Cerebral Cortex Advance Access originally published online on June 12, 2006
Cerebral Cortex 2007 17(5):1117-1128; doi:10.1093/cercor/bhl021
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Mechanisms of Sensitivity Loss due to Visual Cortex Lesions in Humans and Macaques
Center for Visual Science and Department of Ophthalmology, University of Rochester Medical Center, Rochester, NY 14642, USA
Address correspondence to William H. Merigan, Department of Ophthalmology, Box 314, University of Rochester Medical Center, Rochester, NY 14642, USA. Email: billm{at}cvs.rochester.edu.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionSupplementary MaterialAppendixReferences |
|---|
This study represents the first use of noise masking and signal detection theory to examine mechanisms of visual loss after lesions of visual cortex. Noise-masked contrast thresholds were increased in 2 macaques and 2 humans at lesion-affected, compared with control, regions of their visual fields. Experiments suggested by the organization of visual cortex examined possible mechanisms of the visual loss. Two experiments tested the hypothesis that damage to feedback connections might eliminate the benefit of comparing test stimuli with remembered representations but neither could account for the sensitivity loss. The third experiment found that extrastriate lesions did increase the trial-to-trial variability of sensory decisions, suggesting this as one mechanism of sensitivity loss. In addition to clarifying mechanisms of lesion-induced contrast sensitivity loss, this study also showed that elevated contrast thresholds, that are subtle in the absence of external noise, became dramatic when measured with masking noise.
Key Words: cortical lesions • human • macaque • mechanisms of visual loss • perceptual decision • signal detection
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSupplementary MaterialAppendixReferences |
|---|
Signal detection theory (Green and Swets 1966
) has provided important insights into mechanisms by which optical or neural defects can limit the performance of the visual system. Masking noise, added to the visual stimulus, can help identify the mechanism of a change in sensitivity (Pelli and Farell 1999). This approach has been used to examine improvements in visual sensitivity during early development of the visual system (Banks and Bennett 1988) or during perceptual learning (Dosher and Lu 1998), as well as decreases of visual sensitivity due to aging (Bennett and others 1999), optical defocus (Pardhan and others 1993), or visual disease (Kersten and others 1988). The signal detection approach mathematically separates several theoretical sources of altered efficiency in visual processing, some of which can be plausibly identified with components of the visual system, such as feedforward or feedback projections.
This study examined the modest reduction in contrast sensitivity that follows extrastriate lesions. Although contrast sensitivity is typically identified with earlier stages of the cortical pathway (Miller and others 1980; Merigan and others 1993), complete lesions of such early areas as striate cortex or the lateral geniculate nucleus produce such devastating visual loss (Miller and others 1980) that signal detection analysis is not possible. Likewise, although extrastriate cortex is most identified with complex shape perception (Merigan 1996; Riesenhuber and Poggio 1999), visual losses due to extrastriate lesions, such as prosopagnosia (loss of face recognition) (Damasio and others 1982), achromatopsia (loss of color vision) (Zeki 1990), and disrupted motion (Baker and others 1991; Pasternak and Merigan 1994), and shape (Gallant and others 2000; Merigan 2000) perception are so profound that signal detection analysis is not possible.
In the experiments described here, we studied impaired visual sensitivity in 2 human and 2 macaque subjects, all of whom had extrastriate cortical lesions that affected a portion of their visual field. Both visual sensitivity and complex vision were measured in the affected portion of the visual field of each subject and compared with comparable measures in an unaffected control region of the visual field. Sensitivity loss was studied to examine possible neural mechanisms giving rise to the loss, whereas complex visual perception was studied only to verify that it occurred in the same region of the visual field as the contrast sensitivity loss.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSupplementary MaterialAppendixReferences |
|---|
General Methods
Subjects
Two macaque monkeys received ibotenic acid lesions of cortical area V4, corresponding to one of the lower quadrants of the visual field (Merigan and Pham 1998). Magnetic resonance imaging was used to map the cortical lesions (Merigan and Pham 1998), and histological analysis confirmed that they corresponded to the lower field representation of the left or right visual field in the 2 monkeys. Prior to the studies described here, the 2 monkeys were tested on a variety of visual discriminations, ranging from simple orientation-based acuity and contrast sensitivity to 3-dimensional (D) shape discriminations (Merigan and Pham 1998). These measures showed that visual deficits were confined to the damaged quadrant of the visual field. All testing described in this report involved a comparison of affected and control regions of the visual field. Experimental protocols were approved by the Animal Use and Care Committee of the University of Rochester and conformed to National Institutes of Health guidelines.
The 2 human subjects had suffered unilateral strokes in or near the fusiform gyrus, resulting in visual dysfunction in the upper left quadrant of their visual field. Informed consent was obtained from both participants after the risks, hazards, and potential discomfort associated with the procedures were explained. Visual loss in one of these subjects, on a variety of visual tests different from those used in the present study, has been reported previously (Merigan and others 1997). Figure 1 shows the approximate location of the lesions and the approximate visual field distribution of impaired vision in the 2 human and 2 macaque subjects. Experiment 5 in the present study measured an index of complex visual perception (illusory contours) for both human subjects and macaques and found it to be severely disrupted in the same portion of the visual fields. Prior to the start of this study, the 2 human subjects had extensive experience with controlled fixation testing of near-peripheral vision (
20 degree eccentricity).
|
Fixation Control
All visual testing was conducted while the subject fixated a small spot near the center of the display. Stimuli were presented to the affected quadrant of the visual field, or as a control, to the corresponding location across the vertical meridian. To permit monitoring of fixation locus, monkeys were fitted with scleral search coils during a surgical procedure prior to their lesions (Judge and others 1980). This system permitted monitoring of fixation within ±0.5 degrees. The testing software presented the fixation target after a short intertrial interval, and the test stimulus was presented after the monkey had fixated this target for 800 ms. Trials were interrupted immediately if the monkey did not maintain fixation within a ±0.75 degree window.
Human subjects viewed the display with their head supported by a chin rest. Eye position was monitored by a remote video system (ISCAN Model 416) using infrared illumination to avoid interference with stimulus visibility. During the experiments described here, we manually monitored fixation with the video system and could reliably detect eye movements of as little as 1 degree. Data were analyzed only for those sessions in which the subject maintained fixation. All testing of human subjects was at an eccentricity of 20 degrees to insure that residual eye movements would not affect performance. Subjects were continually reminded to fixate carefully, and if fixation was unreliable, the entire staircase was repeated.
Display
For all orientation discrimination tasks (Experiments 1–3), monkey and human subjects faced a 17'' display of 5 cd/m2 luminance at a distance of 42 cm for humans and 84 cm for monkeys. For the contrast increment task (Experiment 4), humans sat 57 cm away from the display. The displays were equipped with video attenuators (Pelli and Zhang 1991) to provide precise control of contrast and were gamma corrected by software (Brainard 1997). During testing, monkey subjects had pupil diameters of 3.25–3.5 mm and humans 3.4–3.6 mm.
Procedure
For all subjects, orientation discriminations (Experiments 1, 2, 4, and 5) involved a 2 alternative forced-choice between stimuli of horizontal or vertical orientation. The contrast increment discrimination (Experiment 3) was tested with a 2-interval forced-choice, and subjects chose the interval containing the grating of higher contrast. All subjects pressed one of 2 buttons at the end of each trial to indicate their choices. If correct, macaques received a drop of fruit juice, and then progressed to the next trial after a brief intertrial interval. Incorrect trials were followed by a 3-s time-out tone. A tone following correct responses provided feedback for one of the human subjects (A), but no feedback was provided for the other subject (B) because this subject adopted a win-stay, lose-switch strategy when provided with feedback. Threshold measurements in these studies were determined by staircases with 3:1 up–down ratios (Experiments 1, 2, and 4) or 2:1 up–down ratios (Experiment 3). Single thresholds were determined in daily sessions of approximately 200 trials in monkeys, and multiple thresholds, involving approximately 100 trials each, were determined in each 1-h test session in human subjects. Thresholds were calculated by fitting maximum likelihood Weibull functions to the data of individual staircases, then taking the 75% correct point from the fitted curve.
Individual Experiments
1: Noise-Masked Contrast Sensitivity
Macaques. Examples of the stimuli are shown in Figure 2. Each stimulus measured 2 by 2 degrees and contained a Gabor patch (cosinusoidal grating of 3 cycles/degree, windowed by a Gaussian envelope of 0.6 degree standard deviation [SD]). The orientation of these stimuli was masked by 2D spatial noise of narrow bandwidth. The noise was created by filtering 128 by 128 pixel fields of random Gaussian-distributed gray levels with a 0.2 octave rectangular band-pass filter centered on 3 cycles/degree. Contrast thresholds for the Gabor stimuli were determined at 4 contrasts of the filtered masking noise, with SDs of 0, 4.5, 9, and 18 gray levels out of 256. In units of spectral density (noise power per unit bandwidth), these were 0 x 10–3, 0.73 x 10–3, 1.46 x 10–3, and 2.92 x 10–3 degree2. The stimuli were presented at an eccentricity of 5.6 degrees with a slow onset (300 ms raised cosine) and remained present until a choice response was made.
|
Humans. The stimuli were identical to those used with macaques, except that they were scaled up to 6 by 6 degrees (windowed by a Gaussian of 1.8 degree SD) for presentation at 20 degree eccentricity. These stimuli were presented with the same slowed onset as the stimuli for monkeys. Contrast thresholds were also measured at 4 levels of noise contrast, with SDs of 2.2, 4.5, 9, and 18 gray levels. Spectral densities were 1.1 x 10–3, 2.2 x 10–3, 4.4 x 10–3, and 8.8 x 10–3 degree2.
2: Spatial Frequency Bandwidth of Masking
Example stimuli are shown below Figure 4. These stimuli were similar to those used in the first experiment, except that the 2D masking noise had a center frequency 1, 1.4, 3, or 9 times that of the grating stimulus to be masked. The orientation discrimination was identical to that described above. The contrast of the masking noise was adjusted to make it equally visible (constant multiple of contrast threshold for a single observer) at each of the 4 mask spatial frequencies. The unadjusted noise had a SD of 18 gray levels, leading to adjusted noise spectral densities of 1.5 x 10–2, 0.8 x 10–2, 0.3 x 10–2, and 2.5 x 10–2 degree2.
The masked contrast thresholds were plotted as a function of the center spatial frequency of the mask and fit using a heteroscedastic nonlinear regression model. The expected value of the threshold contrast in the intact and lesion fields was represented as a function of the masking frequency (x) by the function
 |
 |
The solution for this equation is x = b. In order to compare the breadth of the response for the intact and lesion fields, we tested whether the maximum likelihood estimates of b obtained in the 2 different fields were statistically significant. This comparison was carried out for each patient and for each monkey using the likelihood ratio test (Lehmann 1986). Tests with P value smaller than 0.05 were considered significant. The model fittings were performed using Matlab (The MathWorks, Inc. Natick, MA 01760).
3: Added Stimulus Pedestal to Measure Contrast Increment Thresholds
Contrast increment thresholds were measured in the 2 human observers with a 2-interval, temporal forced-choice procedure. High visibility, 0.25 contrast pedestals were used for both stimuli to insure that the location, spatial frequency, and orientation of both stimuli were unmistakable. Thus, the properties of each test stimulus were clear to the observer, although the contrast increment between the 2 stimuli was at threshold. The stimuli were 6 by 6 degrees vertical, Gaussian windowed (SD = 1.8 degrees) patches of gratings of 1 cycles/degree, presented at an eccentricity of 20 degrees. The 2D masking noise consisted of spatially unfiltered fields of 24 by 24 samples, whose gray levels were drawn from Gaussian white noise distributions with SDs of 2.5 and 10 gray levels (out of 256) generated on each trial and added to the test stimuli. These SDs produced spectral densities of 2.1 (0.1 pedestal contrast) and 8.4 x 10–4 degree2 (0.25 pedestal contrast), values found in pilot studies to give moderate threshold elevation.
The 2 stimuli to be compared were presented for 467 ms each, separated by a blank period of 250 ms, and stimulus presentations always followed a 200-ms period of fixation. The patient indicated which grating appeared to have the higher contrast by pressing a computer key (left for the first interval and right for the second interval). A 2-up/1-down staircase controlled the difference in contrast between the 2 intervals.
Data Analysis
Resampling of Thresholds (Experiments 1, 2, and 3)
Because our human subjects found long test sessions difficult, thresholds were measured with a small number of trials, typically less than 100. To obtain estimates of the variability inherent in these brief test sessions, we performed a resampling procedure (Manly 1991) by pooling all the data for each subject into a single probability correct for each contrast level and from these values generated many psychometric functions. We fit the psychometric functions to the simulated data with the same maximum likelihood Weibull function that we used for the psychophysical data. Error bars in the graphs reflect the SD of the resampled thresholds.
4: Consistency of Performance on 2 Passes through a Fixed Stimulus Sequence
This procedure estimated the variability of the decision process by measuring the consistency of response to repeated presentations of a sequence of noise-masked orientation stimuli (same discrimination as Experiment 1). These repeated measures were made in both control and lesion-affected regions of the visual field for the 2 human subjects and one macaque to determine if the variability of psychophysical decisions was greater in the region corresponding to the lesion. Each test set consisted of 200 stimuli, 50 at each of 4 contrast levels, that were chosen for each location to produce overall performance in the 70–90% correct range. For all subjects, control visual field masks varied in contrast energy from 0.5 x 10–2 to 3 x 10–2 degree2; in the lesion-affected field they varied from 5.5 x 10–2 to 2.1 x 10–2 degree2. The noise mask in each of the 200 stimuli was unique. The monkey was tested with a spectral density of 1.5 x 10–3 degree2. Humans were tested with the same noise level in the control visual field, but in the lesion-affected visual field, in order to raise performance, the noise level was reduced to 1.1 x 10–3 degree2. The monkey was tested with each list of stimuli 5 times, once per day. Human subjects were tested with each list twice. When asked after testing, human subjects were unaware that they had been making discriminations with a repeating list of stimuli.
Calculation of consistency.
We compared trial-by-trial responses for pairs of passes through the stimulus set at control and lesion-affected regions of the visual field. We fitted the data to equation (1) (described in the Appendix [Supplementary Material]), which represents the following model. Each stimulus is cross-correlated with a vertical and a horizontal template (identical to the stimuli, but without masking noise), and the difference between these cross-correlation values is the decision variable of the noiseless ideal observer. The decision variable divided by the SD of the masking noise yields the d' (visual sensitivity) of the ideal observer for masking noise of that contrast. The sensitivity of the ideal observer is much greater than that of real observers (Banks and others 1987), and this difference is given by the scaling parameter alpha. Because the masking noise on pairs of trials is identical, the ideal observer will always make the same decision on each paired trial. The consistency of the real observer is described by the parameter k, which reflects the ratio of the internal (processing) noise to the external (masking) noise, with larger values of k indicating more internal noise, and therefore less agreement. We determined a single value of k across 4 contrasts of masking noise for the lesion-affected and control visual fields of each of the 3 tested subjects. However, we determined 4 values of alpha (one for each contrast) in each visual field.
To estimate the expected variability in our determinations of consistency (k), we performed a resampling procedure (Manly 1991), in which we determined values of k for shuffled data equated to the observed consistency values, and took the SD of the simulated k values as an estimate of the true variability in our measurements of consistency (vertical error bars in Fig. 6).
5: Discriminating the Orientation of Illusory Contours
In this experiment, the monkey or human subject was presented with one of the stimuli illustrated in Figure 7 and required to report the vertical or horizontal orientation of the contour across the middle of the target. These stimuli had been found in previous unpublished studies to be indiscriminable after V4 lesions in monkeys (for similar stimuli, see Merigan 2000) but were easily discriminated by patients in their normal visual field. Each stimulus consisted of 8 concentric hemicircles that were displaced by one half of their separation along a radius either horizontally or vertically through the figure. Contour thickness was 1% of the entire figure, and circle separation was 7%. The 3 outer circles were stepped down in contrast by 50% from the neighboring circle to minimize the visibility of the outer circle. Circular figures to the left in each axis of Figure 7 had real contours (value of 1) across the middle, but the proportion of real contour decreased 2-fold in circular figures to the right, until the rightmost stimulus contained only an illusory contour, marked by misalignment of the hemicircles. A 3:1 up/down staircase adjusted the line termination thickness, increasing thickness after errors and decreasing after correct responses. The radius of the circular figures was 2 degrees at 5.6 degree eccentricity for monkeys and 6 degrees at 20 degree eccentricity for humans.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSupplementary MaterialAppendixReferences |
|---|
Experiment 1: Noise-Masked Contrast Thresholds
Figure 3 shows noise-masked contrast thresholds as a function of the contrast energy of the 2D masking noise (stimuli for this experiment are shown in Fig. 2). Contrast thresholds in the control field of all subjects (open symbols) increased with the contrast of the masking noise. However, in the lesion-affected field (filled symbols), thresholds were markedly elevated relative to the control field for all subjects. Lesion field thresholds were approximately proportional to control field thresholds as indicated on the log axes of Figure 3 by the vertical displacement of the threshold curves. The extent of threshold elevation ranged from about 4-fold in 3 subjects to approximately 10-fold in Patient A. No differences in response latency between control and lesion fields were seen in any subjects.
|
No Evidence of Increased Additive Noise
Increased noise in signal detection analysis can be either additive or multiplicative (independent of or increasing with noise contrast). On the log axes of Figure 3, an increase in additive noise caused by the extrastriate lesion would correspond to a reduced slope of the lesion field threshold data, relative to that for the control location, but such a shift is not evident in any of the plots. Statistical analysis of additive internal noise calculated from intercepts of the data on linear axes (Table 1) shows no increase in additive noise in any subject.
|
All 4 Subjects Showed Decreased Calculation Efficiency
In Figure 3, the vertical shift of the threshold contrast versus masking contrast functions indicates a decrease in calculation efficiency for all 4 subjects (Bennett and others 1999). Minimal stimulus noise had a small effect on thresholds, whereas large masking noise greatly increased thresholds.
Likely mechanisms by which damage to extrastriate cortex could give rise to this change in calculation efficiency (Lu and Dosher 1999) were evaluated in the next 3 experiments. Three possibilities were considered: a consistent mismatch between templates; an increase in the uncertainty about which template to use for the matching; and an increase in signal-dependent or multiplicative internal noise. These possibilities are diagrammed in Figure 8.
Experiment 2: Consistent Mismatch between Internal Templates and Test Stimuli. Spatial Frequency Bandwidth of Noise Masking
Figure 4 shows the elevation of contrast thresholds by 2D noise as a function of the center spatial frequency of the narrow-band noise masks. For all 4 subjects, contrast thresholds in the lesion-affected portion of the visual field were higher than those in the control field, and this was true for both masked and unmasked thresholds measured in this experiment. However, the spatial frequency bandwidth of the masking was not significantly increased for any subject except patient A for whom an increase in bandwidth was significant at the 0.001 level (Table 1). This result indicates that increased susceptibility to masking by dissimilar stimuli (at least on the spatial frequency dimension) cannot be an essential basis of increased contrast thresholds after cortical lesions.
|
Experiment 3: Increased Intrinsic Uncertainty in the Lesioned Field. Noise-Masked Contrast Increment Thresholds
Contrast increment thresholds masked by low and high contrasts of masking noise (0.001 and 0.04 corresponding to spectral densities of 2.1 x 10–4 and 8.4 x 10–3, respectively) are shown in Figure 5 for the lesion-affected and control visual fields of the 2 human subjects. Even in the presence of a visible template, contrast thresholds were substantially elevated at the lesion location for both subjects, making it unlikely that inability to use internal templates was a major contributor to the sensitivity loss.
|
Experiment 4: Consistency of Orientation Judgments: 2-Pass Measurement
Figure 6 shows the consistency (percent agreement) of a series of stimulus orientation judgments on 2 passes through an identical stimulus set as a function of the average percent correct for the 2 human subjects and one macaque. The lines are best fits of equation (1) (Appendix). This model scales the percent correct performance of the ideal observer at each contrast by a parameter alpha to match the performance of the human observer. The percent agreement is determined by a second parameter, k, that depends only on the ratio of the internal noise in the visual system to the external noise in the stimulus. The k ratios shown for each subject in Figure 6 were significantly changed (Student's t-test [P < 0.05]). This indicates that the internal multiplicative noise in the lesioned field was higher than in the intact field for all 3 subjects.
|
Experiment 5: Illusory Contour Perception
Finally, we measured the perception of complex shape at the visual field locations where subjects had reduced visual sensitivity, using a stimulus difference that is easily discriminated by normal human subjects but that cannot be discriminated despite substantial parametric variation after V4 lesions in monkeys (for a discussion of related stimuli, see Merigan 2000). Figure 7 demonstrates that all 4 subjects also showed a deficit in complex visual perception (illusory contour orientation) in the lesion-affected region of the visual field compared with the control region. Each graph shows percent correct performance as a function of the level of difficulty of the horizontal–vertical discriminations illustrated under the abscissa, which ranged from a highly visible real contour, indicated by 1, to an entirely illusory contour indicated by 0. All subjects performed all levels of difficulty near perfectly in the control portion of the visual field (open symbols), but in the region corresponding to the extrastriate lesion (filled symbols), all subjects performed near chance levels with illusory contour stimuli.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSupplementary MaterialAppendixReferences |
|---|
Elevations of noise-masked contrast thresholds, as well as disrupted complex perception, were found in visual field locations corresponding to lesions in 2 macaques and 2 humans who had retinotopic lesions of extrastriate visual cortex. The degree of threshold elevation at the lesion site increased with the contrast of the masking noise, making the contrast sensitivity loss much more evident during masking with high-contrast noise. This decreased calculation efficiency is a type of processing deficit consistent with several possible mechanisms. Likely mechanisms of this multiplicative sensitivity loss within the signal detection model were explored in the subsequent experiments. We first attempted to examine the internal representation of the stimuli by measuring the spatial frequency bandwidth of the masking function and found that it was not broadened in most subjects by the cortical lesion, suggesting that frequency mismatch with some internal template was not the source of the visual loss. The exact form of such a high-level internal template is unknown; for the sake of simplicity in our model we used an exact copy of the stimulus (also known as a "matched filter"). We further evaluated a possible mismatch of stimulus and internal template by providing a highly visible external template consisting of a contrast pedestal, but the magnitude of sensitivity loss remained severe, indicating that this experiment did not support template uncertainty as a mechanism of the visual loss. However, in a separate experiment, we found that the consistency of near-threshold decisions was decreased by the lesions, suggesting increased variability in the decision process as a central component of the decreased vision. We did not examine the basis for the retinotopically impaired complex perception of the subjects in the lesion-affected region of the visual field, but it is likely that similar mechanisms could be responsible.
The Affected Region of the Visual Field Corresponded to the Anatomical Locus of the Lesions
The visual field locus of the loss in the 2 macaques (illustrated in method Fig. 1) was contralateral to ibotenic acid lesions that had been placed in the lower field representation of the retinotopic cortical area V4 (Merigan and Pham 1998). Physiological mapping of area V4 in the macaque (Gatass and others 1988) has shown that the representations of the vertical and horizontal meridia of the visual field lie close to the lunate and superior temporal sulci, respectively, whereas the foveal representation is inferior, near the end of the inferior occipital sulcus. Our reconstruction of these lesions, using psychophysical and magnetic resonance measures (Merigan and Pham 1998), confirmed that lesions were confined to the representation of one quadrant of the visual field within V4 and that visual loss was limited to a single quadrant extending to a few degrees from the vertical and horizontal meridia.
We also established, using a variety of perceptual measures (W.H. Merigan, unpublished data), that the visual loss in the human subjects was confined to one quadrant of the visual field, suggesting that the lesions either were within retinotopic cortical areas, as in the monkeys, or interrupted fiber pathways subserving a quadrant of the visual field (Plant 1991). The lesion in subject A was small (1.3 cm diameter) and located near the fusiform and lingual gyri, anterior to the highly retinotopic areas that have been mapped in the human brain with functional magnetic resonance imaging (e.g., Sereno and others 1995). This location is consistent with the lesion location reported in other human subjects who show quadrant visual loss (Plant 1991; Gallant and others 2000). This region was also affected in patient B, although this patient's lesion was broader and extended more posteriorly.
No Increase in Additive Noise
Additive noise (Fig. 8) comprises all sources of variability of visual processing that are independent of stimulus contrast, producing visual loss that is independent of the contrast of added noise. Additive noise in the intact visual system is caused by such variables as optical blur, the limited quantum catch of photoreceptors, and the spatial pooling of photoreceptor signals by bipolar cells (Banks and Bennett 1988). Increased additive noise can also be caused by cataracts (Pardhan and others 1993) or decreased retinal illuminance (Nagaraja 1964) but not by amblyopia or optic neuritis (Kersten and others 1988). The present results show that it is also not caused by extrastriate cortical lesions.
|
Decreased Calculation Efficiency
The signature of decreased calculation efficiency is a deficit in visual processing that is more pronounced the higher the contrast of the added noise. Previous research has shown that such losses can be caused by aging (Bennett and others 1999) or by amblyopia (Kersten and others 1988). Within the model of template matching used in this study (Fig. 8), the decreased calculation efficiency we found could result from a disruption of template matching or from decisional aspects of the visual processing. There are 3 components of template matching that may be vulnerable to extrastriate dysfunction: the development of internal templates, the trial-by-trial choice of optimal templates, and the fidelity with which the template matches the test stimuli.
Development and Use of Templates
Internal representations of expected visual stimuli can be built up inductively by repeated exposure to test stimuli during the course of testing, a capability the method of single stimuli (McKee and others 1986) takes advantage of in measuring thresholds. Psychophysical observers often show a progressive improvement in sensitivity during extended practice (Ahissar and Hochstein 1996; Furmanski and Engel 2000) that is highly specific to the training stimuli, and this improvement may reflect learning of the location, size, direction of motion, color, and other characteristics of the stimuli. Features of the masking noise itself can also be learned if the noise remains constant from trial to trial, and the observer can actively discount the mask to better detect or discriminate stimuli (Beard and Ahumada 1999). Thus, the learned component of visual perception may be thought of as the tuning of an internal representation or template for the test stimuli (Gold and others 1999). Alternatively, internal templates may be developed without extended training if the set of possible stimuli are displayed or described to the subject (Burgess 1985; Gold and others 1999). In either case, the derivation and use of an internal stimulus template involves learning and memory, processes typically associated with more central stages of the visual pathways (Brunel 2003). The finding that learning improvements can be specific to the trained portion of the visual field (Dosher and Lu 1999) parallels the localized retinotopic losses found in the present study, suggesting involvement of visual neurons prior to inferotemporal cortex, such as those in extrastriate cortex that have retinotopic receptive fields.
Trial-by-Trial Selection of Optimal Template
Template-matching models (Fig. 8) envision observers having access to multiple internal visual templates (often termed visual channels) but monitoring only the most likely channels in order to minimize impaired visual sensitivity due to the spontaneous noise in the additional visual channels being monitored (Pelli 1985). Uncertainty about which stimulus will be presented forces the observer to monitor multiple channels, thereby reducing visual sensitivity. Indeed, visual performance is degraded when any stimulus features are made unpredictable, such as direction of motion (Ball and Sekuler 1981), spatial frequency (Davis and Graham 1981), or stimulus location (Cohn and Lasley 1974; Burgess and Ghandeharian 1984). We tested the hypothesis that extrastriate lesions produce similar uncertainty by using contrast increment discrimination (Legge and others 1987) in Experiment 3 to provide the subject a clearly visible representation of the stimulus and thus minimize stimulus uncertainty. We reasoned that an observer with damage to extrastriate cortex might behave like an uncertain observer unable to use visual memory of previous test trials to predict which stimuli were likely to be presented. However, to the extent that provision of an external template reduces uncertainty, our findings do not support this hypothesis.
Reduced Precision of Template Matching
A third possible mechanism for the present finding is reduced visual sensitivity when the internal templates to which test stimuli are compared deviate in some way from the actual stimuli. For example, if the templates used by observers in the present study differed in color, spatial scale, spatial frequency, etc., template matching would suffer, due to an increase in stimulus noise being allowed through the mismatched filter. Experiment 2 tested one possible mismatch, spatial frequency, and found that the template was not mismatched to the stimuli because the spatial frequency tuning of masking was not broadened. Subjects were still able to discount the masking frequencies. Of course, test stimuli have other potentially mismatched dimensions, and some deviations might have been found if we had examined other stimulus features. One possible basis of increased masking bandwidth is internal templates that are slightly less extensive in space than the actual stimuli. This could be caused by impaired spatial integration resulting from damage to extrastriate cortex, which has larger receptive fields than earlier cortical areas that project to it. Spatial frequency and orientation bandwidths of a Gabor patch are inversely proportional to spatial extent (Graham 1989), hence a decrease in the size of an internal template could result in increased masking bandwidth. Such an account seems particularly probable given the finding of spatial pooling by subunits of like orientation preferences within the receptive fields of V4 neurons (Pollen and others 2002).
The width of spatial frequency channels measured psychophysically in adaptation and masking studies (Blakemore and Campbell 1969; Legge and Foley 1980) closely match those of individual V1 neurons recorded under similar stimulus conditions (Bradley and others 1987). This match may suggest that the narrowness of spatial frequency masking bandwidths reflects the width of internal templates (neural channels), as long as an observer can select a single channel to monitor. On the other hand, psychophysically measured bandwidths for spatial frequency and other stimulus dimensions might be broader if stimulus uncertainty or extrastriate lesions forced observers to monitor multiple channels.
Consistency of Perceptual Decisions
Multiplicative noise is a theoretical construct not found in all models of visual perception. It is a necessary component of our model (Fig. 8) because of the results of Experiment 4, namely, the decrease in consistency of subjects' answers on 2 passes through an identical stimulus set. At the same level of accuracy, subjects were simply less consistent in their decisions when stimuli were presented in the lesioned visual field. Models that rely purely on template mismatch to explain mistakes would not produce this result. According to such models, observers would make mistakes due to the stimulus noise, but they would make the "same" mistakes during every pass through the same stimulus set.
The 2-pass method (Burgess and Colborne 1988) used in this study has been used previously to examine the consistency of perceptual decisions as a possible mechanism of changing sensitivity in a noise-masked discrimination. Extended practice on a noise-masked face identification task resulted in improved performance (Gold and others 1999), but the consistency of choices was not changed during the period of learning. However, the findings of the present study suggest that decisional consistency may be decreased by aging and by amblyopia that, like extrastriate lesions, reduce the efficiency of visual performance (Kersten and others 1988; Bennett and others 1999).
The inconsistency of contrast sensitivity we observed in the lesioned field was probably not due to unsteady or inconsistent fixation. During all conditions, the subjects fixated with central vision, and we saw no differences in fixation stability when testing was in the control or lesion visual fields. All subjects fixated very close to the fixation spot, the macaques not using the full ±0.75 degree fixation window, and human subjects rarely deviating from narrow fixation. Furthermore, in past studies (e.g., Merigan 1996), we have not found the extrastriate lesioned visual field of monkeys to show more heterogeneous contrast sensitivity than the control field. Thus, trial-to-trial variability in fixation locus could not account for the observed inconsistency of decisions.
Physiological studies of neural activity related to perceptual decisions in macaques suggest it is dominated by high-level cortical areas, involving broad networks of neurons across multiple brain regions. Cortical areas potentially involved include prefrontal (Kim and Shadlen 1999), inferotemporal (Dudkin and others 1995), and parietal cortices (Platt and Glimcher 1999). Correlates of perceptual decisions have also been demonstrated in extrastriate neurons in the dorsal stream (Newsome and others 1989; Mazurek and others 2003) but have not yet been studied in ventral extrastriate cortex. Reduced consistency of perceptual decisions after extrastriate lesions in the present study could represent damage to extrastriate neurons involved in decisions or simply interruption of the connection through extrastriate cortex of the peripheral visual system to the high-level neural networks engaged in perceptual decisions. The present findings indicate that perceptual decisions can be degraded over only a portion of the visual field, confirming the finding that perceptual decisions show a retinotopic layout.
Implications for Using Noise-Masked Stimuli to Test Cortical Integrity
Previous research has found that visual noise can exacerbate the loss caused by lesions of extrastriate visual cortex. In a study in macaques (Rudolph 1997; Rudolph and Pasternak 1999), motion direction noise severely exacerbated the effects of middle temporal area (MT) lesions on direction discriminations, whereas 2D spatial noise potentiated the effects of V4 lesions on orientation discriminations. The basis of the increased noise masking could be broadened bandwidth of masking, which, as in the present study, might make previously irrelevant noise interfere with discriminations. This mechanism might account for some otherwise inexplicable effects, such as the observation in a patient who had a lesion of the dorsal cortical pathway (Baker and others 1991), that direction discrimination, tested with dot stimuli moving at 10 degrees/s, was disrupted by the presence of static noise. Static noise does not normally disrupt the perception of rapid motion because the mechanisms responsible do not overlap in spatiotemporal frequency sensitivity.
Our findings suggest that the use of noise-masked stimuli may be particularly valuable for detecting visual loss in patients with lesions of extrastriate visual cortex. In the present study, elevations of contrast thresholds by extrastriate lesions were small in the absence of external noise but obvious when noise was added to the stimuli. This property of added visual noise, if general to different tasks and modalities, could make even basic measures of visual sensitivity useful for testing the integrity of high-level sensory cortical areas (Pelli and Farell 1999).
In addition to the impaired visual sensitivity observed in Experiments 1 through 3 of the present study, extrastriate lesions also severely disrupt complex vision (Experiment 5 and Merigan 2000). Such loss could also be examined with a signal detection/noise analysis by setting the visual channels (templates) equal to the objects or patterns being discriminated (Tjan and others 1995; Gold and others 1999). The complex stimulus selectivity of extrastriate neurons (Kobatake and Tanaka 1994; Gallant and others 1996; Hinkle and Connor 2002) may provide a template for such complex discriminations. Thus, in addition to clarifying the reduced visibility of simple grating stimuli in the present study, future signal detection analysis of extrastriate lesions may also illuminate the severely disrupted perception of complex stimuli.
|
Supplementary Material |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSupplementary MaterialAppendixReferences |
|---|
Supplementary material can be found at: http://www.cercor.oxfordjournals.org/.
|
Appendix |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSupplementary MaterialAppendixReferences |
|---|
Measurement of consistency (k) from response probability in Experiment 4 (In Supplementary Material).
|
Acknowledgments |
|---|
This work was supported in part by National Institutes of Health grant EY 08898 and an unrestricted grant from Research to Prevent Blindness. We thank Walter Makous and David Knill for comments. Conflict of Interest: None declared.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSupplementary MaterialAppendixReferences |
|---|
Ahissar M and Hochstein S. (1996) Learning pop-out detection—specificities to stimulus characteristics. Vision Res 36:3487–3500.[CrossRef][Web of Science][Medline]
Baker C, Hess RF, Zihl J. (1991) Residual motion perception in a "motion-blind" patient, assessed with limited-lifetime random dot stimuli. J Neurosci 11:454–461.[Abstract]
Ball K and Sekuler R. (1981) Cues reduce direction uncertainty and enhance motion detection. Percept Psychophys 30:119–128.[Web of Science][Medline]
Banks MS and Bennett PJ. (1988) Optical and photoreceptor immaturities limit the spatial and chromatic vision of human neonates. J Opt Soc Am A Opt Image Sci 5:2059–2079.[Web of Science][Medline]
Banks MS, Geisler WS, Bennett PJ. (1987) The physical limits of grating visibility. Vision Res 27:1915–1924.[CrossRef][Web of Science][Medline]
Beard BL and Jr Ahumada AJ. (1999) Detection in fixed and random noise in foveal and parafoveal vision explained by template learning. J Opt Soc Am A Opt Image Sci Vision 16:755–763.
Bennett PJ, Sekuler AB, Ozin L. (1999) Effects of aging on calculation efficiency and equivalent noise. J Opt Soc Am A Opt Image Sci Vision 16:654–668.
Blakemore C and Campbell FW. (1969) Adaptation to spatial stimuli. J Physiol 200:11P–13P.[Medline]
Bradley A, Skottun BC, Ohzawa I, Sclar G, Freeman RD. (1987) Visual orientation and spatial frequency discrimination: a comparison of single neurons and behavior. J Neurophysiol 57:755–772.
Brainard DH. (1997) The psychophysics toolbox. Spat Vision 10:433–436.[Web of Science][Medline]
Brunel N. (2003) Dynamics and plasticity of stimulus-selective persistent activity in cortical network models. Cereb Cortex 13:1151–1161.
Burgess A. (1985) Visual signal detection. III. On Bayesian use of prior knowledge and cross correlation. J Opt Soc Am A Opt Image Sci 2:1498–1507.[Web of Science][Medline]
Burgess AE and Colborne B. (1988) Visual signal detection. IV. Observer inconsistency. J Opt Soc Am A Opt Image Sci 5:617–627.[Web of Science][Medline]
Burgess AE and Ghandeharian H. (1984) Visual signal detection. II. Signal-location identification. J Opt Soc Am A Opt Image Sci 1:906–910.[Web of Science][Medline]
Cohn TE and Lasley DJ. (1974) Detectability of a luminance increment: effect of spatial uncertainty. J Opt Soc Am 64:1715–1719.[Medline]
Damasio AR, Damasio H, Van Hoesen GW. (1982) Prosopagnosia: anatomic basis and behavioral mechanisms. Neurology 32:331–341.
Davis ET and Graham N. (1981) Spatial frequency uncertainty effects in the detection of sinusoidal gratings. Vision Res 21:705–712.[CrossRef][Web of Science][Medline]
Dosher BA and Lu ZL. (1998) Perceptual learning reflects external noise filtering and internal noise reduction through channel reweighting. Proc Natl Acad Sci USA 95:13988–13993.
Dosher BA and Lu ZL. (1999) Mechanisms of perceptual learning. Vision Res 39:3197–3221.[CrossRef][Web of Science][Medline]
Dudkin KN, Kruchinin VK, Chueva IV. (1995) Neurophysiologic correlates of the decision-making processes in the cerebral cortex of monkeys during visual recognition. Neurosci Behav Physiol 25:348–356.[CrossRef][Medline]
Furmanski CS and Engel SA. (2000) Perceptual learning in object recognition: object specificity and size invariance. Vision Res 40:473–484.[CrossRef][Web of Science][Medline]
Gallant JL, Connor CE, Rakshit S, Lewis JW, Van Essen DC. (1996) Neural responses to polar, hyperbolic, and Cartesian gratings in area V4 of the macaque monkey. J Neurophysiol 76:2718–2739.
Gallant JL, Shoup RE, Mazer JA. (2000) A human extrastriate area functionally homologous to macaque V4. Neuron 27:227–235.[CrossRef][Web of Science][Medline]
Gatass R, Sousa APB, Gross CG. (1988) Visuotopic organization and extent of V3 and V4 of the macaque. J Neurosci 8:1831–1845.[Abstract]
Gold J, Bennett PJ, Sekuler AB. (1999) Signal but not noise changes with perceptual learning. Nature 402:176–178.[CrossRef][Medline]
Graham N. (1989) Visual pattern analyzers(New York: Oxford University Press, Inc).
Green DM and Swets JA. (1966) Signal detection theory and psychophysics(Wiley, New York).
Hinkle DA and Connor CE. (2002) Three-dimensional orientation tuning in macaque area V4. Nat Neurosci 5:665–670.[CrossRef][Web of Science][Medline]
Judge SJ, Richmond BJ, Chu FC. (1980) Implantation of magnetic search coils for measurement of eye position: an improved method. Vision Res 20:535–538.[CrossRef][Web of Science][Medline]
Kersten D, Hess RF, Plant GT. (1988) Assessing contrast sensitivity behind cloudy media. Clin Vision Sci 2:143–158.
Kim JN and Shadlen MN. (1999) Neural correlates of a decision in the dorsolateral prefrontal cortex of the macaque [see comments]. Nat Neurosci 2:176–185.[CrossRef][Web of Science][Medline]
Kobatake E and Tanaka K. (1994) Neuronal selectivities to complex object features in the ventral visual pathway of the macaque cerebral cortex. J Neurophysiol 71:856–867.
Legge GE and Foley JM. (1980) Contrast masking in human vision. J Opt Soc Am 70:1458–1471.[Medline]
Legge GE, Kersten D, Burgess AE. (1987) Contrast discrimination in noise. J Opt Soc Am A Opt Image Sci 4:391–404.[Web of Science][Medline]
Lehmann EL. (1986) Testing statistical hypotheses 2nd edn. (Springer Verlag, New York).
Lu ZL and Dosher BA. (1999) Characterizing human perceptual inefficiencies with equivalent internal noise. J Opt Soc Am A Opt Image Sci Vision 16:764–778.
Manly BFJ. (1991) Randomization and Monte Carlo methods in biology(Chapman & Hall, New York).
Mazurek ME, Roitman JD, Ditterich J, Shadlen MN. (2003) A role for neural integrators in perceptual decision making. Cereb Cortex 13:1257–1269.
McKee SP, Silverman GH, Nakayama K. (1986) Precise velocity discrimination despite random variations in temporal frequency and contrast. Vision Res 26:609–619.[CrossRef][Web of Science][Medline]
Merigan WH. (1996) Basic visual capacities and shape discrimination after lesions of extrastriate area V4 in macaques. Vis Neurosci 13:51–60.[Web of Science][Medline]
Merigan WH. (2000) Cortical area V4 is critical for certain texture discriminations, but this effect is not dependent on attention. Vis Neurosci 17:949–958.[CrossRef][Web of Science][Medline]
Merigan WH, Freeman A, Meyers S. (1997) Parallel processing streams in human visual cortex. Neuroreport 8:3985–3991.[Web of Science][Medline]
Merigan WH, Nealey TA, Maunsell JH. (1993) Visual effects of lesions of cortical area V2 in macaques. J Neurosci 13:3180–3191.[Abstract]
Merigan WH and Pham HA. (1998) V4 lesions in macaques affect both single- and multiple-viewpoint shape discriminations. Vis Neurosci 15:359–367.[CrossRef][Web of Science][Medline]
Miller M, Pasik P, Pasik T. (1980) Extrageniculostriate vision in the monkey. VII. Contrast sensitivity functions. J Neurophysiol 43:1510–1526.
Nagaraja NS. (1964) Effect of luminance noise on contrast thresholds. J Opt Soc Am 54:950–955.
Newsome WT, Britten KH, Movshon JA. (1989) Neuronal correlates of a perceptual decision. Nature 341:52–54.[CrossRef][Medline]
Pardhan S, Gilchrist J, Beh GK. (1993) Contrast detection in noise: a new method for assessing the visual function in cataract. Optom Vision Sci 70:914–922.[Web of Science][Medline]
Pasternak T and Merigan WH. (1994) Motion perception following lesions of the superior temporal sulcus in the monkey. Cereb Cortex 4:247–259.
Pelli DG. (1985) Uncertainty explains many aspects of visual contrast detection and discrimination. J Opt Soc Am A Opt Image Sci 2:1508–1532.[Web of Science][Medline]
Pelli DG and Farell B. (1999) Why use noise? J Opt Soc Am A Opt Image Sci Vision 16:647–653.
Pelli DG and Zhang L. (1991) Accurate control of contrast on microcomputer displays. Vision Res 31:1337–1350.[CrossRef][Web of Science][Medline]
Plant GT. (1991) Disorders of colour vision in diseases of the nervous system. In Foster DH (Ed.). Inherited and acquired color vision deficiencies(CRC Press, Boca Raton, FL) pp. 173–198.
Platt ML and Glimcher PW. (1999) Neural correlates of decision variables in parietal cortex. Nature 400:233–238.[CrossRef][Medline]
Pollen DA, Przybyszewski AW, Rubin MA, Foote W. (2002) Spatial receptive field organization of macaque V4 neurons. Cereb Cortex 12:601–616.
Riesenhuber M and Poggio T. (1999) Hierarchical models of object recognition in cortex. Nat Neurosci 2:1019–1025.[CrossRef][Web of Science][Medline]
Rudolph K and Pasternak T. (1999) Transient and permanent deficits in motion perception after lesions of cortical areas MT and MST in the macaque monkey. Cereb Cortex 9:90–100.
Rudolph KK. (1997) Motion and form perception following lesions of areas MT/MST and V4 in the macaque monkey(Rochester, NY: University of Rochester).
Sereno MI, Dale AM, Reppas JB, Kwong KK, Belliveau JW, Brady TJ, Rosen BR, Tootell RB. (1995) Borders of multiple visual areas in humans revealed by functional magnetic resonance imaging [see comments]. Science 268:889–893.
Tjan BS, Braje WL, Legge GE, Kersten D. (1995) Human efficiency for recognizing 3-D objects in luminance noise. Vision Res 35:3053–3069.[CrossRef][Web of Science][Medline]
Zeki S. (1990) A century of cerebral achromatopsia. Brain 113:1721–1777.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||