Learn Login
Question of the Week
Video Library
Live Webinars
Ophthalmology Books & Manuals
Cybersight Atlas of Eye Diseases
The Ophthalmology Minute
Nursing Education
Eye Care Equipment
Orbis Program Features
Free Online Journals
Ophthalmology Links
Frequently Asked Questions (FAQ)
Print ViewPrint this Page
Visual Field Defects : 

Essentials of Automated Perimetry

Examination Techniques  |  Essentials of Automated Perimetry  |  Topical Diagnosis  |  Color Defects  |  Positive Perversion of the Visual System  |  Functional Visual Loss

Accurate assessment of the visual field is of great importance in many neuro-ophthalmologic disorders. Historically, ophthalmologists have relied on manual techniques such as the tangent screen and Goidmann kinetic perimeter to map a patient's peripheral vision. Over the past two decades, automated perimetry has increased in popularity and is currently a frequently employed formal method to evaluate a patient's peripheral vision. Reasons for this change include a more reproducible, standardized test that is readily quantifiable and that provides the opportunity for data storage, statistical analysis, and comparability among patients and offices. In addition, because automated perimetry uses a computer-controlled test algorithm, the role of the physician or technician performing the test may be less demanding in terms of time and training required. The Humphrey perimeter is currently the most widely used automated perimeter in the United States, and example. In this chapter have been largely obtained with this instrument.

The hill of vision may be mapped by kinetic (moving stimulus) or static (stationary stimulus) methods. Kinetic perimetry uses a stimulus of a constant size and intensity, which is moved from nonseeing to seeing areas of the visual field (Fig. 17.8). Accurate detection of the boundary between nonseeing and seeing requires a sloping hill of vision in the tested area. Adjacent similar-sensitivity boundary points are connected to produce linear maps of transition zones, termed isopters. Kinetic techniques are not optimal for the examination of relatively flat areas of the visual field. Manual perimetry typically uses kinetic testing.

fig. 17.8

Figure 17.8. Kinetic perimetry uses a stimulus of uniform size and intensity, which is varied in location from nonseeing areas of the field until the patient sees the stimulus. This technique is relatively insensitive to shallow depressions in the field.

perimetry evaluates stationary locations with test objects of a constant size. It is the stimulus intensity that is varied, and the results are generally displayed in terms of threshold intensity at each location tested (Fig. 17.9). Automated techniques of stimulus presentation are particularly suited to making static measurements since the computer algorithms are relatively straightforward. Threshold is defined as the stimulus intensity that has a 50% probability of being seen at that particular location.

fig. 17.9

Figure 17.9. Static perimetry uses a uniformly sized stimulus that is presented in a single location at different stimulus intensillen to determirie threshold.

static perimetry uses a stimulus intensity that should be seen everywhere in the visual field; i.e., it is slightly above the predicted threshold value for each location. The predicted threshold values can be based on age-matched normal data or on the results of an individual's prior threshold testing. Stimulus intensity for suprathreshold tests may be constant over the entire field (Fig. 17.10) or may conform to the slope of the visual hill (threshold-related, Fig. 17.11). Static suprathreshold strategies were devised by Armaly for screening glaucoma defects. While this type of screening is rapid, it does not detect some early abnormalities such as shallow localized depressions or increased variability of responses in localized areas. Suprathreshold screening techniques may be used to make qualitative estimates of the 
visual field. Threshold measurements are required to obtain the quantitative data needed for the early diagnosis and careful follow-up of patients.

fig. 17.10

Figure 17.10. Suprathreshold static perimetry uses a uniformly intense stimulus that should be seen in most of the field.

fig. 17.11

Figure 17.11. Suprathreshold static testing uses a stimulus that is slightly brighter than expected in that location. These stimuli may be based on normal patient data or on prior threshold results for that particular patient.

Manual perimeters, such as the Goldmann and the Tubingen, are capable of quantitative threshold measurements when operated by an experienced perimetrist. Such techniques are tedious and require highly trained personnel. The greatest advantage of computerized perimetry lies in its ability to make static threshold measurements in an acceptable length of time under standardized conditions. Static threshold measurements are relatively sensitive to shallow depressions of the visual field when the stimuli are sufficiently close together. (Fig. 17.12). This sensitivity has led to a higher detection rate of early visual field defects than with manual techniques and has enhanced the ability to meaningfully compare successive visual field examinations.

fig. 17.12

Figure 17.12. Static perimetry is often more sensitive in detecting shallowing depressions of the field than kinetic perimetry.

The human visual system possesses an exceptional adaptive ability, and generally, our
ability to estimate absolute magnitude of light intensity is poor. The human visual system has, however, a remarkable aptitude to perceive contrast. Thus, it is the differential light sensitivity of stimulus against a constantly illuminated background that is measured by static perimetry.

The measurement of light in units based on the response of the eye is termed photometry, with the apostilb (asb) as the unit of measure for the luminance of a perfectly diffusing surface. Most automated perimeters use neutral-density filters, graded in decibels (dB), over a maximally emitting bulb to vary stimulus intensity. Retinal locations of reduced sensitivity require brighter stimuli to reach threshold, represented by lower decibel values. Higher decibel threshold values represent more sensitive retinal locations (Fig. 17.13). Each decibel 0.1 log unit. Thus, 10 dB equals 1 log unit or a 10-fold change in intensity, and 30 dB equals 3 log units or a 1000-fold change in intensity. The maximum bulb intensities vary; Goldmann and Octopus perimeters generate a maximum stimulus luminance (0 dB) of 1,000 asb, while the Humphrey perimeter uses a 10,000-asb bulb (0 d 13). Background luminance also varies; Goldmann and Humphrey instruments use 31.5 ash, while most Octopus models use 4 asb.

fig. 17.13

Figure 17.13. Graytone key to the Humphrey perimeter. Note that more sensitive areas of the visual field depicted as light symbols require higher neutral-density filters graded in dB to produce dimmer threshold stimuli. Note that the maximum bulb intensity with the Humphrey perimeter (dB = 0) is 10,000 asb.

Perimetry is a subjective psychophysical test requiring the patient's cooperation, effort, and communication. As in any diagnostic test, the response to a specific question has an associated error. The frequency-of-seeing curve is a useful construction to emphasize the importance of probabilistic considerations in estimating a location's threshold (Fig. 17.14). In mathematical terms, threshold is the stimulus luminance that is seen on 50% of repeated presentations.

fig. 17.14

Figure 17.14. Sample of frequency of seeing curve. Note that as stimulus intensity increases, the frequency of seeing also increases. Threshold is defined as the stimulus intensity that is detected 50% of the time. Note that false-negative and false-positive responses will limit the maximal and minimal frequency of seeing values, respectively.

A frequency-of-seeing curve is generated by testing repeatedly at a single location. For example, a location in visual space might be tested 10 times each with stimuli of 39, 37, 35, 33, 31, 29, and 27 dB, for a total of 70 questions. For each stimulus intensity, the frequency of the positive responses is plotted. Figure 17.14 indicates that these probabilities can never be 0 or 100% because of the influence of false-positive and false-negative responses, respectively.

Several recent studies have explored the characteristics of frequency-of-seeing curves in glaucoma. The slope of the curve, a measure of uncertainty in determining the threshold, is highly correlated to actual threshold or threshold deviation (threshold deviation is the deviation from age-appropriate normal values at a particular location). Thus, areas of high retinal sensitivity (normal central locations) tend to test with high reproducibility, while locations with reduced sensitivity (abnormal central locations or peripheral locations) have a more shallow slope of the frequency-of-seeing curve, which is associated with greater uncertainty (Fig. 17.15).

fig. 17.15

Figure 17.15. Influence of reduction in threshold value on frequency of seeing curve shape. Regions of the visual field with high retinal sensitivity typically have sharp frequency of seeing curves with a good estimate of true threshold. Regions or the field with a reduction of stimulus intensity characteristically produce a broadening of the frequency of seeing curve, indicating increased uncertainty in defining the threshold.

Numeric measurements of a physiologic parameter produce variation around the mean of a test result. Quantitative measure-ments are not usual with manual perimetric techniques; therefore, fluctuations in sensitivity that cause these variations are not easily recognized. In addition, there is a strong tendency for the perimetrist to make the current visual field test conform to the results of previous tests. Computerized threshold perimetry has no such built-in bias. Careful objective static measurements have uncovered fluctuations in visual field thresholds. Bebie et al. described several components of this fluctuation. These are short-term fluctuations (STFs) and the homogeneous and heterogenous components of long-term fluctuation (LTF).

STF is the variation of responses that occurs over the performance of a single test. It is calculated by measuring the sensitivity at a location several times within the context of a single testing session. Some authors consider it important to examine both local and global STF values. While double threshold determinations accurately measure global STF, it may be necessary to rethreshold a location as many as five times to accurately assess the local STF. In clinical practice, the Humphrey machine calculates global STF by thresholding 10 locations twice during the course of a given test and displays this number on the test printout.

STF is caused by a combination of the instability of the threshold being tested and the level of cooperation and attentiveness of the patient. Similar to the broadening of the frequency-of-seeing curve seen in locations with reduced sensitivity, patients with glaucomatous visual field loss have higher STF than normal subjects. These data are consistent with earlier clinical experience with manual techniques in which variable responses in localized areas were interpreted as early manifestations of glaucomatous damage.

LTF, the fluctuation between tests, occurs over days, months, or years. By definition, LTF excludes learning effect and STF. Homogeneous LTF refers to a unidirectional change in sensitivity throughout the entire visual field and is typically about 1 dB in normal eyes. Heterogeneous LTF refers to different directions and amounts of change in sensitivity at different visual field locations. Heterogeneous LTF varies according to location and presence of disease, with a typical normal value of 1.5 dB. The LTF at a single test location increases by approximately 0.2 dB for each 1 dB decrease in sensitivity. As a rough guideline, a location with a 10-dB defect may fluctuate by as much as 10 dB without reaching the 95% confidence interval for a change. The magnitude of LT11 is usually greater than that of STF and is not routinely presented to the examiner numerically.

The causes of LTF are not yet fully established. In normal patients, LTF may increase with age and increases as the intertest time interval increases. In a group of clinically stable glaucoma patients, LTF was correlated with initial sensitivity (Fig. 17.16) and with distance from fixation. Although STF is weakly correlated with LTF, the relationship is not strong enough to accurately predict LTF from STF in an individual patient. Knowledge of the magnitude of LT11 in an individual is necessary for comparisons of visual fields for change over time. This represents a limitation in the current ability to detect subtle visual field changes.

fig. 17.16

Figure 17.16. The relationship of initial sensitivity to long-term fluctuation. Note that areas with high initial sensitivities have fairly narrow 95% confidence intervals for subsequent tests. As initial sensitivity falls, the long-term fluctuation increases. With initial sensitivity below 10 dB, the 5th percentile value approaches 0.

A basic understanding of the test algorithm used by automated perimeters employing a full threshold strategy is essential to interpreting the results and troubleshooting problematic fields.

Bracketing Strategy

Since the visual threshold of a given point in the retina is the luminance at which 50% of the presented stimuli are perceived, a patient undergoing a threshold examination may see only half-of the presented stimuli. This can be a source of frustration to patients, who may feel they have performed poorly. Understariding the phenomenon requires knowing the bracketing strategy used to make threshold measurements at each location of the visual field. Commonly used algorithms to estimate threshold employ a double crossing of the threshold (Fig. 17.17). For instance, if the initial stimulus is subthreshold (not seen), intensity is increased in 4-dB steps until the patient responds with a "yes" (seen). The stimulus intensity is then decreased in 2-dB steps until the patient does not respond (not seen). The visual threshold is thereby crossed twice.

fig. 17.17

Figure 17.17. Graphic depiction of the 4-2 double crossing test algorithm. o, a nonseen stimulus; x, a seen stimulus. Because question number 1 is subthreshold (not seen), the stimulus is increased in intensity by 4 dB. In this case, stimulus mini. ber 2 is also not seen, and with an additional 4 dB in intensity, stimulus 3 is seen. The algorithm then lowers stimulus number 4 by 2 dB which, in this case, is also seen. Stimulus number 5 is further reduced by 2 dB and is not seen. For the Humphrey perimeter, the last-seen stimulus (number 4) is taken as threshold.

If the initial stimulus is suprathreshold, stimulus intensity is decreased by 4-dB steps until the threshold is crossed, then increased in 2-dB steps (the threshold again is doubly crossed). With this strategy, accurate threshold estimates are achieved by presenting, on average, approximately five stimuli per test location. Stimulus presentations are not performed sequentially at a single location but are moved randomly throughout the entire visual field. This discourages "cheating," since the patient does not know where to expect the next stimulus presentation.

Foveal Sensitivity

Measurement of the foveal sensitivity is an option that, if selected, occurs at the very beginning of the test. This option should generally be left on, as it takes very few stimulus presentations and provides information about the most valuable portion of the visual field. The patient is asked to maintain gaze on an illuminated diamond that is projected inferior to the standard central fixation target used throughout the remainder of the test. The initial stimulus intensity is 30 dB, and the regular 4-2 bracketing strategy is used to determine foveal sensitivity. Once this portion of the test is completed, the fixation diamond is removed, and the patient is asked to fixate on the central target.

Seed Locations

Important time savers are used to reduce the number of stimuli necessary to estimate the threshold level and thereby shorten the test. Starting points for threshold determinations usually depend on results from already thresholded primary locations. This is because a location's threshold result is statistically correlated with its neighboring location's threshold value. In practice, the Humphrey machine initially tests four "seed locations," one per quadrant, located 9° from the horizontal and vertical meridians. The initial stimulus intensity at these four seed locations is 25 dB, and the full 4-2 strategy is used. Each location is thresholded twice, and the results from these four seed locations are used to determine the starting stimuli in adjacent areas (Fig. 17.18). Threshold results from these adjacent areas are, in turn, used to determine their neighbor's starting,locations, until the entire test is completed.

fig. 17.18

Figure 17.18. Seed and short-term fluctuation map superimposed on 24-2 grid. Early in the test strategy, the Humphrey perimeter tests 4 seed locations (circles), each located 9° from the horizontal and vertical meridians. These locations are each tested twice and used to determine starting values for surrounding locations. Throughout the remainder of the test, an additional 6 locations (squares) are intentionally tested twice, and these 10 locations are used to calculate the global short-term fluctuation. Throughout the remainder of the test, additional locations that deviate unexpectedly from normal values are also thresholded twice, with the results presented in parentheses.

Not infrequently, patients are attentive during the initial seed-location threshold determination process and then rapidly fatigue with deficient further responses. This results in a characteristic clover leaf—shaped field.

Catch Trials

Throughout the performance of the test, patient fixation and level of alertness is periodically assessed. The Humphrey perimeter estimates fixation with the Heijl-Krakau technique of projecting a stimulus in the anticipated blind spot location. If the blind spot checks are not seen, fixation is assumed to be central, which is not necessarily the case. A high number of fixation losses may result from wandering fixation, but they may also result from a displaced blind spot or from many false-positive responses. High plus lenses tend to shift the blind spot toward fixation, while myopic correction moves the blind spot peripherally.

Another technique to reduce the percentage of fixation losses is to instruct the technician to set the machine to replot the blind spot if high fixation losses are detected early in the test and seem to have an optical cause. The machine then executes a short subprogram that presents densely packed stimuli in the region of the expected blind spot until the actual blind spot is mapped.

The technician's description of patient fixation is also extremely valuable in detecting pseudo—loss of fixation. The absence of a low patient reliability message should not lull the examiner into a false sense of security. Consider the patient who falls asleep at the machine. Clearly, that patient is unlikely to respond to stimuli presented in the blind spot, despite poor fixation.

False-positive errors are tested by periodically withholding a stimulus presentation, although the faint noise that usually accompanies a stimulus projection is created. False-positive responses tend to indicate anxious, "trigger-happy" patients. The rate of false-positive responses can often be improved if the perimetrist coaches patients to respond only when they are certain that they have seen the stimulus. "

False-negative errors are tested by projecting a 9-dB suprathreshold stimulus in a region already thresholded. Failing to respond to this markedly suprathreshold stimulus indicates patient fatigue. False-negative errors are less influenced by coaching; however, the perimetrist should ensure that the patient is awake and consider giving the patient a short break. False-negative errors are produced both by patient inattentiveness and by a diseased, easily fatigued visual system. If the threshold results are markedly reduced, the machine may not be able to generate suprathreshold stimuli. This may lull the inexperienced perimetrist into a false sense of patient responsiveness.

Single Test Printout

The single field printout from Programs 30-2 and 24-2 of the Humphrey perimeter contains a large amount of data, with various analyses presented in multiple ways (Fig. 17.19). The inexperienced examiner may find the printout overwhelming at first, but familiarity with the overall organization and the derivation of the plots and indices will greatly ease interpretation. The printout can be conveniently divided into six sections; (a) general information, located at the top; (b) reliability, located in th second row, at the left; (c) raw data, located in the second row, at the right; (d) total deviation, a plot located in the third row, at the left; (e) pattern deviation, a plot located in the third row, in the middle; and, (f) global indices, located in the third row, at the right. At the bottom of the printout is a general legend that explains the graytone and probability symbols.

fig. 17.19

Figure 17.19. A sample lest printout of the Humphrey perimeter. The printout can conveniently be divided into six locations, (a) At the top, general information; (b) middle left, reliability data; (c) middle right, raw and graytone data; (d) lower left, total deviation plot with accompanying probability symbols; (e) middle bottom, pattern deviation plot with accompanying probability symbols; and (f) lower right, general indices. To create the pattern deviation value, the seventh-best total deviation value is used to adjust the entire total deviation plot. In this case, most locations are corrected by 2 dB, producing a slightly less significant pattern deviation plot. The general indices include mean deviation (MD), pattern standard deviation (PSD), short-term fluctuation (SF), and corrected pattern standard deviation (CPSD).


Positioned at the top of the printout, the general information section displays important data about the individual patient as well as particular test variables. Included here is the program name (e.g., Central 24-2), patient name, patient birth date, stimulus size, background illumination, blind spot check size, threshold strategy, fixation target type, patient identification number, test time and date, optical correction, pupil diameter, and Snellen acuity. Many of these variables can significantly affect the raw or calculated data, and they can be invaluable in interpreting results. For example, a miotic pupil or incorrect refraction can reduce threshold values, while an incorrectly entered birth date will create wrong age-compared deviations.


The reliability data, located below and to the left of the general information section, indicate which eye (left or right) was tested and displays a calculated patient age. The number of questions asked typically averages five per test location. The number of fixation losses estimates how stably the patient main-tained gaze at the fixation target. The number of false-positive errors aims to identify the "trigger-happy" patient, while the number of false-negative errors indicates patient fatigue. The final items of information displayed with the reliability data are test duration and the optional selection, foveal sensitivity.


The two largest plots on the printout, located to the right of the reliability indices, are the raw data printouts. These display, in numeric and gray tone format, the actual threshold values in decibels. The same 10 locations are always measured twice and are used to calculate the STF of the test session (Fig. 17.18). The duplicate values are dis-played in parentheses in the numeric print-out and are averaged for the graytone display. The testing algorithm also rethresholds locations where the initially obtained values deviated greatly from the age-matched normal data (these doubly determined values are not included in the STF calculation). The graytone plot is extrapolated from the numeric plot, and although it implies uniform sampling of the 30° field, in reality, less than 1% of this area is actually tested. The gray-tone plot remains useful to alert the examiner to problem areas and is an effective way of showing visual field results to the patient.


Since the introduction of the Humphrey perimeter, the manufacturer has upgraded the machine with increasingly sophisticated statistical analysis packages. The first of these, Statpac, allows comparison of the raw threshold data with age-matched normal values at each location: accompanying probability symbols indicate the significance of any abnormality. This plot is termed the total deviation plot. These plots are displayed in the lower left portion of the printout, both numerically and with probability symbols. The p values take into account the wider range of normal values as the distance from fixation increases. For example, a value reduced 14 dB in sensitivity, compared with the age-matched normal value that is located 27° superior to fixation, may be only marginally significant (p < 5%), while the same 14-dB deviation located 10° inferotemporal to fixation may be highly significant (p < 0.5%). The examiner should always keep in mind that statistical significance does not always mean clinical significance.


Patterns of visual field loss can be conveniently divided into generalized depression, which uniformly affects the entire field by a similar amount, and localized ("scotomatous") loss, which is frequently more diagnostic. Generalized depression is most commonly caused by cataract and can hide underlying scotomatous loss. Statpac modifies the total deviation plots in an attempt to display any superimposed patterns of localized loss hidden under generalized depression. This is done by correcting the seventh-best deviation value within the Program 24-2 test grid to zero deviation and "adjusting" the entire field by that value (Fig. 17.19). The resulting plot is termed the pattern deviation plot. The threshold value of this seventh most elevated location, has been termed the general height value, although it is not routinely displayed on the printout. For example, if the entire visual field is diffusely reduced in sensitivity by 7 dB, with an underlying moderate superior arcuate defect of an additional 12 dB, the deviation plot will be reduced by 7 dB and displayed in the area labeled pattern deviation, and the 12-dB relative depression will become more apparent. A probability analysis is again displayed on these adjusted deviation values.


In the lower right corner of the single field printout, Statpac displays the global indices, which describe the entire visual field in four numeric values: (a) MD, mean deviation, is a location-weighted mean of the values in the total deviation plot. It is essentially a distilled value that represents the average height of the entire hill of vision. Negative values represent depression. MD is relatively insensitive to localized defects and is strongly affected by generalized trends; (b) PSD, pattern standard deviation, represents the unevenness of the surface of the hill of vision. It is calculated by taking a location-weighted standard deviation of all the threshold values. PSD is insensitive to the overall average height and is strongly affected by localized defects; (c) STF, short-term fluctuation, is the standard deviation of the 10 doubly thresholded locations. STF increases in inconsistent patients. This increase may be due to poor patient cooperation or attention, but STF also tends to Increase in scotomatous areas, particularly at their borders; (d) CPSD, corrected pattern standard deviation, calculated because STF influences PSD. This attempts to better represent the unevenness of the surface of the hill of vision by accounting for the influence of STF. Statpac provides probability values for each global index value, compared with age-matched normals. For example, if the MD value is accompanied by a p< 2% symbol, the MD of the field is depressed by an amount greater than that found in 98% of the same age normal population.

Glaucoma Hemifield Test

Statpac 2, Humphrey's more recent statistical upgrade of the single field analysis printout, introduced an additional statistical analysis titled the glaucoma hemifield test (GHT). The software produces the GHT result by dividing each of the upper and lower halves of the field into five mirror-imaged zones. Each zone is subsequently scored ac-cording to its pattern deviation values, and each upper zone is then compared with the corresponding lower zone. In addition, a general height of the field is determined by analyzing the most normal region of the field. The GHT has not been specifically validated in neuro-ophthalmic patients, but several of the responses may be useful. These include the result WNL, which is usually a reliable indicator of a normal field. Abnormally high sensitivity indicates a "trigger-happy" patient. General reduction in sensitivity indicates uniform field loss. The other two possible GHT descriptions, borderline or abnormal, may result from asymmetric loss across the horizontal meridian. While these descriptions more typically result from glaucomatous visual field defects, they can be seen with any nerve fiber bundle defects.

3-6 Alternative Tests

Increasing the number of test locations and the precision with which they are tested does not necessarily provide a more accurate picture of the visual field. Lengthy tests become fatiguing to the patient and may result in greater variability of responses. A number of strategies have been devised in an attempt to shorten the test and reduce the number of tested points while still providing an accurate representation of the visual field. These strategies are described in the sections that follow. Multiple other tests available with the Humphrey perimeter (e.g., Temporal Crescent, Neurologic NO, Neurologic 50, Program 24-1, and Program 30/60-1) are rarely used.

Grid Size

The standard Humphrey Program 30-2, one of the more commonly used tests, samples 76 locations with a uniform 6° grid extending to 27° from fixation (Fig. 17.20). All Humphrey programs ending in -2 (e.g., 30-2, 24-2, 10-2) are offset from the horizontal and vertical meridians. Programs 30-2 and 24-2, which use 6° spacing, are thus offset by 3°.

fig. 17.20

Figure 17.20. Program 24-2 and 30-2 test grids. Program 24-2 deletes the outer row of the 30-2 test grid with the exception of the two nasal locations. This produces a 30% reduction in test time. Because the outer row of locations are typically least reliable, this is often an attractive tradeoff between test speed and the ability to detect disease.

To threshold every location, generate reliability indices, measure STF, and rethreshold unexpected values, approximately 550 questions are asked in a typical test, which takes about 15 minutes per eye. As the distance from fixation increases, the normal threshold values decrease, with a corresponding increase in the intratest and test-retest variabilities, providing diminishing returns. One ap-proach to shortening the test is to delete the outer row of locations. Program 24-2 only tests out to 21°, except for preserving the important-nasal extent of Program 30-2. The resulting test contains 54 locations, a 29% reduction compared with the Program 30-2 grid, and considerably shortens the test duration. This represents an attractive tradeoff in patients who fatigue with additional testing.

Fast Threshold

A further decrease in time can be realized by choosing the fast threshold strategy (different from Fastpac), which performs the entire bracketing process only at locations that are not seen with a predicted 2-dB suprathreshold stimulus. The stimulus intensity can be calculated from age-matched normal data or preferably from the results of the patient's prior conventional threshold tests. In this way only "abnormal" points are bracketed. Fast strategies can cause misinterpretation in the initial evaluation of patients whose visual field may be a bit supranormal.


The recently introduced Fastpac uses an entirely different testing strategy. Instead of the standard 4-2 full strategy with a double crossing of threshold, Fastpac adjusts the stimulus intensity by 3-dB increments until the threshold is crossed once. Fastpac saves up to 40% of test time in normal or near-normal fields and provides less advantage in patients with larger amounts of field abnormality. The time savings is accompanied by a small reduction in the estimate of scotoma extent and severity and higher STF. In comparisons of Fastpac and full threshold strategies, Fastpac shortens test time by 35 to 40%. Fastpac may be most suitable for following up reliable patients with previously near-normal results, although few longitudinal data are available on its use in these populations, and the effect on LTF is presently unknown. One potential advantage of Fast-pac over the fast threshold program is the ability to use the Statpac 2's programs that analyze for change over time.

Programs 30-1 and 24-1

Programs 30-1 and 24-1 test a uniform 6° grid out to 30° and 24°, respectively, but are not offset from the horizontal and vertical meridians. Because scotomas centered on the meridians are difficult to classify (superior vs. inferior, nasal vs. temporal), these locations have less diagnostic, localizing value. Therefore, these programs are infrequently used.

Program 10-2 and Macula Test

Program 10-2 provides a high-resolution test of the central 10°, with a tight 2° grid, offset 1° from the meridians (Fig. 17.21A). A total of 68 locations are used. Central tests are useful in carefully defining central or paracentral scotomas and are more sensitive in detecting subtle progression within the central visual field. In patients with ad. vanced damage and small remaining central islands of vision, Program 10-2 can be performed with stimulus size V. This strategy provides the advantage of testing more areas with measurable threshold and seems to in. crease patient cooperation and reduce patient fatigue.

An even more localized test is the macula test, which thresholds 16 locations within the central 5°. Each location is thresholcled three times to provide better estimates of local STF (Fig. 17.21B).

fig. 17.21

Figure 17.21. A. Program 10-2. This grid uses 2° spacing offset 1° from the meridians and tests out to 9°. This test is useful to better define dense paracentral defects as well as small remaining central islands of Vision. B. The macular threshold test thresholds the central 16 locations of the 10-2 test grid. Each location is tested 3 times to provide a better estimate of fluctuation. This test is useful in measuring small central and paracentral scotomas.

Program 30/60-2

Additional programs allow exploration or the peripheral visual field (beyond 30°), Prograrri 30/60-2 extends the test out to 60°, with a uniform grid testing 68 additional locations.

Nasal Step Program

Patients with possible nasal steps can be further explored with the nasal step program, which tests 12 locations beyond 30° nasally. Two locations in the temporal visual field are also included to reduce the predictability of the questions to the patients.

Stimulus Size Option

Most programs are performed with stimulus size III, which subtends a diameter of 0.43° in visual space. This size is derived from the 4 mm2 size 3 used by the Goldmann perimeter. Testing with size III allows application of the Statpac 2 sophisticated statistical analysis. In fields where most test locations are markedly reduced, it is often preferable to increase the stimulus size to V (1.72° in diameter). This larger size is often preferred by the patient and may reduce fluctuation, although the option of using the glaucoma change analysis is lost.

Follow-up Printout

When evaluating a series of automated fields performed over time, the perimetrist may find the integration of the massive amount of data overwhelming. The Humphrey instrument allows the creation of several serial printouts to ease confusion and allow statistical analysis of change over time.

Overview Printout

The overview printout (Fig. 17.22) simply presents a sequential listing of condensed single field printouts chronologically. For each test session, four plots are displayed (from left to right): graytone threshold plot, numeric threshold plot, total deviation p value plot, and pattern deviation p value plot. Above these four plots are listed the glaucoma hemifield test results, reliability data, pupil size, and Snellen acuity. Below the plots are listed foveal sensitivity and global indices. The viewer familiar with the single field analysis printout will have no difficulty understanding terms used in the overview printout.

fig. 17.22

Figure 17.22. Overview printout in this case, three sequential 30-2 tests are presented. The plots from left to right include graytone, numeric dB, total deviation, probability, and pattern deviation probability plots. In this case, over the first three fields, a significant learning effect has occurred.

Change Analysis Printout

The change analysis printout represents each field as a box plot, depicted graphically over time (Fig. 17.23). To create this plot, deviation values at each location are ranked from least to most depressed. As can be seen from the legend at the left side of the print-out, each box contains the 15th to 85th percentile deviation values from this ranking, with a central line representing the median value. In general, a small box with long tails suggests a clumping of values with a few outliers. If a small box is near 0 dB, most of the field is normal. If the box changes in location over time with a stable size, a generalized change is likely occurring. A change in box dimension usually indicates a more localized process.

fig. 17.23

Figure 17.23. Change analysis printout. In this printout, each field is depicted as a box plot at the top of the printout. The box plots are created by ranking total deviation values and displaying the 0, 15th, 50th, 85th, and 100th percentile graphically. In this case, notice the learning effect over the first four fields. A change in box plot location with a retention in dimension indicates a generalized trend, while lengthening of tails indicates more localized processes. At the bottom of the printout, the four general indices are graphed over time. A linear regression of the MD slope is performed with an accompanying significance of value.

The change analysis printout also plots the four global indices over time, with threshold levels for statistical significance. Probably most useful is the MD value over time. If this plot is not worsening, it is likely that most of the field is not worsening, although a stable MD with an increasing CPSD may indicate early progression of a localized scotoma. If the MD value is declining, the examiner must inspect the other indices and actual fields to discern confounding developments such as cataract formation. For the plot of MD over time, a linear regression analysis, titled MD slope, is performed, which describes the slope in decibels per year and assigns a significance level.

Glaucoma Change Probability Printout

Statpac 2 introduced the glaucoma change probability printout (Fig. 17.24A), which allows the examiner to average two initial fields into a baseline field and then perform a point-by-point statistical comparison of each subsequent field, looking for significant change (Fig. 17.24B). The "expected" fluctuation values have not been validated in nonglaucoma patients.

fig. 17.24a

fig. 17.24b

Figure 17.24. Glaucoma change probability. A. The glaucoma change probability averages the a and b first two fields into a base-line field. B. Each subsequent field then undergoes a point-by-point comparison with baseline presented as a change in dB from baseline with accompanying probability symbols. These symbols were derived from the fluctuation of a population of glaucoma patients tested four times over a 1-month period. They have not been validated for other diseases.

Learning Effect

The results of many psychophysical tests improve as the subject gains more experience performing the test. The learning effect in automated perimetry seems to be small in most patients who have had experience with manual perimetry. Some patients, however, may demonstrate a dramatic improvement in the second test, compared with the first, despite previous experience with manual perimetry (Fig. 17.22). Occasionally, patients continue to improve over the initial three, four, or (rarely) five automated fields. The variability of test results decreases significantly with experience. Whenever possible, a patient new to perimetry should undergo several test sessions to establish a baseline for subsequent comparisons. The magnitude of the learning effect can be reduced by an attentive, thoughtful, operator who takes the requisite time to explain the examination thoroughly to the patient.

Interpreting a Single Test

Armed with a solid understanding of the test printout algorithm and derivation of catch trials, the examiner is better prepared to interpret a single test. Of great concern is the reliability of the particular patient. The patient with a tendency toward high false-positive errors can be thought of as trigger-happy, eager to perform well. These patients frequently respond to machine noise instead of perceived visual stimulus. The typical high false-positive printout will demonstrate physiologically supranormal sensitivity values, which will be depicted as "white scotomas" on the graytone printout. The general height adjustment of the total deviation plot to produce the pattern deviation plot may artifactually depress most of the field, producing a highly significant pattern deviation plot. The mean deviation general index is often above +2 dB, and these patients typically have high pattern standard deviation and STF (Fig. 17.25).

fig. 17.25

Figure 17.25. High false-positive patient. Note the thresholds in the 50s in peripheral areas of the field, which are nonphysiologic. The seventh-best total deViation value of +16 is used to define the general height of the field and correct it to 0. This produces a markedly abnormal-appearing pattern deviation plot. The glaucoma hemifield test has correctly identified the field as containing abnormally high sensitivity values, The general index mean deviation is extremely positive, and the field contains high pattern standard deviations and short-term fluctuation.

The patient with a tendency toward high false-negative responses can be thought of as the easily fatigued patient who gradually becomes less responsive during the test. Because the outer edges of the test grid are tested last, the easily fatigued patient will tend to produce patchy reduction in sensitivity toward the periphery of the field. Although distinguishing such peripheral depression from true defects can prove difficult, the high false-often produce a result that does not respect anatomic boundaries and is not consistent with other aspects of the ophthalmologic examination. The reader is reminded that patients with diseased visual systems are also more easily fatigued. One suggestion for confirming or disproving the absence of true peripheral defects is to test the patient with a different-sized test grid to see if the defects are reproducible in location. The influence of high false-negative responses on the printout is, in many ways, the opposite of the high false-positive patient. If the seventh-best value on the total deviation negative patient will plot is affected by fatigue, then the general height correction of the total deviation plot will create a false good-appearing pattern deviation plot. The mean deviation index will become more negative in a high false- negative patient (Fig. 17.26).

fig. 17.26

Figure 17.26. High false-negative errors due to patient fatigue. Note the patchy reduction in sensitivity near the periphery as well as the high Use-negative error ratio in the reliability section. The seventh-highest value in the total deviation plot of –3 has been corrected to 0, wIth a reduction in the abnormality of the pattern deviation plot. The MD value Is very negative, and there are high PSI), SP, and CPSD.

General Reduction in Sensitivity

One of the more common ocular conditions that can affect the visual field is cataract. Cataract formation characteristically reduces the field in a fairly uniform fashion. Moderate nuclear sclerosis, even with retention of 20/20 Snellen acuity, can reduce the visual field uniformly by several decibels. Several clues to diagnosing general reduction sensitivity include a reduced foveal sensitivity. The graytone plot is uniformly darkened, and the general height correction of the total deviation plot will produce a fairly normal-appearing pattern deviation plot. Although the mean deviation index may be statistically significant, the PSD, SF, and CPSD should be normal (Fig. 17.27).

fig. 17.27

Figure 17.27. Generalized depression due to cataract. This patient with 20/40 acuity had 3+ nuclear sclerosis. This has reduced the foveal sensitivity as well as fairly uniformly reducing the total deviation plot. The general height correction does a good job of removing the generalized depression, leaving a clean-appearing pattern deviation plot indicating absence of localized scotomas. This is also correctly identified by the glaucoma hemifield test. The only general index abnormality is a negative mean deviation. The normal PSD, SP, and CPSD also indicate absence of localized scotomas.

Localized Defect

Table 17.1 lists graded minimal criteria for defining localized loss. This table emphasizes the importance of excluding far peripheral values as well as values around the physiologic blind spot because of high fluctuation.

table 17.1