Influence of Background Complexity on Visual Sensitivity and Binocular Summation Using Patterns with and without Noise

Abstract

Purpose.: To investigate how background complexity influences visual sensitivity and binocular summation.

Methods.: Using two noise backgrounds (noise-sparse and noise-dense) and two corresponding noise-free backgrounds with the same luminance for each noise background, monocular and binocular thresholds were measured in six visually normal subjects (average age, 27.3 ± 1.1 years). The noise-sparse and noise-dense backgrounds respectively had 312 and 936 white-light dots projected on them—the same size white-light dots (0.431° of visual angle) as those that were used for the white-spot target in the threshold measurement. The target was tested at the fovea and at 3° intervals on the 45°, 135°, 225°, and 315° meridians. A total of 25 locations were tested.

Results.: The monocular threshold for the noise-dense background was higher than that for its corresponding noise-free background, with significant differences seen at 15° and 18° (P < 0.01). No significant differences in the binocular threshold were seen, either between the noise-dense and its corresponding backgrounds or between the noise-sparse and its corresponding backgrounds. The binocular summation ratios for both noise backgrounds were significantly higher than the ratios for the noise-free backgrounds, and the difference increased with eccentricity, with significances seen at 15° and 18° (P < 0.01).

Conclusions.: Only the monocular threshold increases with background complexity. The binocular summation increases with background complexity in the periphery. When the background becomes more complex and the monocular visual processing reaches its limit, binocular interaction functions efficiently.

Clinical evaluation of visual sensitivity is essential for diagnosis and determination of visual field (VF) loss progression in glaucoma and optic nerve diseases. In a standard clinical measurement, the two eyes are usually tested separately with a white-light target presented at various test locations on a dome-shaped, solid-white background. The luminosity discrimination threshold for each test location is determined by varying the intensity of the target luminance. The luminosity discrimination threshold is defined as the difference between the luminance levels of the background and target (i.e., the increment threshold). Because subtle VF defects may indicate the possibility of various ocular diseases such as glaucoma, detection of a subtle VF defect is essential to early diagnosis, effective follow-up, and good care for the patient's quality of vision (QOV). However, the clinical measurement of visual sensitivity may not represent the true visual function in daily life. In daily surroundings, the human vision usually functions binocularly and the background is much more complex than the white-solid background used in the clinical measurement. Therefore, the influences of background complexity and a binocular viewing condition should be carefully considered when evaluating visual sensitivity. 

Ever since Pirenne ¹ reported binocular summation as one of the binocular functions in 1943, different binocular summation ratios that express the ratio of binocular sensitivity to monocular sensitivity have been presented by many researchers for various stimulus factors. Factors such as target size, ^2,3 stimulus contrast, ^{4 –6} and the type of task ^7,8 can all affect the amount of binocular summation. In assessing binocular sensitivity, investigators have used either a model in which the results of the two separate monocular fields were used to predict binocular sensitivity or a method in which the higher sensitivity of each eye was selected at each test point. ^{9 –11} Since the binocular summation ratio has been shown to increase when the monocular function has reached its limit in a task that has become too difficult, ^8,12 binocular sensitivity results from binocular interaction and should not be the simple addition of two monocular sensitivities. Therefore, the clinical assessment of the two separate eyes may not be sufficient to reflect a patient's true QOV. 

To detect any decreased sensitivity, a solid background is used to remove any factors that might influence visual sensitivity in clinical measurement of the luminosity discrimination threshold. However, most tasks in daily life require not only target detection but also target discrimination from various surroundings, and yet, it is not clear whether and how visual sensitivity will be affected when a nonsolid background is used in the measurement of the luminosity discrimination threshold. That is, the relationship between background complexity and visual sensitivity has not been quantitatively evaluated. Elucidation of this relationship is helpful not only for assessing a patient's visual function in daily life but also for understanding the fundamental attribute of visual sensitivity. 

Engel ^13,14 used test objects of slightly different forms in an identical background of straight lines of random slant and location to detect the size of the effective VF. They concluded that the size of the effective VF becomes smaller as the form of the test object becomes more similar to the random straight lines in the background. In another study, a target embedded in distracting stimuli was used to investigate the relationship between visual sensitivity and target localization problems in the elderly. ^15,16 With a Gabor target, the detection thresholds decrease when binocular disparity separates the target from the mask. ^17,18 To et al. ¹⁹ reported that the perceived magnitude decreases when the target is surrounded by the same flankers rather than the different flankers in the peripheral area. Although these researchers did not use solid backgrounds and have investigated how target and/or background could influence the task of target discrimination, to our knowledge, no group has investigated either the relationship between background complexity and the visual sensitivity obtained under a binocular viewing condition or the influence of background on binocular summation. 

In the present study, we sought to investigate whether and how background complexity influences visual sensitivity and binocular summation. As a measure of background complexity, the two noise backgrounds used in this study had different dot densities. In addition, they incorporated random dots of the same size and shape as the target dots used in the visual sensitivity measurement. By assessing the increment threshold, the influence of background complexity on visual sensitivity and binocular summation was investigated. We further evaluated its effect on monocular and binocular conditions and between the central and peripheral vision within the central 30° VF. 

Materials and Methods

Subjects

Subjects were six female volunteers between 25 and 28 years of age without any systemic or ophthalmic disease that were likely to alter their visual function. We selected the subjects within a narrow age range to avoid any possible variation in visual sensitivity with age. All the subjects were experienced in psychophysical experiments. The ophthalmic inclusion criteria were as follows: best corrected visual acuity of 1.2 (−0.1 logMAR equivalent) or better, refractive error within ±3.00 D sphere and ±0.75 D astigmatism, normal stereopsis (with foveal stereopsis of 60 seconds of arc or better on the TNO stereo test), normal ocular alignment, and normal ocular motility. All experiments were performed in accordance with the Declaration of Helsinki for research involving human subjects. Informed consent was obtained from the subjects after explanation of the nature and possible consequences of the study. 

Apparatus and Stimulus

To project a VF background, a perimeter (Octopus 900; Haag-Streit International, Köniz, Switzerland) combined with a microprojector (KAGA Components, Tokyo, Japan) was used. The microprojector contained an LED with a luminous flux of 140 lm. 

In addition to the regular noise-free background, two noise backgrounds with random-dot patterns were used study: a sparse background (noise-sparse) with 312 white-light dots (0.2 dot/deg²) and a dense background (noise-dense) with 936 white-light dots (0.7 dot/deg²) projected on the background (Fig. 1). Both backgrounds used the same white-light dots with a dot size of 0.431° of visual angle, which was the same dot size used for the white-spot test target in the visual sensitivity measurement. To differentiate the background dots from the test target, we arranged the two random dot patterns in such a way that no background dots intersected with the 25 test locations. The minimum distance between the background dot and the target was 5.62 minutes of visual angle. In this experiment, we measured eight areas without dots on each of the noise backgrounds, and each area was measured twice. We then averaged these values and used it for the luminance level of the noise background measured. The luminance levels of the noise-sparse and -dense backgrounds were 42.95 asb (13.68 cd/m²) and 64.89 asb (20.67 cd/m²), respectively. For comparison, each noise background had a corresponding noise-free background with the same luminance level; four backgrounds were used in the study. 

Figure 1.

 

The two noise backgrounds.

Twenty-five test locations were used: at the fovea and at 3° intervals on the 45°, 135°, 225°, and 315° meridians in the central 30° VF (Fig. 2). A white-spot test target with a size of 0.431° of visual angle was projected on the cupola of the perimeter (Octopus 900; Haag-Streit) with one of the four backgrounds. Visual sensitivities were determined using a 4–2–1-dB bracketing staircase measurement procedure. The stimulus duration was 100 ms and the viewing distance was 30 cm. Fixation was checked with a small infrared video camera to monitor the pupil's position in each eye. The measurement was continued only when the subject fixated on the central point within the cupola. 

Figure 2.

 

Schematic representation of the target locations. The target was tested at the fovea and at 3° intervals on the 45°, 135°, 225°, and 315° meridians. A total of 25 locations were tested.

Measurement Procedure

Thresholds were measured for the right, left, and both eyes using the four backgrounds: the noise-sparse and its corresponding noise-free backgrounds with luminance of 42.95 asb (13.68 cd/m²) and the noise-dense and the corresponding noise-free backgrounds with luminance of 64.89 asb (20.67 cd/m²). Each background was tested twice, and the average was used for all the subsequent calculations for that session. The subject was instructed to press the button on perceiving the target while fixating on the central point within the cupola. No subjects required refractive corrections, because their refractive errors were within ±3.00 D sphere and ±0.75 D astigmatism. For the measurement of monocular threshold, the nontested eye was occluded with an opaque cover so that the subject could perceive only the background luminance in the cupola and not the target. Instead of using the average value, we used the lower threshold of the two eyes for the monocular measurement. Because the nasal retina of the measured eye corresponds to the temporal retina of the fellow eye under a binocular viewing condition, the differential threshold (particularly in the case of peripheral threshold) between the nasal and temporal retinas should be taken into account. In addition, we could overestimate binocular summation by using the average value of the two eyes. The orders of the eyes and backgrounds to be tested were determined randomly for each subject. The examination could be interrupted at any time during the test at the subject's request. 

Statistical Analysis

Differences in visual sensitivity between the tested eye(s) (monocular, binocular) and background and between eccentricity and background were analyzed by two-factor factorial ANOVA and the Bonferroni/Dunn test. The difference between the actual and predicted binocular visual sensitivities was analyzed by one-factor factorial ANOVA and the Bonferroni/Dunn test. Differences reaching P < 0.05 were statistically significant. 

Results

Comparison between Monocular and Binocular Thresholds for the Noise and Noise-Free Backgrounds

Differences between monocular and binocular thresholds for the noise and noise-free backgrounds were analyzed separately for the noise-sparse and -dense backgrounds by two-factor factorial ANOVA, with the tested eye(s) and background as factors. No significant interaction was found between these two factors (F _(4.35) = 0.98, P = 0.33 for noise-sparse and F _(4.35) = 2.72, P = 0.11 for noise-dense). Binocular thresholds were significantly lower than monocular thresholds for both noise-sparse and -dense backgrounds (F _(4.35) = 25.43, P < 0.01 for noise-sparse; F _(4.35) = 26.20, P < 0.01 for noise-dense; and P < 0.01 by the Bonferroni/Dunn test; Table 1). 

Table 1.

 

Monocular and Binocular Thresholds for the Noise and Noise-Free Backgrounds

	Noise-Sparse	Corresponding Noise-Free
Monocular threshold	4.43 ± 0.01**	4.41 ± 0.01**
Binocular threshold	4.40 ± 0.00	4.39 ± 0.00

	Noise-Dense	Corresponding Noise-Free
Monocular threshold	4.61 ± 0.02**	4.59 ± 0.00**
Binocular threshold	4.58 ± 0.01	4.58 ± 0.02

 Data are the mean asb × min² ± SD for the mean threshold of the six subjects. The threshold for the right eye or the threshold for the left eye, whichever was the lower, was used for the monocular threshold. Probabilities are for the differences between monocular and binocular thresholds (**P < 0.01).

Differences in Thresholds between the Noise and Noise-Free Backgrounds at the Seven Eccentricities

Threshold differences between the noise and noise-free backgrounds were separately evaluated for the monocular and binocular conditions (Fig. 3). In the monocular threshold, two-factor factorial ANOVA with background and eccentricity as factors indicated a significant interaction (F _(2.23) = 2.80, P = 0.02 for noise-sparse and F _(2.23) = 3.74, P < 0.01 for noise-dense). The noise-dense background had a significantly higher monocular threshold than its noise-free background at 15° and 18° (F _(3.98) =16.5, P < 0.01 and P < 0.01 by the Bonferroni/Dunn test; Fig. 3A). In the binocular threshold, two-factor factorial ANOVA with background and eccentricity as factors did not yield a significant interaction (F _(2.23) = 1.47, P = 0.20 for noise-sparse and F _(2.23) = 1.17, P = 0.33 for noise-dense). No significant differences in the binocular threshold were seen between the noise and noise-free backgrounds for both noise-sparse and noise-dense backgrounds (Fig. 3B). 

Figure 3.

 

(A) Differences in monocular threshold between the noise and noise-free backgrounds at each of the seven eccentricities. The monocular threshold for the noise-dense background was significantly higher than that for the corresponding noise-free background at 15° and 18° eccentricities (P < 0.01). (B) Differences in binocular threshold between the noise and noise-free backgrounds.

Difference between the Actual and the Predicted Binocular Visual Sensitivities for the Noise and Noise-Free Backgrounds

According to Nelson-Quigg et al., ¹⁰ binocular sensitivity can be predicted by the square root of the summed squares of the two monocular sensitivities. Using this formula, we calculated the predicted binocular sensitivity and evaluated the difference between the actual and predicted binocular visual sensitivities. The obtained sensitivity differences for the noise and noise-free backgrounds were compared. Significant differences were found among the four backgrounds (F _(2.69) = 6.23, P < 0.001 by one-factor ANOVA). The sensitivity differences between the actual and predicted values for the noise-dense background were significantly greater than the differences for its noise-free background (P < 0.01 by the Bonferroni/Dunn test; Fig. 4). However, the differences for the noise-sparse and its noise-free backgrounds were not significant. 

Figure 4.

 

Differences between the actual and predicted binocular visual sensitivities. The values (actual minus predicted) for the noise backgrounds were plotted against those for their corresponding noise-free backgrounds. The values for the noise-dense background were significantly greater than those for the noise-free background (P < 0.01).

Comparison of Binocular Summation Ratios among the Noise and Noise-Free Backgrounds

The binocular summation ratio was compared between the noise and noise-free backgrounds (Fig. 5). Two-factor factorial ANOVA with background and eccentricity as factors showed a reliable interaction (F _(2.23) = 2.87, P = 0.01 for noise-sparse and F _(2.25) = 2.91, P = 0.01 for noise-dense). The binocular summation ratios for both noise backgrounds were higher than the ratios for their corresponding noise-free backgrounds (F _(3.98) = 8.49, P < 0.01 for noise-sparse and F _(3.98) = 15.58, P < 0.01 for noise-dense). Significant differences were seen at 15° and 18° eccentricities (F _(2.23) = 2.87, P = 0.01 for noise-sparse and F _(2.23) = 7.36, P < 0.01 for noise-dense; P < 0.05 for noise-sparse and P < 0.01 for noise-dense by the Bonferroni/Dunn test; Fig. 5). The highest binocular summation ratio was seen at 18° with the noise-dense background. No significant differences in the binocular summation ratio were seen between the two noise-free backgrounds. 

Figure 5.

 

Differences in the binocular summation ratio between the noise and noise-free backgrounds. The ratio difference increased with eccentricity beyond 12° eccentricity for both noise backgrounds, with significance seen at 15° and 18° eccentricities (*P < 0.05; **P < 0.01). Error bars, ±SD (n = 6).

Discussion

This study clearly showed that as the background became more complicated, the difference in the monocular visual sensitivity between the noise and noise-free backgrounds became statistically significant, and the amount of binocular summation also increased in the peripheral area within the central 30°. However, the binocular visual sensitivity did not differ significantly between the noise and noise-free backgrounds. 

Many previous studies have reported that various factors influence binocular summation. The amount of binocular summation increases with low-contrast, ^{4 –6} decreasing stimulus size, ³ increasing eccentricity, ^3,6 a blur condition, ²⁰ and a recognition task rather than a detection task. ^7,8 Although these studies indicated how binocular summation is affected by the properties of the test target, the influence of background on binocular summation is still unknown. Targets that have the same properties but are presented on different backgrounds could affect visual sensitivity and binocular summation differently. Our results showed that as the background became more complicated (noise-free versus noise, and noise-sparse versus noise-dense), the amount of binocular summation increased but it did not differ significantly between the two noise-free backgrounds with different luminance levels adjusted to the corresponding noise backgrounds. We therefore considered that, like the properties of the test target, the background complexity could also be a factor that influenced binocular summation. Moreover, the background complexity rather than the luminance level might have a greater impact on the amount of binocular summation. 

Regarding the relationship of binocular summation with monocular and binocular sensitivities, Bearse and Freeman ⁵ reported that stimulus contrast level and exposure duration are the factors in binocular summation. With short durations and low contrasts, monocular thresholds are higher than binocular thresholds. In addition, monocular discrimination thresholds decrease to a greater degree than binocular thresholds with higher contrast and longer durations. Our study of reaction time at suprathreshold ¹² and the present study of background complexity showed similar results, indicating a higher correlation between binocular summation and monocular sensitivity. The difference in monocular sensitivity caused by background complexity could be a factor in the amount of binocular summation. Furthermore, the investigated factors in the above-mentioned studies all appeared to contribute to binocular summation more when monocular perception had become more difficult. 

The monocular visual sensitivity was not influenced by the background complexity within the central 10° VF, whereas the sensitivity decreased as the complexity increased beyond 10° in the periphery. Only the monocular visual sensitivity for the noise-dense background significantly decreased in the peripheral area within the central 30°. We considered two possible causes for this observation. The first was the perceived contrast difference between the background dots and the target dot that varied in different areas. In VF testing, the fovea is known to have the highest sensitivity, and sensitivity decreases with increasing eccentricity. ²¹ This result can be attributed to the distribution of ganglion cells and the fact that the cortical magnification factor is greater in the central area than in the periphery. ^{22 –24} The required luminance for the target dot to be detected would be lower in the central area and much higher in the periphery. In this study, the random dots on each noise background had the same luminance. Thus, the perceived contrast difference between the background dots and the detectable target dot was greater in the central area and smaller in the peripheral area within the central 30°. From this finding, we deduced that the perceived contrast difference could influence the performance of a detection task. The greater the perceived contrast difference was, the easier the task became, and vice versa. 

The second possible cause was the resolution difference between the peripheral and central areas on the noise background. We have previously reported the difference between the detection and resolution sensitivities using parallel-line targets. ⁸ The resolution sensitivity is lower in the periphery than in the fovea as the width of the parallel lines becomes narrower. Besides, the resolution sensitivity is lower than the detection sensitivity, and this difference increases in the periphery. In the present study, the average distances between any two background dots were approximately 2.03° of visual angle in the noise-sparse background and 0.99° of visual angle in the noise-dense background. The equivalent visual acuities for these visual angles were approximately 0.01 decimal visual acuity (2.1 logMAR) for the noise-sparse background and 0.02 decimal visual acuity (1.8 logMAR) for the noise-dense background. The minimum distance between the background dot and target dot was approximately 5.62 minutes of visual angle, which was equivalent to 0.2 decimal visual acuity or 0.7 logMAR. Regarding the relation between visual acuity and eccentricity, previous studies reported that the human eyes have visual acuity of approximately 0.2 to 0.3 at 8° and 10° eccentricities and 0.1 beyond 10° in the periphery. ^{25 –27} Therefore, the subjects in this study should be able to discriminate between the background dots and the target dot, as well as between any two background dots within the central 10°. The discrimination however became more difficult beyond 10° in the periphery, and this was particularly true of the noise-dense background because it required higher resolution than the noise-sparse background in the peripheral area. Under the measurement conditions of this study, only the monocular sensitivities significantly differed between the noise-dense and corresponding noise-free backgrounds at 15° and 18°. This suggested that the binocular interaction prevented a decrease in the binocular visual sensitivity, even in the peripheral area where the visual sensitivity was low for each eye. 

In conclusion, we have demonstrated that as the background became more complex, only the monocular visual sensitivity decreased, with a significant decrease in the peripheral area within the central 30°. We further showed that the binocular summation increased significantly when the eyes detected the target projected to the peripheral area on the noise background. This was concordant with our previous result indicating that binocular interaction functions efficiently when the monocular visual processing has reached its limit. Although our present results may be more relevant to the basic understanding of retinal sensitivity than practical applications in patient's daily life, they could serve as a base for the quantitative assessment of a patient's daily visual function. In the future, we intend to further investigate whether and how the influence of background complexity on visual sensitivity varies in eyes with various visual dysfunctions and to develop a method of quantitatively assessing a patient's daily visual function. 

Footnotes

 Disclosure: A. Wakayama, None; C. Matsumoto, None; K. Ohmure, None; M. Inase, None; Y. Shimomura, None

The authors thank Reiyo Tahara for editorial support.