British Journal of Ophthalmology, 1987, 71, 58-65
Spatial contrast sensitivity and the diagnosis of amblyopia
A C W VOLKERS,' K H HAGEMANS,2 G J VAN DER WILDT,' AND P I M SCHMITZ' From the 'Department ofBiomedical Physics and Technology, Erasmus University, PO Box 1738, NL-3000 DR Rotterdam; 2Health Centre, De Linie 2, Capelle a. d. IJssel; and the 'Institute ofBiostatistics, Erasmus University, Rotterdam, The Netherlands
SUMMARY The relationship was studied between spatial contrast sensitivity function and type of pathology of the visual system. Two characteristics were found to be typical for amblyopia: (1) there is a discrepancy between the high-frequency cut-off ('grating acuity') and the Snellen acuity; (2) the contrast sensitivity is strongly dependent on the width of the stimulus. In optic nerve degeneration a decrease in contrast sensitivity is found at low spatial frequencies. Decreased contrast sensitivity at high spatial frequencies is atypical but occurs in those disorders associated with decreased Snellen visual acuity. From these findings a relationship is defined which enables spatial contrast sensitivity tests to be used to differentiate between amblyopia and optic nerve degeneration.
The ophthalmologist is faced with the task of diag- contrast or, in other words, how well coarser struc- nosing the cause of any visual impediment. After tures can be seen at low contrast.' optimal correction and examination by slit-lamp or The study of contrast sensitivity function as it is ophthalmoscope there remain some patients in related to several types of pathology is not new. whom the diagnosis is not clear. Their visual loss may Various workers have reported on contrast sensi- be due to optic nerve degeneration or to amblyopia tivity in amblyopia''9 and in optic nerve disorders.45 ('lazy eye'). It would be helpful if a simple method The CSF has been described in optic nerve lesions'0-'2 were available to distinguish between these two and in cataract. 13 Most of these publications deal with possibilities. Determination of spatial contrast sensi- rather small numbers of cases, and almost all concern tivity function (CSF) might be such a relatively our understanding of specific disorders of the visual simple method. system, but there are no reports on the evaluation of The Snellen optotype chart is a measure of the CSF as a diagnostic aid in neuro-ophthalmological minimum angle of resolution at high contrast. In CSF practice. the maximum discernible spatial frequency can also The aim of this study was threefold: first, to be interpreted as a measure of the resolving power: determine to what extent the shape of the CSF is the 'grating acuity' can be calculated from this related to the particular type of pathology; secondly, frequency. Although the stimuli with which the visual to evaluate the effectiveness of spatial contrast sensi- acuity and the grating acuity are determined are quite tivity tests for the diagnosis of amblyopia, especially different (optotypes and line patterns respectively), with regard to the differentiation between amblyopia it is interesting to compare the visual acuity and the and optic nerve degeneration with decreased visual grating acuity because both values are measures of acuity; and thirdly to compare the visual function in the spatial resolution at high contrast. The CSF optic nerve degeneration combined with normal provides more information than the Snellen acuity, visual acuity with visual function in a group of normal while it also estimates the spatial resolution at low eyes. Correspondence to A C W Volkers, Erasmus Universiteit Rotter- In a pilot study by Hagemans and Van der Wildt' in dam, Postbus 1738, 3000 DR Rotterdam, The Netherlands. a small group of patients it was found that the 58 Spatialcontrastsensitivity and the diagnosis ofamblyopia 59 dependence of contrast sensitivity on the width of the The CSF was determined for spatial frequencies stimulus was much greater for amblyopes than for ranging from 0-1 to 25 6 cycles per degree (c/deg) in normal subjects or patients with other pathology. steps of a factor 2. Up to 0-4 c/deg the viewing distance was 50 cm, resulting in a stimulus width of Subjects and methods 320; from 0-8 to 25-6 c/deg the viewing distance was 200 cm, which gave a width of 8°. For lower fre- The subjects who took part in this study were chosen quencies the stimulus had to be wider, because if from among the outpatients of an ophthalmological there are fewer than 5 cycles within the stimulus the practice. They were divided into several groups number of cycles has an influence on the CS.'5-8 The according to the type of pathology, thus: amblyopia CS is defined as the reciprocal of the value of the (n=98), optic nerve degeneration (n=42), and contrast at the threshold. cataract (n= 10). A group of normal persons (n=26), When the CS was determined as a function of the that is, without any known pathology or complaints, width of the stimulus, the monitor was at a viewing were also examined. distance of 100 cm; the width of the stimulus was The group of patients with optic nerve degener- changed from 160 to 10 in steps of a factor 0.5. The ation comprised those who had suffered from optic width dependence was determined at a spatial fre- neuritis in the previous three years, with or without quency of 1-6 c/deg. other evidence of multiple sclerosis. Optic neuritis The CS data were automatically plotted in double was electrophysiologically diagnosed in that period. logarithmic form against spatial frequency or against This group was divided into two subgroups, those the width of the stimulus, depending on the type with normal visual acuities (greater than or equal to of determination. The apparatus with which the 1 0; n=16) and those with diminished visual acuities measurements were performed was specially de- (less than 1 0; n=26). The group of cataractous eyes signed for clinical use."9 The measuring procedure was examined to enable a comparison to be made runs automatically and is controlled by a micro- between groups of eyes with decreased visual acuity processor system. The whole measuring sequence originating from different causes. including patient instruction takes about 25 minutes Cases with evidence of multiple pathology were for both eyes. excluded. From a statistical point of view each For each of the CSFs the corresponding grating patient was allowed to occur only once in the series. acuity was calculated by extrapolating the high fre- Hence in cases of binocular disease only one eye quency part of the CSF to the maximum contrast (chosen at random) was used. Five amblyopes had (CS= 1). A quadratic function was used for. the binocular disease, while 93 had monocular. The extrapolation, as this type of curve yielded a good fit numbers of binocular and monocular cases of optic for the appropriate part of the CSF. The grating nerve degeneration were 7 and 35, and of cataract 4 acuity is expressed as the reciprocal of the minimum and 6, respectively. angle of resolution in minutes of arc, as is the Snellen The mean age was 35 years, standard deviation visual acuity. The acuity A (1/min of arc) is equiv- 10 years, for both normal and amblyopic subjects; 45, alent to a spatial frequency of 30xA [c/deg]. SD 17, years for subjects in both optic nerve degener- The slope of the CS between 10 and 40 width was ation subgroups; and 59, SD 11, years for the subjects taken as the measure for the width dependence of the in the cataract group. CS. This was expressed as: The visual acuity of all subjects was determined by S=(log(CS(4°))-log(CS(10))/log(4). the Snellen optotype chart. Results DETERMINATION OF CONTRAST SENSITIVITY The contrast sensitivity (CS) was determined by Point by point averaging of the data points of the presenting vertical sinusoidal gratings on a television CSFs within each of the groups yielded the group- monitor. The contrast of the stimulus was variable; specific CSFs. The curves for the group of normal the contrast threshold was determined for each eye eyes and for the group ofamblyopic eyes are depicted separately by a modified von B&k6sy tracking tech- in Fig. 1. The mean visual acuity for the group of nique.'4 The spatial frequency of the grating and the normal eyes was 1 34; the mean grating acuity was width ofthe stimulus could be adjusted. The height of 1-18. the stimulus always subtended 50 visual angle; the As is to be expected from the lower mean visual average luminance of the stimulus was 5 cd/m2. The acuity in the amblyopic eyes (0-39), contrast sensi- examinations were performed in a dark room. The tivity at the higher spatial frequencies (CS(HF)) was grating was counterphased every 0-6 s to prevent low. This was associated with a decrease in the after-images. grating acuity, which had a mean value of 0-80. 60 A C W Volkers, KHHagemans, GJ van der Wildt, and PIM Schmitz
1. O - 0 'Z 0. 5 - ce 0. 3- Fig. 1 Lower left: Averaged CSFs 0. 2- for26 normal visualsystems andfor 500 98 amblyopic eyes. The dashed 300 lines atthehighfrequency endofthe 200 CSFs indicate extrapolation to the ,5 100 grating acuity. Lower right: The 4-0 v 50 relationships between contrast C sensitivity and width ofthestimulus UA 30 ('width dependence')for normal * 20 eyes andfor amblyopic eyes. Upper tau left and right: Ratio ofthe two w 10 curves in the lowerpart with the 0 5 normal one as the reference. The 3 o normal (n=26) curve on the left (the ratio ofthe two 2 eyes CSFs) is called the 'visuogram'. \ '\ * amblyopic eyes (n=98) 0.1 0.2 0.5 1 2 3 5 10 20 3050 1 2 3 5 10 20 Spatial frequency (c/deg) Width (deg)
Hence the resolving power calculated from the CSF contrast is needed to reach the threshold. In the was about twice the value determined with the amblyopia group the CS decreased more rapidly as Snellen letter chart. The graph at the top of Fig. 1 the width of the grating was reduced than in the shows the ratio ofthe value for amblyopic eyes to that normal group: the slope between 1° and 40 was much for normal eyes. This type of curve is called a greater in the amblyopic visual systems. This result, 'visuogram. '20 The difference between amblyopic and which was anticipated in the pilot study performed by normal eyes became more pronounced at higher Hagemans and van der Wildt,9 was confirmed by this frequencies. study of 98 amblyopic eyes. The results were almost The right-hand graphs show the relationship be- identical among subdivisions of the group of tween CS and the width of the stimulus. Below a amblyopic eyes: 41 cases of anisometropia, 32 cases width of 40 the CS decreased considerably, which of strabismus, and 25 cases of unknown origin. These means that for a narrower grating presentation more results are not shown here.
1.0o
0 0. 5 -
0. 3 - A A 0. 2- 500 - 300 Fig. 2. Lower left: Averaged CSFs 200 - for26 normal eyes, for optic nerve degeneration with normal visual 100 - acuity (n= 16), and with decreased visualacuity (n=26). Lower right: 50 - a) 30 - Width dependencefor thesame U, groups. Upper: Ratio curves with 20 - the normal curve as the reference. L. 1 0 - 0 u 5 - 3 - O normal eyes (n=26) 0 * OND VA >- (n=l 6) 2 - A OND VA <1 (n=26)
0. 10. 2 0. 5 1 2 3 5 10 20 30 50 1 2 3 5 10 20 Spatial frequency (c/deg) Width ( deg ) Spatial contrast sensitivity and the diagnosis ofamblyopia 61
1. 0 - ._0 0. 5 - 0. 3 1 .pppp 0.2-2 500 - 300 - 200 - >v ._ 100 Fig. 3 Lower left: Averaged CSFs 50 for26 normaland 10 cataractous ._ 0' eyes. Lower right: Width 30 *dependencefor thesame groups. 4) 20 - Upper: Ratio curves with the normal (A, curve as the reference. 0 - 0 u 5 3- o normal eyes (n=26) \ ' * cataract (n=10) 2-
g \~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I' 0.1 0.2 0.5 1 2 3 5 10 203050 1 2 3 5 10 20 Spatial frequency (c/deg) Width(deg)
The averaged results for the two subgroups of eyes Discussion with optic nerve degeneration are shown in Fig. 2; the CSF for the normal group is given as a reference. VISUAL ACUITY AND GRATING ACUITY The CSs for both optic nerve degeneration subgroups The Snellen test of visual acuity (VA) is an important were lower than for the normal group at all spatial tool in ophthalmological practice. It is a measure of frequencies. For both subgroups the CS in the low the performance ofthe visual system at high contrast, spatial frequency range (CS(LF)) was reduced by and it is claimed that it evaluates the patient's reading about the same amount, while the CS(HF) was lower ability. It is expressed as the reciprocal of the in the low vision subgroup than in the subgroup with minimum angle of resolution in minutes of arc. The normal visual acuity. The calculated grating acuity grating acuity (GA) may also be expressed in the values were 0-71 and 1-01 respectively. The corres- same units; half a cycle of the highest visible fre- ponding visual acuities were 049 and 1.14. In both quency is taken as the minimum angle of resolution. subgroups the width dependence (see right-hand Some relationship between VA and GA may be graph) had a similar form to that in the reference expected, though the stimuli are different (respect- group, but the whole curve was lower. This is to be ively optotypes and gratings). expected from the lower CS value at 1-6 c/deg in the Our results indicate that there is a discrepancy CSF. The results were almost the same among between the two estimates for the spatial resolution. subgroups containing 23 cases of optic neuritis with- For the normal group and for the optic nerve out and 19 cases of optic neuritis with multiple degeneration subgroup with normal VA the VA is on sclerosis. This is not shown here. average greater than the GA, while for the group of Comparison of the averaged results for the amblyopic eyes the VA is much lower than the GA. cataractous eyes with those of the normal group in The VA is also lower than the GA for the low vision Fig. 3 indicates a reduction in CS(HF). This was subgroup with optic nerve degeneration, albeit to a expected from the lower mean visual acuity in this lesser extent. Only for the group of cataractous eyes group (0-45); the grating acuity (0-48) was in accord- are VA and GA nearly equal. ance with this. In this study it is not proved whether in amblyopia On the basis of these results we can recognise at the GA is an overestimation of the real spatial least three differences between some of the groups. resolution or the VA an underestimation. Levi and These differences, which we shall study in further colleagues found identical differences 22 and suggest detail, can be seen in: (1) the discrepancy between it is because amblyopic visual systems have edge the two types of spatial resolution estimate-visual sensitive mechanisms broader than normal.2324 In five acuity and grating acuity; (2) the dependence of the amblyopic subjects we also determined the VA with CS on the width of the stimulus; (3) the contrast single letters (Sheridan-Gardiner test), and in four of sensitivity for the low spatial frequency range. them this measure for VA coincided with the GA, 62 A C W Volkers, K H Hagemans, GJ van der Wildt, and P1 M Schmitz
40 Fi WIDTH DEPENDENCE 1- x 1 00% 30 - n The second differentiating finding is the dependence 20 normal eyes of CS on the width of the stimulus. To detect a 10 n=26 narrower grating more contrast is needed. In our unI experimental data the largest effect was found 40 between 1° and 2° width. To obtain a statistical 30 amblyopia spread the slope S was calculated for the range from 20 10 to 4°. The slope is much higher for amblyopic eyes 10 n=98 than for non-amblyopic eyes, as shown in Figs. 1 to 3. 0 An explanation for differences in width depend- 40 ence between amblyopic and other eyes may be 30 OND VA AI found in an influence of the sharp luminance discon- 20 tinuity on the contrast needed for perception of the 10 n=16 grating when the edges of the narrowing stimulus 0 1 H-1. approach each other. It is not known whether the 40 edge sensitive mechanism of Levi et al.24 is sufficient 30 OND VA <1 as an explanation for the higher slope S in amblyopia, 20 n=26 as the former seemed to play a part at approximately 10 of while the effect in the latter Xm| rmH 10 min arc, already 0 t happens at a half width of 120 min of arc. If an edge 40 sensitive mechanism explains differences in 30 cataract VA-GA as well in width dependence among several 20 of eyes, then there must be a correlation 10 n=10~groupsbetween the two variables. Indeed for the groups of H] 0 26 normal and 98 amblyopic eyes there is a significant -48 -36 -24 -12 0 12 24 36 correlation between VA-GA and the slope S (using
VA-GA (cideg) a=0.05), though the association is weak (r= -0.39). Fig. 4 Frequency distribution ofdifference bbetween visual Research in this field is continuing.' acuity andgrating acuity in the groups ofnornnal, amblyopic, Fig. 5 gives the normalised frequency distributions optic nerve degenerative (normal and low visit alacuity), and for width dependence for all five groups. It is evident cataractous eyes. The distributions are normalisedto a total that the width dependence distribution for amblyopic of100%. eyes lies to the right, whereas the one for cataractous eyes lies to the left, with the other distributions somewhere between. There is a considerable over- though the values were still subnormal. This so called lap. crowding phenomenon or separation diffficultyll26 can The result of the analysis of variance revealed a be explained by the same edge sensitive mechanism. statistically significant difference between the The discrepancy between VA and C3A (which is groups: F(4,175)=21.0; p<00001. This suggests that greatest in the amblyopia group) was further studied the slope can also be used for differentiating between for its potential as a diagnostic measure..The discrep- amblyopia and the other groups. Tukey's multiple ancy was expressed as the difference Ibetween VA range test indicated a significant difference (using a= and GA. Fig. 4 gives the frequency disstributions of 0-05) between the amblyopic group of eyes and all VA-GA for the groups of normsal eyes and other groups. amblyopic eyes, for both subgroups of e!yes with optic nerve degeneration, and for the group c f cataractous CONTRAST SENSITIVITY FOR THE LOW SPATIAL FREQUENCY eyes. Each distribution is normalised to a total of RANGE 100%. An analysis of variance" shove,ved that the The two differences already discussed were between distributions cannot be regarded as samples from the the amblyopia group and the other groups. A third same population: F(4,175)=29-7; p - 50 Fi x 100% 40 Fi I 00% 40 - n - n 30 normal eyes 30 - 20- normal eyes n=26 20 - n=26 10 i. k- 10 - u 0 -1 40 40 30 amblyopia 30 - 20 amblyopia n=98 20 - 10 10 n= 98 0 I 0 40 40 30 OND VA 1 30 - OND VA>1 20 - 20- n=16 20 - in - F] 0 n=16 0 I mr mI I m 40 40 30 OND VA <1 30 OND VA<1 20 20 r n=26 10 n=26 10 aiFnlIVTTi, 0 0 40 40 30 cataract 30 cataract 20 20 n=10 10 10 n=10 0 0 3 5 H1 1 I I2 3 5 H- 102 -1 0 1 2 2 3 5 1 0 20 30 50 1 00 200 Slope (logunits CS/logunit width) Contrast sensitivity (low Fs) Fig. 5 Normalisedfrequency distributionforthe width Fig. 6 Normalisedfrequency distributionfor contrast dependence ofcontrastsensitivity;further as Fig. 4. sensitivity at low spatialfrequencies;further as Fig. 4. ation subgroups in general have lower values, though Tukey's multiple range test not only showed signifi- there is a certain overlap with the other groups. It is a cant differences (using a=0-05) between both optic common finding that some cases of optic neuritis or nerve degenerative subgroups on the one hand and multiple sclerosis show decreased CS for only low, the groups with amblyopic and normal eyes on the middle, or high spatial frequencies, or for combi- other, but also that the CS(LF) for cataractous eyes nations.45I0 The cataract group is somewhere in was significantly lower than for normal eyes. between; the distributions in Fig. 6 show that the one for cataract is not bimodal and so does not support COMBINATION OF FINDINGS the two-type classification of Hess and Woo. 13 Each of the three variables so far described can be Inspection of the individual CSFs of cataractous eyes used to differentiate between different visual defects. shows that low CS(LF) in this group may be due to We investigated the diagnostic value of combining the fact that in a considerable number of cases the the three variables in one test to differentiate CS(HF) was decreased to such an extent that not only between some of the groups, as stated in the intro- the high frequency range but also the low frequency duction: (1) between amblyopia and optic nerve range was affected. Further, for the group of degeneration with low visual acuity; and (2) between amblyopic eyes we did not find a bimodal distribution normal eyes and optic nerve degeneration with of CS(LF), as Hess and Howell postulated.2 normal vision. We used logistic discriminant The analysis of variance indicates a significant analysis' to answer the two questions. difference between the groups: F(4,175)=16-2; Diagnosis on the basis of our chosen variables was p 100 Table 1 Results oflogistic discriminant analysisfor the group ofamblyopic eyes andfor the low vision optic nerve 90 degeneration (OND) subgroupfor a specific choice (see text) 80 v Diagnosis resulting Known diagnosis from tests 70 OND (VA -Il 40 C low (that is, many amblyopia cases are erroneously 30 classified as optic nerve degeneration). 20 The clinician has to decide whether it is acceptable to miss a number of amblyopia cases in order to 10 obtain correct diagnosis for a greater proportion of optic nerve degeneration cases. Table 1 gives one 0 example of such a choice, in which both the 0 1 0 20 30 40 50 60 70 80 90 100 sensitivity and the specificity are about 80%. This 100 * [1 - specificity (amblyopia) I choice corresponds to the following linear combi- Fig. 7 ROC curve showing, for the combination ofthe three nation of the three variables discussed above: variables, the relationship between the diagnosticsensitivity Z=6-7+0-05*(VA-GA)-0-18*CS(LF)-0-4*S. for optic nerve degeneration with low visual acuity and the Whenever Z>- 1-4 the patient is assigned to the optic diagnosticspecificityfor amblyopia. nerve degeneration group, otherwise to the amblyopia group. By means of this choice it turns out optic nerve degeneration cases (sensitivity of the that a correct diagnosis is made in 80/98 or 82% of the tests), and the percentage of incorrectly diagnosed amblyopes and in 21/26 (81%) of the optic nerve amblyopia cases (1-specificity). The curve shows degeneration group, while an erroneous decision was that if one chooses limits which ensure that the taken in 23 out of 124 cases (19%). The predictive sensitivity is high (that is, many cases of optic nerve value of the test for amblyopia is high: 80/85 or 94%. degeneration are diagnosed correctly) specificity is For optic nerve degeneration the predictive value is considerably lower: 21/39 or 54%. Of course these 100 last values depend on the incidences of the two types of disorders in the tested population. 90 In a similar way results for the comparison between 80 the optic nerve degeneration subgroup which has 5- normal VA and the group of normal eyes are given in a 70 Fig. 8 (ROC curve) and Table 2 (result of logistic z 60 discriminant analysis). If limits for normal tests are 0 chosen so that 95% of the 26 normal eyes can be 50 considered as such, it turns out that 12 out of 16 cases of optic nerve degeneration with normal vision ._ 40 en (75%) are not assigned to the normal group despite 30 the normal VA. Yo CD 20 Table 2 Results oflogistic discriminant analysisfor the group ofnormal eyes andfor the optic nerve degeneration 10 (OND) subgroup with normal visual acuity for a specific choice (see text) 0 1 0 20 30 40 50 60 70 80 90 100 Diagnosis resulting Known diagnosis from tests 100 * [1 - specificity (normal eyes)] OND (VA _J) Normal eyes Total 8 ROCcurveshowing, forthe combination ofthe three Fig. OND 12 1 13 variables, the relationship between the diagnostic sensitivity Normal eyes 4 25 29 for optic nerve degeneration with normal visualacuity and Total 16 26 42 the diagnostic specificityfornormal eyes. Spatial contrastsensitivity and the diagnosis ofamblyopia 65 CONCLUSIONS 9 Hagemans KH, van der Wildt GJ. The influence of the stimulus width on the contrast sensitivity function in amblyopia. Invest In the introduction several questions were raised Ophthalmol Vis Sci 1979; 18: 842-7. which we can now answer. 10 Regan D, Silver R, Murray TJ. Visual acuity and contrast The shape of the spatial contrast sensitivity sensitivity in multiple sclerosis-hidden visual loss; an auxiliary function is partly related to the type of pathology. diagnostic test. Brain 1977; 100: 563-79. 11 Frisdn L, Sjostrand J. Contrast sensitivity in optic neuritis; a Specific for amblyopia is a discrepancy between preliminary report. Doc Ophthalmol Proc Ser 1979; 17: 165-73. visual acuity and grating acuity (VA