The Effect of Background Luminance on Cone Sensitivity Functions

Transcription

1 October 1969 Vol. 30/10 Investigative Ophthalmology & Visual Science Articles The Effect of Background Luminance on Cone Sensitivity Functions Tsaiyoo Yeh, Vivionne C. Smith, and Joel Pokorny Implementations of the Wald-Marre technique have employed fixed luminance backgrounds to isolate cone sensitivity mechanisms. We evaluated the effect of varying the adaptation level on the relative isolation of the different cone types. For MWS and LWS cone isolation we used a 15 Hz flickering test light to isolate the achromatic channel, and we modelled the resulting spectral sensitivity functions as a linear sum of LWS and MWS input. We found only mild improvement in relative cone isolation with increasing adaptation level. The LWS and MWS cone mechanisms showed decreasing sensitivity with adaptation level and reached limiting Weber behavior above 1000 Td. SWS cones were isolated with a Hz flickering light. SWS cone isolation improved with adapting level, reaching a plateau above 1000 Td. The SWS cone mechanism showed decreasing sensitivity with adaptation level but did not reach a limiting Weber region. Our data indicate that the use of fixed high adaptation levels has different effects on the cone mechanisms. Absolute sensitivity loss for LWS or MWS mechanism will not be revealed. LWS and MWS thresholds will appear normal unless there is an adaptation abnormality. On the other hand, the SWS cone thresholds would be sensitive to both absolute and increment sensitivity loss. More than one adaptation condition is needed to separate different types of sensitivity loss characteristic of eye disease. Invest Ophthalmol Vis Sci 30: , 1989 In recent years there has been considerable interest in the use of test sensitivity increment threshold techniques to estimate the sensitivities of cone receptor systems in clinical populations. A typical procedure involves using a chromatic background to adapt two cone types leaving the third relatively more sensitive. A yellow background may be used to favor detection by the short-wavelength sensitive cones (SWS cones), a purple background for middle-wavelength sensitive cones (MWS cones) and a blue background for longwavelength sensitive cones (LWS cones). With this method, the maxima of the sensitivity functions lie at about 440, 545 and 570 nm, respectively, 12 in general agreement with estimates of the spectral sensitivity functions of the three cone types measured by a variety of other techniques. 3 Marion Marre modified Wald's technique for clinical use, 4 " 6 employing less From The University of Chicago, Eye Research Laboratories, Chicago, IL. Supported in part by EY (JP) and EY (VCS). Submitted for publication: March 3, 1989; accepted June 2, Reprint requests: Vivianne C. Smith, The University of Chicago, Eye Research Laboratories, 939 East 57th Street, Chicago, IL radiant adaptation fields. Specifically, spectral sensitivities were estimated using a 1.2, 50 msec test field of variable wavelength on backgrounds of about 1500 Td. 7 Although the cone spectral sensitivities are not completely isolated by the Wald-Marre technique, the measurement has been of great importance in the evaluation of cone function in acquired color vision defects that may accompany ophthalmic and systemic disease. Lutze, Pokorny and Smith 8 attempted to improve cone isolation in a test sensitivity paradigm by the use offlickeringtest lights. The rationale for this stimulus paradigm came from both recent theoretical and empirical studies. According to modern color vision theory, chromatic detection involves combination of signals from three orthogonal or cardinal channels: the "achromatic" (LWS + MWS) channel, and two "chromatic" channels, a red-green opponent (LWS - MWS) channel, and a channel signaling SWS cone activity. The "achromatic" channel is treated as a linear sum of LWS and MWS cone signals. Recent studies 9 "" suggested that a flickering test (>15 Hz) would be a more effective stimulus for isolation of the spectral sensitivity of the MWS and LWS cone types than the frequency employed pulse stimulus. Since the achromatic channel is believed to show better 2077

2 2078 INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE / Ocrober 1989 Vol. 30 temporal resolution than the chromatic channel, the use of flicker serves to minimize the input from chromatic channel activity and aid in the isolation of MWS and LWS cones. Although the modified Wald-Marre technique did give improved isolation for both the LWS and MWS cones, the isolation of MWS cones at longer wavelengths remained incomplete. 8 Figure 1 shows the theoretical isolation of the individual cone types for several adaptation wavelengths. The more sensitive mechanisms are adapted strongly by the background revealing the less sensitive mechanism at other test wavelengths. Once the sensitivities of the two mechanisms are equivalent at the test wavelength matching the adaptation wavelength, further increase in adaptation affects both mechanisms equally and no further isolation is obtained. Maximal isolation is given by adaptation at a wavelength of maximal difference in sensitivities of the unadapted cone types: that is, a 600 nm or longer background for SWS cones, a 650 nm or longer background for MWS cones, and a 460 nm background for LWS cones. In the present study, cone sensitivity functions were studied using the modified Wald-Marre technique 8 and varying the background luminance level. First, isolation of cone sensitivity functions under different adaptation levels was compared. Then, using the idea that the achromatic channel, as isolated by theflickeringstimulus, sums LWS and MWS cone signals linearly, we computed the relative isolation by fitting linear combinations of LWS and MWS cone sensitivity functions to the data. The thresholdluminance curves were also examined, to assess the use of the Wald-Marre technique in evaluation of different sensitivity loss characteristics in eye disease. Apparatus Materials and Methods A computer-controlled two-channel Maxwellian view system 8 was used to measure test sensitivities on chromatic backgrounds. Background luminance was varied in 0.5 log unit steps over a 3.0 log unit range. For the yellow adaptation field (Wratten 22) used for SWS cone isolation, the background luminance Fig. 1. Theoretical isolation of cone types. The top panel shows estimation of the spectral sensitivities of the cone photopigments. 12 The height of the LWS and MWS sensitivity functions are set so that they sum for the photopic luminosity function. The height of the SWS sensitivity function is arbitrary. The three lower panels show the expected heights for the three cone sensitivity functions under condition of optimal isolation for each of the selective chromatic adaptation conditions WAVELENGTH YELLOW Hz PURPLE IS Hz BLUE 15 Hz J WAVELENGTH J WAVELENGTH WAVELENGTH

3 No. 10 CONE SENSITIVITY FUNCTIONS / Yeh er ol 2079 i i i o.ochbb 1.6 OEG -. YELLOW / 0 PURPLE / 0 BLUE / / -l.c - 0.W- -2.C -2.C- -3.C m 0-3.C- i i i LOG TR0LRND LOG TR0LRND Fig. 2. The test thresholds of each background for three subjects. The solid circles in each panel represent the data obtained for a yellow adaptation field. The open circles represent the data obtained for a purple adaptation field. The solid squares represent the data obtained for a blue adaptation field. The left panels show the 1.6 field data. The right panels show the 5.5 field data. CL 1.6 DEG o.oc- YELLOW o PURPLE BLUE -l.c- -2.C- -3.Ci - ^y^ i i i // i LOG TR0LRND oc -CL 5.5 OEG YELLOW 0 PURPLE / n BLUE V^ l.c 2.C 3.C - A-" ' LOG TR0LRND o.oc. TY 1.6 DEG YELLOW 0 PURPLE BLUE i I 0.0(. TY 5.5 OEG YaLOW o PURPLE BLUE / -2.C - ' /.< r -2.C 0-3.C 1 1 i LOG TR0LAN LOG TR0LRND5 ranged from 1.7 log to 4.7 log trolands; for the purple (Wratten 35) and blue (Wratten 47 B) adaptation fields used for MWS and LWS cone isolation, the background luminance ranged from log to log trolands. The test field size was fixed at either 1.6 or 5.5. Test wavelengths were selected from a random list in steps of 20 or 30 nm. For the yellow background, the range of test wavelengths was from 420 to 520 nm; for the purple background, the range was from 420 to 650 nm; and for the blue background, the range was from 450 to 670 nm. The test alternation rate was Hz on the yellow background, and 15 Hz on the blue and purple backgrounds. A 3 log unit neutral density wedge was used to control test luminance. Energy calibrations for the test wavelengths and for the relative transmission of the neutral density wedge were done by an EG & G photometer/radiometer. The relative energy output for the different test wavelengths was measured with a laboratory constructed spectroradiometer. Procedure A chin rest aided the observer in maintaining alignment with the 2 mm Maxwellian view artificial

4 2080 INVESTIGATIVE OPHTHALMOLOGY G VISUAL SCIENCE / Ocrober 1989 Vol. 30 Table 1. The least mean square residual errors for each adaptation field. (A) The fit using single Smith- Pokorny fundamentals, (B) the fit using the best combination templates Field and subjects Background luminance (log Tds) (A) (B) BB ].6 CL Yellow background (fitted by SWS cone template) Purple background (fitted by MWS cone template) TY Blue background (fitted by LWS cone template) BB CL Purple background (fitted by the best MWS + LWS cone combination template) Blue background (fitted by the best MWS + LWS cone combination template) TY pupil. Following an adaptation of 2 min for the yellow background, and 1 min for the purple and blue backgrounds, test thresholds were measured using a computer-controlled tracking procedure. To begin a trial, the neutral density wedge was set to increment (0.125 log unit/sec) from its dense end. When the flickering test appeared the observer pushed a button to reverse the direction of the wedge to decrement (0.065 log unit/sec). When the test disappeared, the observer released the button which reversed the direction to increment (0.065 log unit/sec). The average of six reversals that fell within a criterion variance was calculated as threshold. Spectral sensitivity curves for each background were obtained by fitting the data to Smith-Pokorny fundamentals 12 with a least mean square procedure. The template heights which were specified by the template peaks were taken to indicate the relative sensitivities of each cone type contributing to the response. Observers The observers (BB: female, 18 years old; CL: male, 28 years old; and TY: female, 24 years old) all had normal color vision as defined by the Ishihara and

5 No. 10 CONE SENSITIVITY FUNCTIONS / Yeh er ol TEST WAVELENGTH TEST WflVELENGTH -3, TEST VRVELENGTH TEST WflVELENGTH -3 DO TEST WflVELENGTH TEST WflVELENGTH Fig. 3. The best MWS and LWS combination templates fitted to the purple and blue background data for a 1.6 field. Luminance levels decreased from bottom to top (solid circles with solid line: log trolandfield;open circles with dashed line: log trolandfield;solid squares with solid line: log troland field; open squares with dashed line: log troland field; solid triangles with solid line: log troland field; open triangles with dashed line: log troland field; solid circles with solid line: log troland field). Standard Pseudoisochromatic Plate tests, the Neitz anomaloscope and the Farnsworth-Munsell 100- Hue test. Observer BB was a protan carrier (denned by family history) who had reduced flicker photometric sensitivity to long wavelength light. 13 Results Threshold-luminance data are shown in Figure 2. The log of reciprocal template peak is plotted as a function of log luminance of the background for the three observers and both field sizes. The sensitivity decreased as background luminance increased for all three chromatic adaptation conditions. For the blue and purple backgrounds, the data approached a Weber region (slope of 1 on these plots) above 1000 trolands. For the yellow background, the sensitivity changed slowly with background luminance and did not reach a Weber region. Similar results were obtained with both 1.6 and 5.5 fields. We took the residual least mean square value to indicate the goodness of fit between the data and the theoretical spectral sensitivity functions (Table 1A). For the yellow background, the least mean square values decreased as background luminance increased and did not asymptote until the adaptation field exceeded 1000 trolands. Thus, isolation of the SWS cone sensitivity function is incomplete below 1000 trolands. For the blue and purple backgrounds, there was little change in the residual least mean square values with change in luminance. Based on our rationale that detection of 15 Hz flicker on blue or purple backgrounds occurred in the achromatic channel, we proposed that these data can be fitted well by a linear sum of the LWS and

6 2082 INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE / October 1989 Vol. 30 BB 1.6 DEC BB 9.5 OEG CL 1.6 OEG CL 5.S DEG a- o 6 Fig. 4. The templates weights for the purple (A) and blue (B) backgrounds data. Solid circles with solid line are the LWS cone component for each experimental condition, and open circles with dashed line are the MWS cone component. TY 1.6 OEG TY 5.5 DEG.O' o <* " MWS cone types, with one or the other cone type dominating the spectral sensitivity functions for each of the chromatic adaptation conditions. We denned a new template as the sum of LWS and MWS fundamentals which yielded the least mean square residual error. The combination templates can be used to indicate the relative isolation of cone types only when these templates are good description of the data. The combination templates gave consistent good fits to the data. Therefore, we suggest that the linear combination of cone sensitivity functions can be used as a quantitative index for the relative isolation of MWS and LWS cone sensitivities. The proportion of the two fundamentals that gave the least mean square residual value was taken to indicate the relative isolation of each background and the residuals for this fitting procedure are shown in Table IB. Table 2 shows the relative isolation for the blue and purple backgrounds at each adaptation level for the 1.6 and 5.5 fields. The relative isolation for the purple background is the proportion of MWS function in the best fitting linear combination of LWS and MWS fundamentals, while the relative isolation for the blue background is the proportion of LWS function in the best fitting LWS and MWS combination. The relative isolation ranged from 0.60 to 1.00

7 No. 10 CONE SENSITIVITY FUNCTIONS / Yeh er ol 2083 B BB 1.6 DEG ' BLUE BACKGROUND 0 MWS CONE COMPONENT BB 9.9 DEG BLUE BACKGROUND o, ' /..9--y> A CL 1.6 DEG BLUE BACKGROUND CL S.9 DEG BLUE BACKGROUND Fig. 4. (B) See legend under Figure 4 (A). TY 1.6 DEG BLUE BACKGROUND TY 5.9 DEG BLUE BACKGROUND for the purple adaptation field, and from 0.33 to 1.00 for the blue adaptation field. The 1.6 data did not show a consistent improvement with increased luminance level. For the 5.5 field, it was more evident that isolation was better at higher background luminances. Figure 3 summarizes the best MWS and LWS combination templates fitted to the data of the purple and blue backgrounds at each luminance level. Template sensitivities increased as adapting background luminance decreased. A good fit is seen as the data points follow the templates throughout the spectrum. However, at luminance levels below 2 log trolands, larger least mean square residuals were found from the template fit, as shown in Table 1. Here we do not report combination templates for the spectral sensitivity functions obtained with yellow adaptation since the rule of combination of the fundamental templates is not as well defined as for the LWS and MWS stimulus conditions described above. Figure 4 shows the template weights needed for the fits of Figure 3 expressed as log sensitivity and plotted

8 2084 INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE / Ocrober 1989 Vol. 30 Table 2. The relative isolation for the blue and purple backgrounds for three observers and both field sizes Field and subjects Background luminance (log Tds) BB 1.6 CL TY BB 5.5 CL TY Purple background Blue background as a function of adapting luminance. A function of the form: log T = log T o + log {(A + A o )/A o } (1) wasfitted to the data using the Marquardt-Levenberg algorithm for nonlinear parameter estimation (as implemented in the Math View Professional software package for the Macintosh). In this equation T refers to test threshold and A to adapting luminance. T o is the unadapted threshold and A o is the adapting luminance where T starts to rise. In preparing these fits, we used the X max of the M cone (540 nm) as the test wavelength for the purple adapting field and the X max of the L cone (570 nm) as the test wavelength for the blue adaptingfield.for the majority of conditions the estimated sensitivities fall on the predictions in an orderly way, indicating that thresholds are following the expected form for increment thresholds for single mechanisms. For the blue background (Fig. 4B), the LWS cone T o is more sensitive than the MWS cone T o at 570 nm in four of six conditions. In general the adapting field has parallel adapting effects on the curves, so that little change in isolation occurs. Only with the large field do two observers show an initially greater MWS sensitivity and in this case the blue adapting field does show a differential effect. For the purple background (Fig. 4A), the unadapted sensitivities at 540 nm are similar for CL and TY. BB showed greater MWS cone T o sensitivity than LWS. The purple background has a pronounced differential effect for most of the six data sets, resulting in greater MWS cone isolation with increase in adapting luminance. This analysis does not indicate any sign of the "super-weber" behavior postulated by Eisner and MacLeod 9 for a similar paradigm. Discussion According to the theory of chromatic adaptation, as adapting luminance is increased, cone isolation should improve until the adapted sensitivities of the cone mechanisms are similar (Fig. 1). This expectation was fulfilled for SWS cone isolation which reached a plateau at 1000 trolands. The expectation was not confirmed for the LWS and MWS cone isolation data at 1.6 which showed only a weak trend for better isolation at higher background luminance. The 5.5 field data were more consistent. For LWS cone isolation (blue background), we did not expect much effect of adaptation since the achromatic channel is pinned near 500 nm. This expectation was confirmed. Only for BB and CL with a 5.5 field did we see a consistent improvement in isolation with the increased background luminance. For MWS cone isolation (purple background), there was a trend toward improved isolation with luminance. Best isolation was obtained in BB who is an observer with normal color vision but shows evidence of being a heterozygote for congenital protan defect. 13 The work of Greenstein, Hood and Carr 1617 suggests two possible types of sensitivity change attributable to eye disease: a reduction in absolute sensitivity and/or a reduction in increment sensitivity. For absolute sensitivity loss, the threshold-luminance curve

9 No. 10 CONE SENSITIVITY FUNCTIONS / Yeh er ol 2085 ABSOLUTE SENSITIVITY LOSS INCREMENT SENSITIVITY LOSS luminance. The single adaptation level of about 1500 trolands as used by Marre 7 cannot separate the two possible causes of the sensitivity loss. For effective evaluation of the Weber region for SWS cones, it is necessary to modify the protocol to include a background whose spectral composition more effectively stimulates SWS cones. The analysis of color defects by the Wald-Marre approach also implies the assumption that thresholds for the different receptor types are measured at a similar adapting level. If different laboratories use different adapting levels to test sensitivity functions in patients, controversial results may result. Reports 18 that the SWS-based Color Vision Mechanism is more sensitive to disease than the MWS- and LWS-based Color Vision Mechanisms need to be reevaluated. Our data and the studies by Kalloniatis and Harwerth 19 and Greenstein et al 20 ' 21 stress the importance of using more than one adaptation level to separate the different types of sensitivity loss in eye disease. Key words: chromatic adaptation, cone isolation, flicker, Wald-Marre functions References Fig. 5. Predicted luminance-threshold curves for two types of sensitivity loss. The continues curve represents the curve for a normal color visual system. The dashed curve represents the predicted curve for absolute sensitivity loss; the dotted curve for increment sensitivity loss. would shift up and to the right by equal amounts compared to the curve for the normal observers (Fig. 5A), which increment sensitivity loss would show only a vertical shift on sensitivity (Fig. 5B). The data showed Weber behavior for the blue and purple backgrounds above 1000 trolands (Fig. 2). A reduction in absolute sensitivity would not be revealed within the Weber region, only reductions in increment sensitivity. In comparison, sensitivity on the yellow background is not within the Weber range. The SWS cones would be sensitive to both absolute and increment sensitivity loss. Thus the response to different types of sensitivity loss in eye disease are not the same for the three cone types at a high adapting 1. Wald G: The receptors of human color vision. Science 145:1007, Wald G: Defective color vision and its inheritance. Proc Natl Acad Sci USA 55:1347, Smith VC and Pokorny J: Spectral sensitivity of color-blind observers and the cone photopigments. Vision Res 2:2059, Marre M: Clinical examination of the three color vision mechanisms in acquired color vision defects. Mod Prob Ophthalmol 11:224, Marre M and Marre E: The influence of the three color visionmechanisms on the spectral sensitivity of the fovea. Mod Prob Ophthalmol 11:219, Marre M: The investigation of acquired colour vision deficiencies. In Colour 73. London, A. Hilger, 1973, pp Marre M: Wald-Marre approach to measurement of the three primary color vision mechanisms (CVMs). In Congenial and Acquired Color Vision Defect, Pokorny J, Smith VC, Verriest G, and Pinckers AJLG, editors. New York, Grune and Stratton, Lutze M, Pokorny J, and Smith VC: Improved clinical technique for Wald-Marre functions. In Vision Deficiencies VIII, Verriest G, editor. The Hague, Dr. W. Junk, Eisner A and MacLeod DI: Flicker photometric study of chromatic adaptation: selective suppression of cone inputs by colored backgrounds. J Opt Soc Am 71:705, Ingling CR Jr and Martinez E: Stiles TT5 mechanism: Failure to show univariance is caused by opponent-channel input. J Opt Soc Am 71:1134, Eisner A: Comparison of flicker-photometric and flickerthreshold spectral sensitivities while the eye is adapted to colored backgrounds. J Opt Soc Am 72:517, Smith VC and Pokorny J: Spectral sensitivity of the foveal

10 2086 INVESTIGATIVE OPHTHALMOLOGY G VISUAL SCIENCE / October 1989 Vol. 30 cone photopigments between 400 and 500 nm. Vision Res 15:161, Pokorny J, Smith VC, and Baron R: How many cones do you need for normal color perception? ARVO Abstracts. Invest Ophthalmol Vis Sci 29 (Suppl):298, King-Smith PE and Carden D: Luminance and opponentcolor contributions to visual detection and adaptation and to temporal and spatial integration. J Opt Soc Am 66:709, Kelly DH and van Norren D: Two-band model of heterochromaticflicker.j Opt Soc Am 67:1081, Greenstein V, Hood DC, and Carr RE: Foveal sensitivity changes in retinitis pigmentosa. Appl Optics 26:1385, Hood DC and Greenstein VC: Increment threshold (tvi) data and the site of disease action. OSA Topical Meeting: Noninvasive Assessment of Visual Function, 1988 Technical Digest Series, 3: Marre M and Marre E: The blue-mechanism in diseased eyes with eccentricfixation.in Colour Vision Deficiencies VI, Verriest G, editor. The Hague, Dr. W. Junk, Kalloniatis M and Harwerth RS: Differential adaptation of cone mechanisms explains the preferential loss of short-wavelength cone sensitivity in retinal disease. Doc Ophthalmol Proc Ser 52:353, Greenstein V, Hood DC, and Carr RE: A comparison of S cone pathway sensitivity loss in patients with diabetes and retinitis pigmentosa. Doc Ophthalmol Proc Ser 52:233, Greenstein V, Hood DC, Siegel IM, and Carr RE: Adaptation dependent sensitivity losses: implications for cone receptor involvement in RP. OSA Topical Meeting: Noninvasive Assessment of Visual Function, 1988 Technical Digest Series, 3:6.