Impaired Temporal, Not Just Spatial, Resolution in Amblyopia

Abstract

Purpose.: In amblyopia, neuronal deficits deteriorate spatial vision including visual acuity, possibly because of a lack of use-dependent fine-tuning of afferents to the visual cortex during infancy; but temporal processing may deteriorate as well.

Methods.: Temporal, rather than spatial, resolution was investigated in patients with amblyopia by means of a task based on time-defined figure–ground segregation. Patients had to indicate the quadrant of the visual field where a purely time-defined square appeared.

Results.: The results showed a clear decrease in temporal resolution of patients' amblyopic eyes compared with the dominant eyes in this task. The extent of this decrease in figure–ground segregation based on time of motion onset only loosely correlated with the decrease in spatial resolution and spanned a smaller range than did the spatial loss. Control experiments with artificially induced blur in normal observers confirmed that the decrease in temporal resolution was not simply due to the acuity loss.

Conclusions.: Amblyopia not only decreases spatial resolution, but also temporal factors such as time-based figure–ground segregation, even at high stimulus contrasts. This finding suggests that the realm of neuronal processes that may be disturbed in amblyopia is larger than originally thought.

Typically, amblyopia, a developmental anomaly of spatial vision, ¹ is characterized as a loss of contrast sensitivity and spatial resolution in an apparently healthy eye, partly due to abnormal binocular interaction. ^2,3 Spatial uncertainty about stimulus position is higher in amblyopic eyes than in their dominant partner eyes, ⁴ apparently due to noise in the system. ⁵ The underlying causes are certainly of neuronal origin, and several experimental results in animals with artificially induced squint demonstrated very clear changes in neuronal connectivity. Hubel and Wiesel ^6,7 first described a dramatically decreased percentage of binocular cortical neurons, which are activated equally well through both eyes, as a result of squint induced in young cats or monkeys. Similarly, patching one eye, thus simulating unilateral lens opacity (cataract), led to a pronounced increase of cortical neurons activated through the open eye while simultaneously decreasing visual acuity ⁷ and the percentage of neurons activated through the patched eye in monkeys. ⁸ More recently, Montero ⁹ found that amblyopia decreases the activation produced by visual stimuli in the visual thalamic reticularis and visual cortex of rats. 

It follows that amblyopia involves changes in neural processing on several levels in the visual system, not just in the striate cortex. These changes seem to be permanent once adulthood is reached, with no therapy available. As a drastic example, enucleation of the dominant (nondeprived) eye in amblyopic monkeys after prolonged lid suture of the deprived eye did not significantly improve spatial or temporal sensitivity in these adult animals. ¹⁰ However, recent evidence suggests that prolonged training may significantly improve visual acuity in amblyopic eyes in humans, ^11,12 possibly through retuning of a decision template ¹³ (see also Refs. 14–16, as well as Ref. 17 for a very atypical case of amblyopia). 

In addition to shifts in the size of cortical representations these representations may also be less organized topographically in amblyopic than in normal eyes (e.g., Refs. 18–23) and inhibition from the dominant eye may play a role. ^14,24  

Several studies indicate that temporal aspects of vision may be altered in amblyopia, too. ²⁵ In the study by Harwerth et al., ²⁶ the difference in contrast sensitivity between the amblyopic versus normal eyes of amblyopic monkeys was larger at high than at low temporal frequencies, with several studies in humans yielding similar results. (See Fig. 2 in Ref. 15, and see Refs. 27 and 28 for longer integration times at high spatial frequencies.) Duration of exposure to stimuli seems to have more of an effect on the difference between amblyopic and normal eyes at high temporal frequencies than at lower ones. ²⁸ Several studies, ^29–31 on the other hand, found no temporal deficits in amblyopic humans if the stimuli were sufficiently large and sufficiently above detection threshold. 

Results of flicker fusion frequency (FFF) for bright, high-contrast stimuli are somewhat controversial: slight decreases, ³² mixed results, ^33,34 and higher FFFs in the amblyopic eye were reported. ³⁵ The differences between study results may partly rely on different diameters of the stimuli used. ³⁶  

Reaction times for stimulus appearance and latencies of evoked potentials were somewhat slower through amblyopic eyes (anisometropic amblyopia) than through their normal partner eyes. ^37,38 In line with these results, latencies of evoked potentials for high spatial frequencies increase in human amblyopia, ³⁹ and stimulation latencies of single cells increase in the feline primary visual cortex. ⁴⁰  

However, as mentioned earlier, the differences between amblyopic and normal eyes of patients usually disappear when the contrast of the stimuli is equated to accommodate the lower contrast sensitivity of the amblyopic eye, indicating that there is no specific loss in the time domain but just an expected loss due to decreased contrast sensitivity. ^4,37 Decreased contrast sensitivity may also at least partly explain the decreased persistence of stimuli presented to the amblyopic eye. ⁴¹ Here, we present evidence that the temporal precision of processing and binding is significantly decreased in amblyopic eyes even in response to large stimuli far above contrast-detection threshold, close to contrast saturation. 

Material and Methods

The basic stimulus in our study was almost identical with the one used in Kandil and Fahle. ⁴² It consisted of a regular grid of 20 × 20 randomly oriented colons (dipoles), which flipped around their imaginary midpoints by 90° at a given frequency (Fig. 1A). The colons flipped synchronously within the target as well as within the background, but asynchronously—that is, with a defined phase delay—between target and background. The target region was defined by a square consisting of 3 × 3 synchronously flipping colons. For sufficiently long delays of the target flips relative to the background flips, the target region became detectable due to the time difference between apparent motions in figure versus ground. Subjects had to indicate the quadrant in the visual field containing the target by pressing the corresponding cursor key on a standard computer keyboard in a four-alternative, forced-choice task. 

Figure 1.

 

Schematic description of the stimuli used to test the temporal resolution for amblyopic eyes. (A) Right: spatial layout of the stimuli that were rotating at different times with no single frame allowing the discrimination between figure and ground. Gray squares (not present in the original stimuli) indicate the possible positions of the target in the four quadrants of the visual field. Left: a short time sequence of two “colons” positioned across the figure–ground border, as indicated on the right. Frame length was varied to determine detection thresholds. (B) Basic principles of the two types of designs used: in the condition varying the phase angle, the (flicker) frequency of the stimulus stayed constant, whereas the phase angle decreased, leading to a decreasing time difference between movements in figure versus ground. In the frequency modulation condition, the phase angle between movements in figure and surround was fixed at 180°, whereas the absolute time difference between the movements in figure and ground again decreased, but due to increasing presentation frequency.

The distribution of the colons remained constant during each stimulus presentation. All colons had the same contrast and were randomly oriented, and motion directions were ambivalent (clockwise versus counterclockwise) and identical in figure and ground. The segregation of the target region was neither signaled by any luminance, orientation, nor global motion cues. It is important to note that figure–ground segregation with this stimulus cannot be achieved by a first-order motion system, but requires a second stage that detects time differences between the onset of motion in figure versus surround. 

We used two designs to determine threshold time delays (Fig. 1B). In the phase-reduction design, the frequency remained fixed at 4.2 Hz, corresponding to an interval of 240 ms between flips, whereas the delay was reduced stepwise from 120 to 10 ms, with minimal step size corresponding to the reciprocal of the frame rate. In the frequency-modulation condition, flip frequency increased from 4.2 to 50 Hz, corresponding to a decrease of the interval between flips from 240 to 20 ms, whereas the phase difference was constant at 180° (i.e., always in counterphase), resulting in delays between 120 and 10 ms. Block size of each condition was 80 presentations, with delay in each block adjusted by means of an adaptive staircase procedure. ⁴³ Thresholds were calculated by interpolating the minimum delay required to obtain 62.5% correct responses, since this level is midway between chance level and perfect performance. 

Stimuli were presented on a 21° color CRT monitor (FlexScan T 662-5; Eizo Nanao Corp., Ishikawa, Japan) with a spatial resolution of 640 × 480 pixels and a frame rate of 150 Hz. Subjects were seated in a dimly lit room and viewed the display from a distance of 40 cm. The stimulus displayed subtended 37° of visual angle. The mean distance between colons was 111 arcmin and the distance between the two dots constituting a colon was 32 arcmin. Each dot had a diameter of 10.5 arcmin. Stimulus luminance was ∼89 cd/m² on a background of 0.3 cd/m², resulting in a Michelson contrast of 0.99. The stimulus was shown for up to 3 seconds or until the observer replied by pressing one of the response buttons. To slow down perceptual learning, we provided no error feedback. 

Additional control experiments were performed with normal observers. In the first control experiment, the described display was used, but the observer's dominant eye was blurred in monocular vision by positive-diopter lenses to produce three levels of reduced decimal visual acuity—namely 0.5, 0.3, and 0.1. In the second experiment, the stimuli were blurred by one of two low-pass Gaussian filters placed on the elements of the stimulus, simulating a reduction of visual acuity to 0.3 or 0.1 (Fig. 2). The Gaussian filters used to blur the original dots of 10.5 arcmin diameter had a half-width of σ = 8.6 arcmin to produce a visual acuity of 0.3 and of σ =15.25 arcmin for 0.1. Both blurring methods were used to test the threshold of temporal delay necessary to detect the target region in the phase reduction, as well as in the frequency-modulation design. 

Figure 2.

 

Diagram of the original stimulus (A) in comparison with two versions blurred by different Gaussian filters (B) and (C).

Forty-one patients who hade amblyopia of different origins (Table 1) participated in the study. They were aged between 17 and 65 years and were naïve to the purpose of the study. With three exceptions, all had normal or corrected-to-normal visual acuity in the dominant eye (Table 1). Ten normal observers, aged between 22 and 41 who were naïve to the purpose of the study served as control subjects. The experiments were approved by the Bremen Ethics Review and Approval Committee. Informed consent was obtained from the subjects after an explanation of the nature and possible consequences of the study and the tenets of the Declaration of Helsinki were strictly observed. 

Table 1.

 

Clinical Data of Patients with Amblyopia

Observer	Age (y)	Affected Eye	Visual Acuity of Amblyopic Eye	Visual Acuity of Dominant Eye	Cause
1	65	L	0.8	1.25	Anisometropia
2	23	R	0.8	1.25	Anisometropia
3	36	L	0.63	1.0	Anisometropia
4	46	L	0.63	1.0	Anisometropia
5	44	L	0.63	0.8	Anisometropia
6	31	R	0.63	1.25	Anisometropia
7	47	L	0.63	1.0	Anisometropia
8	30	L	0.5	1.25	Anisometropia
9	35	L	0.5	1.25	Anisometropia
10	33	L	0.5	1.25	Anisometropia
11	25	L	0.4	1.0	Anisometropia
12	28	R	0.4	0.8	Anisometropia
13	26	R	0.32	1.0	Anisometropia
14	32	L	0.32	1.25	Anisometropia
15	34	R	0.32	1.25	Anisometropia
16	38	R	0.32	1.25	Anisometropia
17	38	L	0.2	1.0	Anisometropia
18	38	L	0.1	1.6	Anisometropia
19	33	L	0.05	1.25	Anisometropia
20	24	L	0.05	1.25	Anisometropia
21	32	L	0.016	1.0	Anisometropia
22	37	L	0.63	1.25	Squint
23	37	R	0.5	1.25	Squint
24	40	R	0.5	1.0	Squint
25	17	R	0.5	1.0	Squint
26	40	R	0.4	1.0	Squint
27	54	R	0.4	1.25	Squint
28	39	R	0.32	0.63	Squint
29	34	L	0.32	1.0	Squint
30	27	L	0.25	0.8	Squint
31	32	R	0.2	1.25	Squint
32	43	R	0.2	1.25	Squint
33	33	L	0.125	1.0	Squint
34	45	R	0.1	1.25	Squint
35	55	L	0.1	1.25	Squint
36	37	R	0.04	1.25	Squint
37	49	R	0.025	1.25	Squint
38	38	R	0.1	1.25	Squint?
39	45	L	0.033	1.0	Squint?
40	32	L	0.8	1.25	Unknown
41	24	L	0.63	1.0	Unknown

Results

The experimental results of our patients are shown in Figure 3. Visual acuity in these amblyopic patients varied over a wide range, from 0.8 (corresponding to 18/20, i.e., normal acuity) to ∼0.05 (i.e., almost a factor of 20 lower). Generally, patients who had lower visual acuity tended to require longer temporal delays between movements in figure and ground to detect the target than did patients who enjoyed a better visual acuity. The correlations were relatively loose but still significant due to the large number of patients (R ² = 0.24; P = 0.001; F-tests; Fig. 3A). The range of variation between temporal delays required by individual patients was smaller than it was for visual acuity, but still accounted to a factor of ∼5 between the lowest and the highest thresholds. Temporal thresholds of the amblyopic eyes as a group were significantly higher than those for the partner eyes (one-sided t-test; P = 0.001) and than those for age-matched control groups. ⁴²  

Figure 3.

 

(A) Results of all amblyopic patients in the phase- reduction (i.e., constant frequency condition). The ordinate plots the temporal delay required by each of the patients to correctly indicate the quadrant in which the target appeared with a probability of 0.63. The abscissa plots the corresponding visual acuity of each patient. Visual acuity varies over a wide range, between 0.8 (corresponding to 16/20) and 0.05 (1/20)—a decrease of acuity by a factor of almost 20. Temporal thresholds tend to be higher and performance tends to be lower, for patients who have low acuities. (B) Results of the same amblyopic patients as in (A), but here are shown the thresholds for the frequency modulation condition (i.e., constant phase of 180°). The results are quite similar to those obtained when varying the temporal phase angle between figure and ground.

Basically, the same was true for the results of the frequency variation condition: a significant correlation between visual acuity and temporal thresholds for figure–ground segregation, with generally better temporal thresholds achieved by those patients with better visual acuity (R ² = 0.22; P = 0.002; Fig. 3B). Results were better (i.e., thresholds were lower) for younger than for older patients. This finding held true for both the constant frequency (Fig. 4A) and the constant phase condition (Fig. 4B). Surprisingly, the decrease in performance with age was far shallower for the amblyopic eyes than for the dominant eyes. In the amblyopic eyes, the correlation was not significant with R ² = 0.04; P > 0.1 for constant frequency and R ² = 0.03; P > 0.1 for constant phase. In the dominant eyes, the decrease with age was highly significant, with R ² = 0.22; P = 0.002 for constant frequency and R ² = 0.16; P = 0.01 for constant phase. Hence, with advancing age, performance of the dominant eyes in this task gradually becomes indistinguishable from that of the amblyopic eyes. 

Figure 4.

 

Threshold for all patients in both (A) the phase-reduction (constant frequency) and (B) the frequency-variation (constant phase) condition, for both the dominant (circles) and the amblyopic eyes (diamonds). In both conditions, the threshold delays required for figure–ground segregation increased with age, indicating decreasing performance. This decrease in performance was significant only for the dominant eyes (circles). As a consequence of the steeper decrease found for the dominant eyes, the difference in performance between the two eyes disappears in older patients. Note the logarithmic scale of the ordinate. The data were fitted by means of an exponential function of the form: y = a · e^bx .

To assure that the loss of temporal resolution in amblyopic eyes is not simply due to the loss of spatial resolution associated with low visual acuity in amblyopia, we blurred the dominant eye of normal observers in two ways, thus matching the visual loss in amblyopia. If temporal deterioration in the amblyopic eyes would be caused by decreased spatial resolution, blurring the stimuli in normal observers should also increase the delay required to detect the target. Somewhat surprisingly, blurring with positive-diopter lenses—resulting in lower acuity—on the contrary even slightly shortened the temporal delay necessary to detect the target at least in the constant-phase condition (Fig. 5). 

Figure 5.

 

Threshold for dominant eyes of normal observers with blur induced by positive-diopter lenses, resulting in visual acuities of 0.1, 0.3, and 0.5, respectively, compared to results without lenses (1.0). (A) Constant-frequency condition; (B) constant-phase condition.

Blurring the stimuli mathematically with a Gaussian filter (i.e., presenting blurred stimuli on the monitor) on average also shortened the temporal delay required in the phase-reduction design and only minimally increased the required temporal delay in the frequency-modulation condition (Fig. 6). The slope of increase differs significantly from the slope obtained in the frequency-modulation design of the amblyopic patients. 

Figure 6.

 

Threshold for dominant eyes of normal observers with blur induced by Gaussian filters, resulting in visual acuities of 0.1 and 0.3, respectively, compared to results without filters (1.0). (A) Constant-frequency condition; (B) constant-phase condition.

Discussion

As mentioned earlier, amblyopia is usually considered as a disturbance of spatial vision, maybe with an additional loss of contrast sensitivity for stimuli of high temporal and spatial frequencies. Our results clearly indicate that amblyopia not only decreases spatial resolution, but obviously patients with deep amblyopia, on average, require significantly higher temporal delays between figure and ground to segregate the two and to perceive a target defined exclusively based on temporal factors. Thresholds increase, on average, by about a factor of three between the patients with the highest compared to those with the lowest visual acuities, while the corresponding visual acuities differ by a factor of ∼20. The lowest threshold, at 20 ms, corresponds roughly to the temporal resolution for flicker detection of the visual system ∼50 Hz. We stress the fact that although our stimuli do move, the task of observers was not to detect any characteristics of this motion such as its velocity or direction, but just its onset. (The same experiment could in principle be performed based, for example, on a change in stimulus wavelength. However, with a stimulus changing color, the figure–ground discrimination could be achieved on a single frame of the stimulus sequence—when part of the stimulus has changed color while the other part has not yet. Hence, such a stimulus would contain a possible confound. Our stimulus, on the other hand, does not allow such a discrimination based on a frozen single frame.) The relevant parameter underlying figure–ground segregation in this experiment is the point in time when motion occurs, an important difference from studies of motion perception in amblyopia. ^44–46  

It is important to note that the increase in thresholds cannot be attributed directly to the decreased spatial resolution for three reasons. First, patients with the same decrease in spatial resolution show widely differing temporal thresholds, and some patients with relatively high spatial acuity (such as 10/20 or 0.5) show strongly elevated temporal thresholds (such as 45 ms). Second high temporal delays between figure and ground will not as such allow patients with low spatial resolution to discriminate between the figure and the ground, since the spatial setting of the stimulus stays the same. Moreover, the spatial resolution required to detect the stimuli was more than sufficient, even for the amblyopic eyes of all patients, given the size of each of the rotators of ∼30 arcmin and the fact that sensitivity to detect motion is far finer than two-point acuity (i.e., it is a so-called hyperacuity ⁴⁷ ). Third, our control experiments with normal observers clearly demonstrate that decreasing visual resolution by blurring or low-pass filtering the stimuli does not as such increase thresholds for time-defined figure–ground segregation. 

A loss in contrast sensitivity for stimuli consisting of high spatial frequencies has been reported repeatedly, but this finding does not have direct consequences for our stimuli with a contrast far above threshold and containing a wide spectrum of spatial frequencies. Therefore, we conjecture that amblyopia not only decreases spatial resolution, but moreover diminishes the ability of patients to group together individual elements on the basis of their temporal common fate (i.e., motion onset) and to discriminate between different groups of elements. However, not all time-related tasks are impaired in amblyopic eyes: For high speeds of motion in spatiotemporal interpolation (which is the basis to perceive motion from the sequences of stationary images in films and TV), performance as originated through channels tuned to low spatial and high temporal frequencies is actually better, and thresholds for detecting an offset are lower in amblyopic eyes than in their normal partner eyes ⁴⁸ (see also Ref. 49). 

The fact that time-based figure–ground segregation is disturbed in amblyopia does not mean that the disturbance is on the neuronal level of figure–ground discrimination—of course, the disturbance could be present already on an earlier level of processing. Although previous studies testing temporal resolution on earlier levels of processing detected minute disturbances at best in amblyopic eyes, the exact nature of the underlying neuronal mechanisms remains to be investigated. In principle, these could be both purely temporal mechanisms or else spatiotemporal ones of the type described for example by Fahle and Poggio. ⁵⁰ In both cases, second order mechanisms ⁴⁴ may be deficient or signal-to-noise segregation may be adversely affected. ¹⁶  

To sum up, amblyopia not only means a distorted spatial representation of the visual world through the affected eye. Quite the contrary, the precision of processing and time-based grouping of neighboring elements obviously is greatly diminished in amblyopic eyes, resulting in a disturbance of spatiotemporal processing. Hence, amblyopia represents a complex disorder of visual processing in (early) visual cortices that not only concerns spatial but also temporal resolution, as is evident in feature grouping. 

Footnotes

 Supported by Grant DFG-FA-119/17-1 from the German Research Council.

Footnotes

 Disclosure: K. Spang, None; M. Fahle, None

Footnotes

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

The authors thank Dennis Trenner for writing the program code, Maren Prass for help with the experiments, and Cathleen Grimsen for discussions.