. Other studies that have not used occlusion, but have instead relied on forms of binocular training^80–81, have also found improvements in amblyopic visual performance including stereoacuity.⁷²

Improvements in amblyopic visual function with computer game use

There is evidence that playing computer games results in visual improvements in people with normal^84,85 and amblyopic⁸⁶ vision. For example, playing action video games can improve the contrast sensitivity of players⁸⁵ something which has been found to result from more tailored training of amblyopic subjects in perceptual learning studies.⁷¹ Computer games contain many of the important ingredients for successful learning (for a review see⁸⁷) and may be a way by which amblyopes could improve their vision in their own homes. Many shooting based games will involve detection of a target and require precise positional placement of a pointer on the target (like Vernier acuity). They will also present images composed of a range of spatial frequencies, contrasts, colours and degrees of crowding. There is feedback and a reward for good performance, engagement in the task, along with an incentive for detecting and correctly aligning stimuli presented at a challenging difficulty level.

Although both perceptual learning studies and studies investigating the effects of repeated sessions of playing a computer game have found improvements in adult amblyopia, the two methods differ: perceptual learning studies attempt to determine the specific stimulus and training characteristics that lead to improvements along with the mechanisms behind the improvements. Once these are understood it will be possible to build these, in a bottom-up manner, in to a game that encapsulates these important components, thus maximising improvements – which is our intention. Studies investigating computer game use have established sizeable improvements in adult amblyopia after a few sessions of gameplay. In an attempt to reverse engineer the critical components of these games, the features which lead to perceptual improvements can be systematically investigated. Both approaches are working on the same problem while tackling it from different angles. There is inevitable overlap and results from each approach will be mutually informative.

Future directions

Perceptual training is not currently available in clinics as a treatment option for amblyopia. It is important that any strategy used to treat adult amblyopia is fully evidence based. There are already a large number of studies that provide strong support for visual improvements in adult amblyopia. However, as highlighted by research on the Cambridge visual stimulator, validation of prospective treatment strategies with randomised controlled trials is important. First, a randomised controlled trial of occlusion in adults is required. Second, the same methodology should be used to assess the benefit of perceptual learning and compare it to that from occlusion therapy. If it is determined that perceptual learning offers a significant clinical benefit, the next step would be the transition of laboratory-based paradigms to clinical or home-based settings – perhaps tailoring the important components of perceptual learning into an optimised computer game design. This is an area that has already received some attention.⁸⁸

We are not suggesting that perceptual learning should replace early interventions, but rather supplement them at an appropriate age. The treatment of amblyopia in adults would offer those detected at an age traditionally deemed too old to be treated the opportunity of improving their vision, potentially opening up career opportunities⁷³ and reducing the likelihood of further visual impairment in later life.⁸⁹ It may also offer those who are not responsive to occlusion therapy a way of improving vision.⁷⁹ Furthermore, it could provide a way of reducing the amount of patching required with occlusion therapy. This would be beneficial given the potential importance of periods of binocular stimulation⁹⁰ and the many reported drawbacks of occlusion therapy including poor compliance,⁶³ a potential reduction in binocularity and negative psychosocial consequences.⁹¹ Given the existing body of evidence, it may be unreasonable not to treat adults with amblyopia if they want to improve their vision.

Conclusion

There is extensive evidence for the ability of the amblyopic visual system to improve after the critical period of visual development. Therefore, the long held and common view that treatment is precluded in amblyopes on the basis that they are too old may need to be revised if we are to offer the best treatment options for amblyopic patients.

Appendices

For details of the authors of this review please see the next page.

Andrew Astle

Andrew Astle received his undergraduate degree in optometry from the School of Optometry & Vision Sciences, Cardiff University in 2002. In 2007 he was awarded a postgraduate research scholarship from The College of Optometrists and completed a PhD in the Visual Neuroscience Group, University of Nottingham in 2010. His research interests include spatial vision, depth perception, and amblyopia.

Ben Webb

Ben Webb graduated in 1998 with BSc in psychology from University of London, Goldsmiths College. He moved to University of Oxford to complete an MSc in Neuroscience, funded by a Medical Research Council Studentship (awarded 1999). The following year he was awarded a Wellcome Trust Prize PhD Studentship at the School of Psychology, University of Nottingham to investigate the transmission of neural signals from lateral geniculate to visual cortex (awarded 2003). After his PhD he took up a post-doctoral position at the Center for Neural Science, New York University to search for the origin(s) of suppressive neural signals in early visual cortex. In 2006, he was awarded a Leverhulme Trust Early Career Fellowship to investigate the neural mechanisms underpinning human perceptual learning. He is currently funded by a Wellcome Trust Research Career Development Fellowship to investigate the neural computations mediating perceptual decisions.

Paul McGraw

Paul McGraw gained an undergraduate degree in optometry in 1990 and after spending a year working at the Glasgow Eye Infirmary, returned to complete his PhD at Glasgow Caledonian University. He then moved to take up a post-doctoral position at the School of Optometry, University of Bradford. In 1997, he was awarded a Vision Research Training Fellowship from The Wellcome Trust to study the mechanisms involved in common misperceptions of spatial vision. At the end of this Fellowship he secured a subsequent 4-year Research Career Development Fellowship from The Wellcome Trust to study information processing in the human visual cortex using non-invasive techniques. In 2004 he was appointed as the Allen Standen Reader in Experimental Psychology at the University of Nottingham. In 2005, he took up a 5-year University Fellowship, funded by the Wellcome Trust, to investigate how the brain processes visual and auditory motion information. He was awarded a personal Chair in Visual Neuro-science at the University of Nottingham in 2006 and is currently the Head of the School of Psychology.