Pattern Discrimination Perimetry in Patients With Glaucoma and Ocular Hypertension Matthew J. N u t a i t i s , L.C.D.R., W i l l i a m C. Stewart, M . D . , D o l o r e s M. Kelly, B.S., H u r s h e l l H. Hunt, P h . D . , a n d M a t t h e w L. Severns, P h . D .
We compared the results of the pattern discrimination perimeter to the program 30-2 on the Humphrey Field Analyzer (Humphrey, Inc., San Leandro, California) in 93 consecutive patients with ocular hypertension and glaucoma and 30 control patients. In 20 patients with ocular hypertension, a significantly greater number of glaucomatous defects were noted on pattern discrimination perimetry (ten patients) than on the program 30-2 (two patients) (P < .05, Wilcoxon signed rank test). The diversity in diagnoses found on pattern discrimination testing was not explained by age, intraocular pressure, refraction, number of glaucoma medicines, race, presence of vascular disease, optic disk status, or pupil size. In contrast, in 73 patients with glaucoma no statistical difference in the severity of diagnoses was noted between perimeters (P > .05, Wilcoxon signed rank test). These results suggest the potential value of pattern discrimination perimetry as a visual function test in patients with glaucoma and in defining subsets of patients with ocular hypertension not found with conventional automated perimetry.
Accepted for publication June 3, 1992. From the Glaucoma Service at the Department of Ophthalmology (Drs. Nutaitis and Stewart, and Ms. Kelley), and the Department of Biostatistics, Epidemiol ogy and System Sciences (Dr. Hunt), Medical University of South Carolina, Charleston, South Carolina; and the Wilmer Eye Institute (Dr. Severns), Johns Hopkins Uni versity School of Medicine, Baltimore, Maryland. This study was supported in part by unrestricted grants from Research to Prevent Blindness, Inc., New York, New York. Dr. Severns is an employee of LKC Technologies, Inc., Gaithersburg, Maryland. The opinions expressed herein are those of the authors and are not official opinions of the Department of the Navy. Reprint requests to William C. Stewart, M.D., Medical University of South Carolina, Storm Eye Institute, 171 Ashley Ave., Charleston, SC 29425-2236.
V I S U A L FIELD TESTING is performed in patients
with increased intraocular pressure to deter mine the amount of damage to the visual sys tem. Most perimeters currently available rely on a uniform white stimulus projected on a white background to assess visual function. Unfortunately, this stimulus detects glaucoma tous changes only after extensive damage to the optic nerve head has already been sustained. 1 The pattern discrimination perimeter, which recently has been described, requires recogni tion of a pattern stimulus, which may have the advantage of earlier detection of glaucomatous damage than conventional automated perime try. 1 We evaluated the use of pattern discrimi nation perimetry compared to automated pe rimetry in routine testing of patients with ocular hypertension or glaucoma.
Subjects and Methods We included in this study consecutively test ed patients with glaucoma and ocular hyper tension who were examined with automated perimetry at the Storm Eye Institute at the Medical University of South Carolina. Patients with glaucoma were defined as having typical glaucomatous-appearing optic disk damage and a history of an intraocular pressure greater than 21 mm Hg with or without visual field damage. Patients with ocular hypertension were defined as having a history of an increased intraocular pressure greater than 21 mm Hg without clinical signs of optic disk damage. We excluded patients with marked neurologic and retinal disease and those who were unable to complete a visual field examination. If both eyes were tested, only one eye was randomly chosen to be included in this study. A control group of normal individuals who had one eye randomly chosen to be examined also was in cluded in this study.
©AMERICAN JOURNAL OF OPHTHALMOLOGY 114:297-301, SEPTEMBER, 1992
297
298
AMERICAN JOURNAL OF OPHTHALMOLOGY
Each subject was tested with the Humphrey Field Analyzer (Humphrey, Inc., San Leandro, California), a fully automated perimeter with a 330-mm bowl that projects a uniform white stimulus on a white background. Subjects were examined with the program 30-2, which uses a staircase technique to determine threshold at 76 locations spaced 6 degrees apart and offset symmetrically across the vertical and horizon tal midlines within the central 30 degrees (Fig. 1)· After testing with the Humphrey Field Ana lyzer, subjects also were examined with the pattern discrimination perimeter (LKC Techno logies, Inc., Gaithersburg, Maryland), de scribed previously. 1 This perimeter's computer generates a display on a 5-inch television moni tor. An optical system relays this image to the eye while providing refractive correction and eccentricity-dependent magnification such that the image appears to be on the surface of a bowl located at optical infinity. The graphics display produces black or white pixels, 640 in number horizontally and 480 vertically. The perimeter has a 30-degree field of view. A fixation target is present in the middle of the graphic display, and fixation is checked by a technician who uses a separate television monitor. Using the pattern discrimination perimeter, we tested 32 locations within the arcuate area of the visual field at the same locations tested by the program 30-2 on the Humphrey Field Analyzer. Testing was limited to the arcuate area to shorten the test time (approximately 15 minutes) and because of previous reports of its importance in glaucoma diagnosis 2 (Fig. 2).
1 0 -2 -5 -5 -4-1 -3 -3 -3 -6 -5 -4 -5 0 -5 -3 -i -i -19-26-32-33 -1 -4 -6-20-17-33 -2 -8 -5-16 -5 -7 -6 -5 -3
-3 -6 -3 -1 -4 -3 -4 -2 -1 0 25--28 -9 -13 -4 -9 -5 -7 -4 -6
-2 -1 -1 2-3 0 1-3 -7-8 -6 -7 -6 -4-2 -4
Fig. 1 (Nutaitis and associates). A right-eye visual field using the program 30-2 of the Humphrey Field Analyzer from a patient with glaucoma showing inferior arcuate changes typical for glaucoma on the depth defect printout. Locations are spaced 6 degrees apart and blind spot values were not included.
98* 67 83 77 84* 72 76 85* 71 64 82* 69 73 71 98* 73 72 95* 83* 68 56 64
September, 1992
78
77 88* 91* 73 85* 63 68 8 0 * 63
Fig. 2 (Nutaitis and associates). Visual field of the same eye as in Figure 1 using the pattern discrimina tion perimeter with threshold levels within the arcu ate area. Asterisks indicate abnormally depressed locations and show a stippled pattern typical for patients with glaucoma undergoing pattern discrimi nation testing. Locations are spaced 6 degrees apart and the blind spot was not tested. The pattern discrimination perimeter tests the visual field by presenting a square checker board stimulus of alternating black and white pixels on a background of randomly arranged black and white pixels. Stimulus recognition is made more difficult by decreasing, by an arith metic progression, the percentage of the stimu lus that is composed of the alternating black and white pixels and increasing the percentage of the stimulus that has the randomly arranged pixel background (spatial coherence). The perimeter determines in a staircase fash ion the lowest spatial coherence the patient is able to discern at each location (pattern dis crimination threshold) (Fig. 3). For this study, the initial stimulus was presented at 60% spa tial coherence. If the subject perceived the ini tial stimulus, the spatial coherence was reduced in 4% steps until the stimulus was not visual ized. The spatial coherence then was increased by 2% steps until observed again. The spatial coherence then was reduced by 1% steps until the subject failed again to see the stimulus, which determined the pattern discrimination threshold. In contrast, if a subject failed to visualize the initial stimulus, then the inverse progression of equal percentage steps was per formed to determine threshold. In this study, each stimulus was presented for 0.6 second and the time interval used between stimulus presentations was 0.3 second. The size of the stimulus was 20 pixels horizontally and 20 pixels vertically. However, the pixel size depended on the distance from fixation. Within 5 degrees from fixation, the pixel size was 7.5
Vol. 114, No. 3
100%
Pattern Discrimination Perimetry
80%
299
40%
60%
Fig. 3 (Nutaitis and associates). Schematic of the pattern discrimination stimulus showing a decreasing percentage of spatial coherence (left to right) of nonrandom black and white pixels projected on a randomly arranged black and white pixel background. The stimulus is located to the upper right of the black central fixation target. minutes; between 5 degrees and 13 degrees from fixation, the pixel size was 15 minutes, whereas further than 13 degrees from fixation, the pixel size was 22.5 minutes. The back ground pattern of nonrandom pixels was updat ed four times per second. These variables were chosen on the basis of findings from our labora tory (Nutaitis and associates, unpublished data, 1991), which indicated these variables provid ed the most reproducible visual stimulus at the lowest spatial coherence. Each visual field was interpreted indepen dently by two of us (M.J.N., W.C.S.). Also, the pattern discrimination field and the program 30-2 field were interpreted independently from each other. After each visual field was inter preted, it was categorized as being within nor mal limits or having either a specific (probably a glaucoma-related change) or nonspecific (not definitely a pathologic change) visual field de fect. Criteria used in this study for interpreting the program 30-2 visual fields have been described previously 3,4 and reflect the global diagnoses commonly used for automated perimetry. Cri teria used in this study for interpreting the pattern discrimination visual fields were based on previous data that individual locations with in the central 30 degrees have an equal diagnos tic value in determining a defect. 1 The number of abnormal locations used to define an abnor mality with the pattern discrimination perime ter was adjusted so the number of specific and nonspecific defects was similar to the program 30-2 visual field within the control population. This definition stipulated that eight or more locations defined a specific defect and six to seven abnormal locations defined a nonspecific defect with the pattern discrimination perime
ter. This definition provided a specificity of 93% in excluding a probable glaucomatous change, which is similar to past information regarding the specificity of the Humphrey Field Analyzer. 5 An abnormal location itself was defined as having a pattern discrimination threshold greater than two standard deviations from the mean threshold of the control popula tion at each tested location. After each visual field was defined (normal, nonspecific defect, or specific defect), the pro gram 30-2 and pattern discrimination visual fields were compared as to the severity of the defects by using the Wilcoxon signed rank test separately for the glaucoma, ocular hyperten sion, and control groups. If a significant differ ence in diagnoses was observed, further statis tical evaluation was performed to attempt to explain the diversity. This evaluation included Goldmann applanation tonometry and patient age analyzed by a Student's f-test, and pupil size, refraction, number of glaucoma medi cines, cup/disk ratio, race, and existence of vascular disease analyzed by a chi-square test. We also evaluated the similarity of the defect location between the two perimeters. To do this we determined which hemifield (superior or inferior) showed the greatest sensitivity loss for each patient for both perimeters separately. The results from both perimeters then were compared for each eye tested.
Results We included in this study 93 eyes (39 left eyes and 54 right eyes) of 93 consecutive patients with glaucoma or ocular hypertension who
300
AMERICAN JOURNAL OF OPHTHALMOLOGY
were examined at the Medical University of South Carolina. The patients had an average age of 66.8 ± 14 years. Thirty control patients (15 right and 15 left randomly tested eyes) who had an average age of 32.4 ± 10 years were included in this study. Two subjects had a specific defect and two had a nonspecific defect on both the program 30-2 and the pattern discrimination perimeter, which was not a statistically significant differ ence (P > .05). The diagnosis agreed between the two perim eters in 55 of the 73 patients (75.3%; 73 eyes) with glaucoma. Five of 73 eyes (6.8%) showed a more severe defect with the pattern discrimina tion perimeter and 13 of 73 eyes (17.8%) showed a more severe defect with the program 30-2. This was not a statistically significant difference (P > .05). The diagnosis was normal on both the pattern discrimination perimeter and the program 30-2 in seven of 20 patients (35.0%; 20 eyes) with ocular hypertension. Both programs showed a specific defect in one eye (5%). However, in three eyes (15%), the pattern discrimination perimeter visual fields were normal, whereas the program 30-2 showed either a nonspecific (two eyes) or a specific (one eye) defect. In the other nine eyes (45.0%), the pattern discrimina tion perimeter showed a specific defect, where as the visual field of the program 30-2 was normal. The difference in diagnoses between perimeters was statistically significant (P < .05). The diversity in diagnoses between pa tients with ocular hypertension who showed a specific defect and those who were normal on pattern discrimination perimetry was not ex plained by differences in the intraocular pres sure, age, refraction, pupil size, optic nerve, race, presence of vascular disease, or number of glaucoma medicines (P > .05). In comparing defect location between perim eters, 36 eyes (49.4%) differed in the hemifield (inferior or superior) with the greatest sensitivi ty loss (Figs. 1 and 2). Most defects were ob served in the superior hemifield with both pe rimeters.
Discussion The diagnosis of glaucoma depends on ob serving an increased intraocular pressure asso ciated with typical glaucomatous-appearing changes in the optic nerve head and on visual function testing. The most common visual func
September, 1992
tion test used to diagnose and monitor patients with glaucoma is automated perimetry, which projects a uniform white stimulus on a white background to determine a differential light threshold. 2 However, automated perimetry does not detect early glaucomatous damage. 6 Electrophysiologic studies have shown that in dividual ganglion cells can respond to light intensities close to normal differential light threshold levels, 78 probably because retinal ganglion cell receptive fields overlap extensive ly. Consequently, many ganglion cells must be lost before a scotoma appears that can be de tected by changes in the differential light threshold. A stimulus that necessitates form recogni tion, however, might require a greater percent age of intact ganglion cells and may be more sensitive to glaucomatous damage than stimu lus recognition based on light sensitivity. 1 The pattern discrimination perimeter is based òn the theory that visual perception by the retina is coded by the weighted responses of all gan glion cells whose receptive fields cover that location. Consequently, early ganglion cell loss might disturb the coding of the relative position of perceived objects before changes in differen tial light threshold become apparent. 1 Compared with automated perimetry, pat tern discrimination testing at individual loca tions within the central 30 degrees tended to be more sensitive in diagnosing ocular hyperten sion and equally sensitive in diagnosing glau coma.1·9 We compared pattern discrimination perime try and conventional automated perimetry in patients with glaucoma and ocular hyperten sion by using global diagnostic criteria that might be applied clinically. This study showed that in patients with ocu lar hypertension, a statistically greater number of abnormal tests were identified with the pat tern discrimination perimeter than the program 30-2. One half of the patients with ocular hy pertension showed numerous (greater than 11) abnormal locations on pattern discrimination perimetry while the other half showed few if any defects. Few defects were noted with the program 30-2. The difference in the pattern discrimination test results between groups of normal and abnormal patients was not ex plained on the basis of intraocular pressure, pupillary dilation, refraction, age, number of glaucoma medicines, race, cup/disk ratio, fami ly history of glaucoma, or the presence of vas cular disease. These results are consistent with the theory that the pattern discrimination stim-
Vol. 114, No. 3
Pattern Discrimination Perimetry
ulus requires a greater number of intact gangli on cells to be observed, and potentially may show visual function damage earlier, than auto mated perimetry. 1 In this study, the separation of patients with glaucoma from normal individuals was similar using both the pattern discrimination perimeter and the program 30-2. However, the involved area of the visual field on the program 30-2 differed from that observed on pattern discrimi nation testing in half of the patients. The rea son for this difference was not explained by our results, but is consistent with the hypothesis that the pattern discrimination perimeter may measure a different aspect of a visual system than automated perimetry. 1 Previous information has indicated that no location more than another signifies a glaucomatous defect in pattern discrimination testing. Consequently, we based the severity of the diagnosis only on the number of locations. This diagnostic approach differs from conventional automated perimetry testing, in which abnor mal locations within the arcuate area and adja cent abnormal locations more strongly indicate a glaucomatous defect.10·11 Although the scale we used was relative, increasing or decreasing the number of locations that indicated a defect did not improve the relative sensitivity and specificity of the pattern discrimination exami nation in our patients. We defined the number of locations that determined a defect to provide a specificity of 93%, which is similar to conven tional automated perimetry. 5 To save time on the pattern discrimination testing, we did not test the peripheral visual field, which previous studies have indicated might be useful diagnostically.1·9 Consequently, the peripheral visual field may be worthwhile for future study in pattern discrimination testing. This study suggests that the separation of patients with glaucoma from normal subjects in pattern discrimination perimetry is similar to that in conventional perimetry. However, in patients with ocular hypertension, the pattern discrimination perimeter identifies subsets of patients either with or without a visual func tion defect. These diagnostic subsets may prove useful in the future as an early diagnostic and prognostic determinant for glaucoma. This study was not designed to show a defi nite clinical importance or reproducibility of the defects found in patients with ocular hyper
301
tension on pattern discrimination testing. Long-term follow-up and further evaluation of the pattern discrimination perimeter is required to understand its potential clinical usefulness. Also, variables of the pattern discrimination perimeter (that is, pixel size, stimulus duration, background update time, and stimulus size) may be worthy of further evaluation to deter mine if more important clinical information is gained.
References 1. Drum, B. A., Severns, M., O'Leary, D. K., Massof, R. W., Quigley, H. A., Breton, M. E., and Krupin, T.: Selective loss of pattern discrimination in early glaucoma. Appi. Optics 28:1135, 1989. 2. Stewart, W. C : Clinical Practice of Glaucoma. Thorofare, New Jersey, Slack, Inc., 1990, pp. 80-83. 3. Stewart, W. C, Shields, M. B., and Ollie, A. R.: Peripheral visual field testing by automated kinetic perimetry in glaucoma. Arch. Ophthalmol. 106:202, 1988. 4. Miller, K. N., Shields, M. B., and Ollie, A. R.: Automated kinetic perimetry with two peripheral isopters in glaucoma. Arch. Ophthalmol. 107:1316, 1989. 5. Trope, G. E., and Britton, R.: A comparison of Goldmann and Humphrey automated perimetry in patients with glaucoma. Br. J. Ophthalmol. 71:489, 1987. 6. Quigley, H. A., Addicks, E. M., and Green, W. R.: Optic nerve damage in human glaucoma. III. Quantitative correlation of nerve fiber loss and visual field defect in glaucoma, ischemie neuropathy, papilledema, and toxic neuropathy. Arch. Ophthalmol. 100:135, 1982. 7. Barlow, H. B., Levick, W. R., and Yoon, M.: Responses to single quanta of light in retinal gangli on cells of the cat. Vision Res. 3(suppl.):87, 1971. 8. Kaplan, E., and Shapley, R. M.: The primate retina contains groups of ganglion cells, with high and low contrast sensitivity. Proc. Nati. Acad. Sci. U.S.A. 83:2755, 1986. 9. Drum, B., Severns, M., O'Leary, D., Massof, R., Quigley, H., Breton, M„ and Krupin, T.: Pattern discrimination and light detection test. Different types of glaucomatous damage. In Heijl, A. (ed.): Perimetry Update 1988/89. Amsterdam, Kugler and Ghedini Publications, 1989, pp. 341-347. 10. Harrington, D. O.: The Bjerrum scotoma. Trans. Am. Ophthalmol. Soc. 62:324, 1964. 11. : Differential diagnosis of the arcuate scotoma. Invest. Ophthalmol. 8:96, 1969.
We compared the results of the pattern discrimination perimeter to the program 30-2 on the Humphrey Field Analyzer (Humphrey, Inc., San Leandro, California) in 93 consecutive patients with ocular hypertension and glaucoma and 30 control patients. In 20 patients with ocular hypertension, a significantly greater number of glaucomatous defects were noted on pattern discrimination perimetry (ten patients) than on the program 30-2 (two patients) (P < .05, Wilcoxon signed rank test). The diversity in diagnoses found on pattern discrimination testing was not explained by age, intraocular pressure, refraction, number of glaucoma medicines, race, presence of vascular disease, optic disk status, or pupil size. In contrast, in 73 patients with glaucoma no statistical difference in the severity of diagnoses was noted between perimeters (P > .05, Wilcoxon signed rank test). These results suggest the potential value of pattern discrimination perimetry as a visual function test in patients with glaucoma and in defining subsets of patients with ocular hypertension not found with conventional automated perimetry.
Accepted for publication June 3, 1992. From the Glaucoma Service at the Department of Ophthalmology (Drs. Nutaitis and Stewart, and Ms. Kelley), and the Department of Biostatistics, Epidemiol ogy and System Sciences (Dr. Hunt), Medical University of South Carolina, Charleston, South Carolina; and the Wilmer Eye Institute (Dr. Severns), Johns Hopkins Uni versity School of Medicine, Baltimore, Maryland. This study was supported in part by unrestricted grants from Research to Prevent Blindness, Inc., New York, New York. Dr. Severns is an employee of LKC Technologies, Inc., Gaithersburg, Maryland. The opinions expressed herein are those of the authors and are not official opinions of the Department of the Navy. Reprint requests to William C. Stewart, M.D., Medical University of South Carolina, Storm Eye Institute, 171 Ashley Ave., Charleston, SC 29425-2236.
V I S U A L FIELD TESTING is performed in patients
with increased intraocular pressure to deter mine the amount of damage to the visual sys tem. Most perimeters currently available rely on a uniform white stimulus projected on a white background to assess visual function. Unfortunately, this stimulus detects glaucoma tous changes only after extensive damage to the optic nerve head has already been sustained. 1 The pattern discrimination perimeter, which recently has been described, requires recogni tion of a pattern stimulus, which may have the advantage of earlier detection of glaucomatous damage than conventional automated perime try. 1 We evaluated the use of pattern discrimi nation perimetry compared to automated pe rimetry in routine testing of patients with ocular hypertension or glaucoma.
Subjects and Methods We included in this study consecutively test ed patients with glaucoma and ocular hyper tension who were examined with automated perimetry at the Storm Eye Institute at the Medical University of South Carolina. Patients with glaucoma were defined as having typical glaucomatous-appearing optic disk damage and a history of an intraocular pressure greater than 21 mm Hg with or without visual field damage. Patients with ocular hypertension were defined as having a history of an increased intraocular pressure greater than 21 mm Hg without clinical signs of optic disk damage. We excluded patients with marked neurologic and retinal disease and those who were unable to complete a visual field examination. If both eyes were tested, only one eye was randomly chosen to be included in this study. A control group of normal individuals who had one eye randomly chosen to be examined also was in cluded in this study.
©AMERICAN JOURNAL OF OPHTHALMOLOGY 114:297-301, SEPTEMBER, 1992
297
298
AMERICAN JOURNAL OF OPHTHALMOLOGY
Each subject was tested with the Humphrey Field Analyzer (Humphrey, Inc., San Leandro, California), a fully automated perimeter with a 330-mm bowl that projects a uniform white stimulus on a white background. Subjects were examined with the program 30-2, which uses a staircase technique to determine threshold at 76 locations spaced 6 degrees apart and offset symmetrically across the vertical and horizon tal midlines within the central 30 degrees (Fig. 1)· After testing with the Humphrey Field Ana lyzer, subjects also were examined with the pattern discrimination perimeter (LKC Techno logies, Inc., Gaithersburg, Maryland), de scribed previously. 1 This perimeter's computer generates a display on a 5-inch television moni tor. An optical system relays this image to the eye while providing refractive correction and eccentricity-dependent magnification such that the image appears to be on the surface of a bowl located at optical infinity. The graphics display produces black or white pixels, 640 in number horizontally and 480 vertically. The perimeter has a 30-degree field of view. A fixation target is present in the middle of the graphic display, and fixation is checked by a technician who uses a separate television monitor. Using the pattern discrimination perimeter, we tested 32 locations within the arcuate area of the visual field at the same locations tested by the program 30-2 on the Humphrey Field Analyzer. Testing was limited to the arcuate area to shorten the test time (approximately 15 minutes) and because of previous reports of its importance in glaucoma diagnosis 2 (Fig. 2).
1 0 -2 -5 -5 -4-1 -3 -3 -3 -6 -5 -4 -5 0 -5 -3 -i -i -19-26-32-33 -1 -4 -6-20-17-33 -2 -8 -5-16 -5 -7 -6 -5 -3
-3 -6 -3 -1 -4 -3 -4 -2 -1 0 25--28 -9 -13 -4 -9 -5 -7 -4 -6
-2 -1 -1 2-3 0 1-3 -7-8 -6 -7 -6 -4-2 -4
Fig. 1 (Nutaitis and associates). A right-eye visual field using the program 30-2 of the Humphrey Field Analyzer from a patient with glaucoma showing inferior arcuate changes typical for glaucoma on the depth defect printout. Locations are spaced 6 degrees apart and blind spot values were not included.
98* 67 83 77 84* 72 76 85* 71 64 82* 69 73 71 98* 73 72 95* 83* 68 56 64
September, 1992
78
77 88* 91* 73 85* 63 68 8 0 * 63
Fig. 2 (Nutaitis and associates). Visual field of the same eye as in Figure 1 using the pattern discrimina tion perimeter with threshold levels within the arcu ate area. Asterisks indicate abnormally depressed locations and show a stippled pattern typical for patients with glaucoma undergoing pattern discrimi nation testing. Locations are spaced 6 degrees apart and the blind spot was not tested. The pattern discrimination perimeter tests the visual field by presenting a square checker board stimulus of alternating black and white pixels on a background of randomly arranged black and white pixels. Stimulus recognition is made more difficult by decreasing, by an arith metic progression, the percentage of the stimu lus that is composed of the alternating black and white pixels and increasing the percentage of the stimulus that has the randomly arranged pixel background (spatial coherence). The perimeter determines in a staircase fash ion the lowest spatial coherence the patient is able to discern at each location (pattern dis crimination threshold) (Fig. 3). For this study, the initial stimulus was presented at 60% spa tial coherence. If the subject perceived the ini tial stimulus, the spatial coherence was reduced in 4% steps until the stimulus was not visual ized. The spatial coherence then was increased by 2% steps until observed again. The spatial coherence then was reduced by 1% steps until the subject failed again to see the stimulus, which determined the pattern discrimination threshold. In contrast, if a subject failed to visualize the initial stimulus, then the inverse progression of equal percentage steps was per formed to determine threshold. In this study, each stimulus was presented for 0.6 second and the time interval used between stimulus presentations was 0.3 second. The size of the stimulus was 20 pixels horizontally and 20 pixels vertically. However, the pixel size depended on the distance from fixation. Within 5 degrees from fixation, the pixel size was 7.5
Vol. 114, No. 3
100%
Pattern Discrimination Perimetry
80%
299
40%
60%
Fig. 3 (Nutaitis and associates). Schematic of the pattern discrimination stimulus showing a decreasing percentage of spatial coherence (left to right) of nonrandom black and white pixels projected on a randomly arranged black and white pixel background. The stimulus is located to the upper right of the black central fixation target. minutes; between 5 degrees and 13 degrees from fixation, the pixel size was 15 minutes, whereas further than 13 degrees from fixation, the pixel size was 22.5 minutes. The back ground pattern of nonrandom pixels was updat ed four times per second. These variables were chosen on the basis of findings from our labora tory (Nutaitis and associates, unpublished data, 1991), which indicated these variables provid ed the most reproducible visual stimulus at the lowest spatial coherence. Each visual field was interpreted indepen dently by two of us (M.J.N., W.C.S.). Also, the pattern discrimination field and the program 30-2 field were interpreted independently from each other. After each visual field was inter preted, it was categorized as being within nor mal limits or having either a specific (probably a glaucoma-related change) or nonspecific (not definitely a pathologic change) visual field de fect. Criteria used in this study for interpreting the program 30-2 visual fields have been described previously 3,4 and reflect the global diagnoses commonly used for automated perimetry. Cri teria used in this study for interpreting the pattern discrimination visual fields were based on previous data that individual locations with in the central 30 degrees have an equal diagnos tic value in determining a defect. 1 The number of abnormal locations used to define an abnor mality with the pattern discrimination perime ter was adjusted so the number of specific and nonspecific defects was similar to the program 30-2 visual field within the control population. This definition stipulated that eight or more locations defined a specific defect and six to seven abnormal locations defined a nonspecific defect with the pattern discrimination perime
ter. This definition provided a specificity of 93% in excluding a probable glaucomatous change, which is similar to past information regarding the specificity of the Humphrey Field Analyzer. 5 An abnormal location itself was defined as having a pattern discrimination threshold greater than two standard deviations from the mean threshold of the control popula tion at each tested location. After each visual field was defined (normal, nonspecific defect, or specific defect), the pro gram 30-2 and pattern discrimination visual fields were compared as to the severity of the defects by using the Wilcoxon signed rank test separately for the glaucoma, ocular hyperten sion, and control groups. If a significant differ ence in diagnoses was observed, further statis tical evaluation was performed to attempt to explain the diversity. This evaluation included Goldmann applanation tonometry and patient age analyzed by a Student's f-test, and pupil size, refraction, number of glaucoma medi cines, cup/disk ratio, race, and existence of vascular disease analyzed by a chi-square test. We also evaluated the similarity of the defect location between the two perimeters. To do this we determined which hemifield (superior or inferior) showed the greatest sensitivity loss for each patient for both perimeters separately. The results from both perimeters then were compared for each eye tested.
Results We included in this study 93 eyes (39 left eyes and 54 right eyes) of 93 consecutive patients with glaucoma or ocular hypertension who
300
AMERICAN JOURNAL OF OPHTHALMOLOGY
were examined at the Medical University of South Carolina. The patients had an average age of 66.8 ± 14 years. Thirty control patients (15 right and 15 left randomly tested eyes) who had an average age of 32.4 ± 10 years were included in this study. Two subjects had a specific defect and two had a nonspecific defect on both the program 30-2 and the pattern discrimination perimeter, which was not a statistically significant differ ence (P > .05). The diagnosis agreed between the two perim eters in 55 of the 73 patients (75.3%; 73 eyes) with glaucoma. Five of 73 eyes (6.8%) showed a more severe defect with the pattern discrimina tion perimeter and 13 of 73 eyes (17.8%) showed a more severe defect with the program 30-2. This was not a statistically significant difference (P > .05). The diagnosis was normal on both the pattern discrimination perimeter and the program 30-2 in seven of 20 patients (35.0%; 20 eyes) with ocular hypertension. Both programs showed a specific defect in one eye (5%). However, in three eyes (15%), the pattern discrimination perimeter visual fields were normal, whereas the program 30-2 showed either a nonspecific (two eyes) or a specific (one eye) defect. In the other nine eyes (45.0%), the pattern discrimina tion perimeter showed a specific defect, where as the visual field of the program 30-2 was normal. The difference in diagnoses between perimeters was statistically significant (P < .05). The diversity in diagnoses between pa tients with ocular hypertension who showed a specific defect and those who were normal on pattern discrimination perimetry was not ex plained by differences in the intraocular pres sure, age, refraction, pupil size, optic nerve, race, presence of vascular disease, or number of glaucoma medicines (P > .05). In comparing defect location between perim eters, 36 eyes (49.4%) differed in the hemifield (inferior or superior) with the greatest sensitivi ty loss (Figs. 1 and 2). Most defects were ob served in the superior hemifield with both pe rimeters.
Discussion The diagnosis of glaucoma depends on ob serving an increased intraocular pressure asso ciated with typical glaucomatous-appearing changes in the optic nerve head and on visual function testing. The most common visual func
September, 1992
tion test used to diagnose and monitor patients with glaucoma is automated perimetry, which projects a uniform white stimulus on a white background to determine a differential light threshold. 2 However, automated perimetry does not detect early glaucomatous damage. 6 Electrophysiologic studies have shown that in dividual ganglion cells can respond to light intensities close to normal differential light threshold levels, 78 probably because retinal ganglion cell receptive fields overlap extensive ly. Consequently, many ganglion cells must be lost before a scotoma appears that can be de tected by changes in the differential light threshold. A stimulus that necessitates form recogni tion, however, might require a greater percent age of intact ganglion cells and may be more sensitive to glaucomatous damage than stimu lus recognition based on light sensitivity. 1 The pattern discrimination perimeter is based òn the theory that visual perception by the retina is coded by the weighted responses of all gan glion cells whose receptive fields cover that location. Consequently, early ganglion cell loss might disturb the coding of the relative position of perceived objects before changes in differen tial light threshold become apparent. 1 Compared with automated perimetry, pat tern discrimination testing at individual loca tions within the central 30 degrees tended to be more sensitive in diagnosing ocular hyperten sion and equally sensitive in diagnosing glau coma.1·9 We compared pattern discrimination perime try and conventional automated perimetry in patients with glaucoma and ocular hyperten sion by using global diagnostic criteria that might be applied clinically. This study showed that in patients with ocu lar hypertension, a statistically greater number of abnormal tests were identified with the pat tern discrimination perimeter than the program 30-2. One half of the patients with ocular hy pertension showed numerous (greater than 11) abnormal locations on pattern discrimination perimetry while the other half showed few if any defects. Few defects were noted with the program 30-2. The difference in the pattern discrimination test results between groups of normal and abnormal patients was not ex plained on the basis of intraocular pressure, pupillary dilation, refraction, age, number of glaucoma medicines, race, cup/disk ratio, fami ly history of glaucoma, or the presence of vas cular disease. These results are consistent with the theory that the pattern discrimination stim-
Vol. 114, No. 3
Pattern Discrimination Perimetry
ulus requires a greater number of intact gangli on cells to be observed, and potentially may show visual function damage earlier, than auto mated perimetry. 1 In this study, the separation of patients with glaucoma from normal individuals was similar using both the pattern discrimination perimeter and the program 30-2. However, the involved area of the visual field on the program 30-2 differed from that observed on pattern discrimi nation testing in half of the patients. The rea son for this difference was not explained by our results, but is consistent with the hypothesis that the pattern discrimination perimeter may measure a different aspect of a visual system than automated perimetry. 1 Previous information has indicated that no location more than another signifies a glaucomatous defect in pattern discrimination testing. Consequently, we based the severity of the diagnosis only on the number of locations. This diagnostic approach differs from conventional automated perimetry testing, in which abnor mal locations within the arcuate area and adja cent abnormal locations more strongly indicate a glaucomatous defect.10·11 Although the scale we used was relative, increasing or decreasing the number of locations that indicated a defect did not improve the relative sensitivity and specificity of the pattern discrimination exami nation in our patients. We defined the number of locations that determined a defect to provide a specificity of 93%, which is similar to conven tional automated perimetry. 5 To save time on the pattern discrimination testing, we did not test the peripheral visual field, which previous studies have indicated might be useful diagnostically.1·9 Consequently, the peripheral visual field may be worthwhile for future study in pattern discrimination testing. This study suggests that the separation of patients with glaucoma from normal subjects in pattern discrimination perimetry is similar to that in conventional perimetry. However, in patients with ocular hypertension, the pattern discrimination perimeter identifies subsets of patients either with or without a visual func tion defect. These diagnostic subsets may prove useful in the future as an early diagnostic and prognostic determinant for glaucoma. This study was not designed to show a defi nite clinical importance or reproducibility of the defects found in patients with ocular hyper
301
tension on pattern discrimination testing. Long-term follow-up and further evaluation of the pattern discrimination perimeter is required to understand its potential clinical usefulness. Also, variables of the pattern discrimination perimeter (that is, pixel size, stimulus duration, background update time, and stimulus size) may be worthy of further evaluation to deter mine if more important clinical information is gained.
References 1. Drum, B. A., Severns, M., O'Leary, D. K., Massof, R. W., Quigley, H. A., Breton, M. E., and Krupin, T.: Selective loss of pattern discrimination in early glaucoma. Appi. Optics 28:1135, 1989. 2. Stewart, W. C : Clinical Practice of Glaucoma. Thorofare, New Jersey, Slack, Inc., 1990, pp. 80-83. 3. Stewart, W. C, Shields, M. B., and Ollie, A. R.: Peripheral visual field testing by automated kinetic perimetry in glaucoma. Arch. Ophthalmol. 106:202, 1988. 4. Miller, K. N., Shields, M. B., and Ollie, A. R.: Automated kinetic perimetry with two peripheral isopters in glaucoma. Arch. Ophthalmol. 107:1316, 1989. 5. Trope, G. E., and Britton, R.: A comparison of Goldmann and Humphrey automated perimetry in patients with glaucoma. Br. J. Ophthalmol. 71:489, 1987. 6. Quigley, H. A., Addicks, E. M., and Green, W. R.: Optic nerve damage in human glaucoma. III. Quantitative correlation of nerve fiber loss and visual field defect in glaucoma, ischemie neuropathy, papilledema, and toxic neuropathy. Arch. Ophthalmol. 100:135, 1982. 7. Barlow, H. B., Levick, W. R., and Yoon, M.: Responses to single quanta of light in retinal gangli on cells of the cat. Vision Res. 3(suppl.):87, 1971. 8. Kaplan, E., and Shapley, R. M.: The primate retina contains groups of ganglion cells, with high and low contrast sensitivity. Proc. Nati. Acad. Sci. U.S.A. 83:2755, 1986. 9. Drum, B., Severns, M., O'Leary, D., Massof, R., Quigley, H., Breton, M„ and Krupin, T.: Pattern discrimination and light detection test. Different types of glaucomatous damage. In Heijl, A. (ed.): Perimetry Update 1988/89. Amsterdam, Kugler and Ghedini Publications, 1989, pp. 341-347. 10. Harrington, D. O.: The Bjerrum scotoma. Trans. Am. Ophthalmol. Soc. 62:324, 1964. 11. : Differential diagnosis of the arcuate scotoma. Invest. Ophthalmol. 8:96, 1969.
