A Residual Deficit for Global Motion Processing after Acuity Recovery in Deprivation Amblyopia

Abstract

purpose. In a case of bilateral deprivation amblyopia due to congenital cataracts, the global motion sensitivity for stimuli that were well within the passband of the amblyopic visual system were assessed.

methods. A stochastic global motion stimulus was used, comprising spatially narrow elements with varied spatial frequency, density, contrast, and area distribution. To determine threshold, a two-alternate, forced-choice direction discrimination task was used.

results. There was a selective deficit for global motion processing that was not due to the visibility of the stimuli and was nonselective for spatial scale. The eye with the more complete recovery (acuity 20/20) from pattern deprivation in childhood exhibited the more severe global motion deficit.

discussion. The results suggest a primary extrastriate deficit in the dorsal pathway, possibly involving the middle temporal (MT) and the medial superior temporal (MST) cortical areas, that is unrelated to the acuity deficit thought to be in area V1. A similar deficit has recently been shown in strabismic amblyopia.

One of the visual processes most studied is that of motion processing. It has a cortical locus and involves a whole array of visual areas extending from area V1, the primary cortex to the medial superior temporal (MST) area in the dorsal pathway leading to the parietal lobe. In V1, cells have spatially localized receptive fields and respond to motion often in a preferred direction. ¹ In higher areas such as the middle temporal (MT), where receptive fields are larger, these V1 local motion signals are thought to be integrated. ² At even higher parts of the dorsal pathway—for example, in the dorsal part of MST—cells respond to radial and concentric forms of global motion. ³ This processing hierarchy extending from local motion in V1 to global integration in MT and to optic flow in MST is apparent both in the physiology and in lesion studies. For example, monkeys ^{4 5 6 7} and cats ^{8 9} whose MT has been bilaterally lesioned, show selective deficits in global motion tasks in which the integration of local motion signals is required. Similarly, humans with lesions to this same area (MT/MST) exhibit selective deficits in global motion processing. The motion-blind patient studied by Baker et al. ¹⁰ is a prime example. In this case, if all local motion signals are going in the same direction, which is the case when integration is not required, motion direction could be correctly judged. However, the addition of just 10% of noise elements (non–signal direction) was sufficient to disrupt performance. Even more surprising, these noise elements need not be moving. Stationary noise elements for a normal observer are trivial to segregate; not so for the patient whose MT/MST is defective. Both the physiology and the psychophysics support the role of MT in global processes involving integration and possibly segregation. 

The cortical deficit in amblyopia is not well understood. There is good evidence from recent functional imaging that, in the case of strabismic amblyopia, a large part of the visual cortex is affected. ¹¹ Recent psychophysical studies ¹² have shown that global tasks are affected in a way that makes it unlikely that the problem is restricted to V1. The nature of the deficit suggests that regions of the extrastriate cortex may be primarily affected, including both the dorsal, motion-processing and the ventral, form-processing extrastriate pathways. ¹³ In form–deprivation amblyopia a similar loss of global form and motion have been reported that is more severe in bilateral than in unilateral cases. ^{14 15}  

We report on a person who had reduced vision in both eyes in childhood due to congenital cataracts that were removed at ages 5 and 6. As an adult, the patient had the acuity in one eye return to normal, although the other eye had a residual acuity loss. We used a global motion task similar to those previously used in MT-lesioned monkeys, the motion-blind patient, and strabismic amblyopes in an effort to see whether the extrastriate cortex that is thought to subserve this task is compromised. Our results showed a deficit in global motion processing that can best be explained by a deficiency at the level of the extrastriate cortex. It was worse in the eye with acuity that had recovered to normal. 

Methods

The experiment was performed by computer (Macintosh G4, 1250 MHz; Apple, Cupertino, CA), using software routines from the Video Toolbox. Stimuli were displayed on a monitor (model P225f; ViewSonic, Walnut, CA), with a screen measuring 35 × 26 cm. The mean luminance of the display was set to 50 cd/m² and was made a linear function of the digital control signals from the graphics card with a light meter (model S370; Graseby Optronics, Orlando, FL). Pseudo 12-bit resolution was achieved using the Video Attenuator. 

Subjects

This study adhered to the tenets of the Declaration of Helsinki. Participating were two normal subjects (RW, TC) and an amblyope (LS). RW was experienced in psychophysics experiments, whereas TC and LS were naive observers. The clinical characteristics of LS are described in Table 1 . 

Stimuli

A random Gabor kinematogram was used in a direction-discrimination task. The Gabors had a limited lifetime of 2, meaning that a dot was in a location for three frames (each frame being 16 ms), then in a second location for another three frames, only to be replaced by a new element of randomly chosen location. If a Gabor exited the stimulus area, it was deleted at the boundary to preserve the stimulus area. At the same time, a new Gabor was randomly generated elsewhere, to maintain element density. For the normal subjects, element density was constant across all conditions: 10 Gabors/deg². In the amblyope, density was lower—six Gabors/deg²—to aid performance by reducing the probability of making false correspondences. This topic will be addressed further in the discussion. 

All elements were Gabor elements and were in one of two populations: signal or noise elements. Signal elements either moved upward or downward, in a coherent direction. Noise elements moved in any random direction, presenting incoherent direction. A constant number of elements was maintained during each trial by adjusting the ratio of signal-to-noise elements. To prevent abrupt spatiotemporal transients, the stimulus area was contrast modulated over time by a cosine function at onset and offset. 

Three conditions of spatial frequency were explored (Table 2) : very low frequency (VLF), medium frequency (MF), and very high frequency (VHF). To analyze the effect of spatial selectivity, it was necessary to maintain a narrow, constant bandwidth. In Fourier terms, the range of spatial frequencies that a Fourier analysis of a waveform would yield was of the same magnitude throughout the experiment. Therefore, bandwidth was kept constant, even as element size varied with each condition (Fig. 1) . 

For each condition, three stimulus area sizes were used. Their dimensions, in pixels, were 225 × 225, 148 × 148, and 98 × 98, subtending 4.08, 2.68, and 1.78 deg² of visual angle, respectively (Fig. 2) . 

Experimental Procedure

The ability of subjects to detect these vertically moving dots on a screen was measured by establishing motion coherence thresholds in a spatial, two-alternative, forced-choice (2AFC) task. The normal observers (RW, TC) and the amblyope (LS) viewed it monocularly, one eye at a time. Viewing took place in a dark room, at a distance of 171 cm from the monitor. The monitor was set at a resolution of 640 × 480 pixels at a 60-Hz refresh rate. There were 50 frames presented per trial, running for a total of 500 ms. Observers were told to fixate a crosshair in the center of the screen and used two keyboard keys to nominate the direction in which coherent motion was seen. Observers were given auditory feedback for both correct and incorrect answers. 

Variables

The two main variables in the experiment were the size of the stimulus area, to observe the effects of summation, and the spatial frequency, to observe which channels were most active. 

As well, to sample the more relevant region of each observer’s psychometric function, three sources of noise were varied: element density, element lifetime, and signal-to-noise ratio (SNR). 

Element density, as mentioned earlier, was lower for the amblyope. Initially, the SNR was set to 100%, and element lifetime was increased until the observer began to see the stimulus better by learning what to look for. Then, the element lifetime was lowered as much as possible, resulting in the smallest possible value of 2 in normal subjects, and ranging from 5 to 20 in the amblyope. Subsequently, with element density and element lifetime thus fixed, the SNR was varied to obtain the best sample of each observer’s psychometric function and determine thresholds. 

Psychophysics

Using the method of constant stimuli, the proportion of signal elements (the SNR) was varied to establish psychometric functions from which detection thresholds were estimated (Table 3) . The data were fitted with a cumulative Gaussian function by using a least-squares algorithm with the fit being weighted by the standard deviations of each empiric data point. A Monte Carlo simulation was performed by computer (MatLab; The MathWorks, Natick, MA) to estimate the threshold and its variability using the Psignifit software package. ^{16 17} Detection thresholds were defined to be the stimulus value corresponding to 75% correct performance on the psychometric function. 

Results

The two important aspects of this stimulus that give one confidence that the visual system is accomplishing the task by using a global, integrative mechanism rather than simply by detecting any one of the local motion directions is the addition of noise and the use of local motions of short lifetimes. The fact that some of the elements are not moving in the signal direction (the noise elements) means that a local motion-detecting strategy would be unreliable, especially when the fraction of noise elements is significant. A normal threshold for this task would be for signal elements to represent ∼ 20% of all the elements. Under these conditions, it is difficult to imagine, on statistical grounds, that the task could be accomplished by anything other than a mechanism that integrates over the sparse signal elements. Combining signal in noise with short lifetimes was intended to ensure further that to detect a globally coherent pattern, observers had to integrate over many dot motions that were individually uninformative. ^{4 18} Thus, even if observers were able to identify individual signal elements, the motion of any one of these, owing to its short lifetime, would be too weak to support performance. Many such signal elements would have to be integrated. 

Under these stringent conditions that ensure that performance is being driven by a global, integrative operation, we could not measure a reliable threshold in either of LS’s eyes. LS’s amblyopic eye could not see the VHF condition due to the patient’s contrast sensitivity deficit, as the contrast sensitivity was biased to lower sensitivity and lower spatial frequencies. Therefore, we attempted to collect data at only the VLF and MF conditions. Although these stimulus conditions fell well within the visibility range of the amblyopic eye, to be able to measure threshold performance at all, we had to increase the element lifetime to 5 for the amblyopic eye and to 20 for the better (i.e., visual acuity), fellow eye. Even this failed to allow us to get reliable data from LS’s fellow eye, even at signal strengths of 100%. The threshold of the amblyopic eye was measurable, but the signal level had to be ∼40%, double the norm, even under the most favorable stimulus conditions (i.e., at MF 225). Examples of these data are presented in Figure 3for the two normal subjects and in Figure 4for both eyes of LS. 

An important feature of such a global task is that, as the stimulus area increases, threshold performance improves because signal levels increase at a greater rate than noise levels (Downing et al. IOVS 1989;30:ARVO Abstract 72). Noise levels do not increase linearly with signal levels, because noise largely superimposes destructively, whereas signal superimposes constructively. The fact that area summation occurs is further evidence that the task is a global one and not simply due to detection of an isolated local motion signal. Because we could measure a threshold for the amblyopic eye we wondered whether we could also show area summation. These results are shown in Figure 5for the two normal subjects and the amblyopic eye of LS. The downward slope in the normal subjects’ graph (RW, TC) shows that summation occurs because threshold decreases with increasing stimulus area. In the amblyope (LS), the slope is almost zero, indicating that the amblyope did not exhibit area summation for this task. 

Discussion

There were two key findings. First, under conditions in which one can ensure that performance on this stochastic motion task is determined by a global integrative mechanism, we could not measure a threshold in either the amblyopic or fellow eye of this deprivation amblyope. Second, performance on this motion task was much worse when the stimulus was viewed by the less amblyopic eye (20/20 acuity). 

Because we had ensured that the stimulus elements were well within the window of visibility for each eye, this deficiency was not due to visibility. That the less amblyopic eye (20/20 acuity) displayed the more severe motion deficit further reinforces this conclusion. Pattern deprivation in early life leads to reduced correctable acuity, a finding reinforced by numerous previous studies on animals ¹⁹ and humans. ^{20 21} It also leads to motion deficits, ¹⁵ and the present study adds further support to the conclusion that global motion is specifically affected. The only conditions under which performance on this task was measurable were ones that invalidate any conclusion based on global integrative function. The use of long lifetimes and minimal noise favors detection by low-level local motion detectors. Unlike those in previous studies, our stimuli were not spatially broadband. We used spatially narrowband Gabors that enabled us to investigate further the global deficit at a number of different spatial scales and to ensure that, at each scale, the contrast of the stimuli was always suprathreshold for the amblyopic eye. Our results show that the deficit is, to a first approximation, independent of spatial scale (i.e., similar results at medium and low spatial frequencies). The other novel feature of the present study is that we show that area summation for global motion detection is absent in the deprived visual system, further suggesting detection by a local rather than a global mechanism. It seems that global sensitivity was so poor in this deprivation amblyope that it did not make any contribution to the task we used. Thus, there is a selective deficit for global motion in this bilateral deprivation amblyope. Of interest, one eye, although it was deprived in early life, exhibited normal visual acuity, and yet the motion deficit was greater in this eye. Not only does this suggest that the motion deficit is unrelated to the acuity loss but it also suggests that acuity and motion-processing follow different courses of development. 

Deficits in global motion-processing have been reported in several clinical cases in which cortical disease is involved. A motion-blind patient who had a bilateral lesion to area MT/MST exhibited a similar inability to detect such a stimulus when it contained even a small degree of noise. ²² Similarly, monkeys with lesions to area MT/MST in the dorsal extrastriate pathway also exhibited performance limited to very high signal coherence. ⁴ The close similarity of the present results to those in these two cases suggests a processing deficit of MT/MST in the extrastriate cortex of LS. 

More recently, it has been shown that strabismic amblyopes exhibit global motion deficits that are not explicable on the basis of the visibility deficit thought to reside in V1. ¹² Such a global deficit is not restricted to motion and has been shown also to involve spatial processing ¹³ ; therefore, both dorsal and ventral pathways are similarly affected. Form deprivation amblyopia shows similar losses ^{14 15} in which the motion deficit is greater when the less affected eye is stimulated. This suggests that the present case, in which the loss in the less affected eye is absolute, is not an isolated curiosity. 

So far, we have considered deficient global processing of motion solely in terms of an inability to integrate signal elements, and we present evidence for this being the case for our deprivation amblyope LS. However, the segregation of signal from noise is also an important, though often overlooked, part of this task. It is not in the interest of best performance for the visual system to integrate blindly all the local motion signals contained within our stimulus. Ideally, signal should first be segregated from noise and then integrated. That we had to reduce the element density for LS to perform the task at all with the amblyopic eye is consistent with deficient segregation of signal from noise. Our more recent results suggest that segregation processes are abnormal when the more amblyopic eye of this deprivation amblyope is used and that it mimics the segregation loss that we have also revealed in the other, more common form of functional amblyopia. ²³  

An interesting false-correspondence effect was observed in each of LS’s eyes. During a practice period, when the stimulus extent was 225 × 225 pixels, 50% to 100% SNR, and an element lifetime of five frames, LS consistently reported clearly seeing motion in the direction opposite the one presented. There are two possible explanations. 

The first is element spacing and receptor size. As a signal element moves to its second location across a given distance, LS does not detect that jump as a fluid motion because she may not have receptors for that size in area V1, perhaps due to spatial undersampling at the level of the local detector. 

The second is relative motion. As a set of signal elements moves in one direction, the noise elements appear to be “left behind,” and they contribute to a percept of relative motion effect in the opposite direction, although they are incoherent. Because the noise elements often outnumber the signal elements, they may be more conspicuous and may be mistaken for signal. 

 Supported by Canadian Institutes of Health Research (CIHR) Grants MT108-18 and MOP533346.

 Submitted for publication February 22, 2005; revised April 8, 2005; accepted April 14, 2005.

 Disclosure: T. Constantinescu, None; L. Schmidt, None; R. Watson, None; R.F. Hess, None

 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.

 

Table 1.

 

LS’s Clinical Details

Deprivation amblyope

Bilateral congenital cataracts, the left one thicker than the right

Right cataract removed at age 5, with lens capsule cleaning

Left cataract a year later at age 6, with lens capsule cleaning

Patched completely at age 7 for 2 months, patched partially for a subsequent period

No atropine treatment

Astigmatism in both eyes, although it is not as pronounced in the right eye

Clear optics with no visible retinal abnormalities

Better visual acuity and contrast sensitivity in right eye (20/20 vs. 20/120)

Table 2.

 

Experimental Details for Each Condition

	VLF	MF	VHF
Spatial frequency (cyc/deg)	2.99	5.30	12.0
Element speed (deg/s)	1.68	0.94	0.42
Michelson contrast (%)	50	50	67

 There was a greater contrast for VHF, to improve contrast detection in that region of the contrast sensitivity function.

Figure 1.

 

Sample stimuli from each of the conditions (from the left: VLF, MF, and VHF). Constant bandwidth was maintained across all conditions.

Figure 2.

 

Sample stimuli for normal subjects at 9.964 dots/deg² (left) and the amblyope at 5.978 dots/deg² (right).

Table 3.

 

The SNR Range Necessary to Establish Psychometric Functions

	SNR Range (%)
RW	5–50
TC	10–70
LS (amblyopic eye)	5–100

 The data for the fellow eye were not reliable, and so they are not included.

Figure 3.

 

Psychometric functions showing performance as a function of the signal–noise level in normal subjects during the direction-discrimination task.

Figure 4.

 

Psychometric functions showing performance as a function of the signal–noise level in the amblyope during the direction-discrimination task.

Figure 5.

 

Summation curves in which direction detection is plotted against stimulus area. The slope is given by the exponent.