Indian Journal of Ophthalmology

: 2001  |  Volume : 49  |  Issue : 2  |  Page : 125--140

Interpreting automated perimetry

Ravi Thomas, Ronnie George 
 Schell Eye Hospital, Christian Medical College, Vellore, India

Correspondence Address:
Ravi Thomas
Schell Eye Hospital, Christian Medical College, Arni Road, Vellore - 632 001, India


Visual field testing is mandatory for many ophthalmic conditions including glaucoma. The current gold standard for visual field testing is automated perimetry. In this article we familiarize the reader with the components of an automated perimetry printout. We describe a systematic approach that leads to a thorough interpretation of the printout. With the help of examples the reader should be able to learn to identify a normal field, detect the presence of a field defect, determine whether it is due to glaucoma, and establish progression, if any.

How to cite this article:
Thomas R, George R. Interpreting automated perimetry.Indian J Ophthalmol 2001;49:125-140

How to cite this URL:
Thomas R, George R. Interpreting automated perimetry. Indian J Ophthalmol [serial online] 2001 [cited 2018 Oct 16 ];49:125-140
Available from:

Full Text

isual field assessment is useful in several ophthalmic conditions, including neuro-ophthalmic diseases. However, its value is most frequently realised in the subspeciality of glaucoma. Automated perimetry is the current gold standard in glaucoma diagnosis and management. In fact, without automated perimetry modern glaucoma management is almost impossible.

In keeping with the norms of the practice section of the journal, this article is intended as a practical guide. It is directed towards those who wish to acquire the ability to interpret perimetry results, but don't care too much for detailed explanations - and certainly not for the numbers and statistics that computers thrive on. The goal of this practice section is to familiarize beginners with automated perimeter printouts. After a reasonably careful appraisal, the reader should at least be able to understand and interpret an automated perimetry printout without much difficulty. We hope to make the process as simple as possible. We also expect this article will help achieve the following additional objectives. The reader should also be able to: (a) identify a field defect, (b) decide whether it is due to glaucoma, and (c) make a reasonable decision as to whether the defect is progressing.

We will deal with the interpretation of reports from the Humphreys Field Analyzer (HFA), since this is the instrument we use in our clinical practice and are most familiar with. We could easily title the 'Normative data' but to avoid confusing some people, we've decided to give it another heading.

 What is normal to an automated perimeter? (Normative data)

The visual threshold is the physiologic ability to detect a stimulus under defined testing conditions. The normal threshold is defined as the mean threshold in normal people in a given age group at a given location in the visual field. It is against these values that the machine compares the patient's sensitivity. For several reasons thresholds are reported in decibels, in a range of 0-50. Fifty decibels (db) is the dimmest target the perimeter can project. It is unlikely that any normal person can detect this dim a stimulus. A young and perimetrically experienced person can probably see the 40db target and that too at the fovea. On the other end of the scale, 0 db is the brightest illumination the perimeter can project. Essentially, the ability to identify a 35db target indicates a better sensitivity of the retina than a point which can identify a 25db target; and this is more sensitive than another point that can identify only a 10db target. In other words, the lower the decibel value, the lower the sensitivity; the higher the decibel value, the higher the sensitivity.

The major advantage of automated perimetry is that it compares the patient's sensitivity to stored values that have been obtained from normal people; in other words, the normative data. The thresholds for normative data are determined by an intensive bracketing strategy that we do not need to discuss here. Alternatively, it can be determined using a smarter program called SITA - the Swedish Interactive Threshold Algorithm. Whichever way they are determined, the normal thresholds are stored in the computer of the perimeter and it is against these normal thresholds that the perimeter compares the patient's data.

Obviously, if one has bought an expensive perimeter with an expensive computer, which can store normal data, the least one expects is a simple X Y Z approach to diagnosis. That is, if X number of points deviate by Y, the computer should be able to make a diagnosis of Z. We should be able to feed the patient's visual threshold sensitivity and a simple set of rules into the computer, which should then be able to make the diagnosis, complete with a "p" value to convince us that everything is "statistically" okay. The problem with such an expectation and approach lies in the way the "normal" range of threshold is determined.

How is the normal range determined? We take 1000 or 10,000 normal people (actually, more like 300), and measure their retinal sensitivity at each location usually tested in automated perimetry. We then arbitrarily decide that the upper 95% of values found in normals are normal. The lower 5% values found in these normal people are arbitrarily labeled as abnormal. Abnormal therefore has a value that is found in 5% of normal people without disease. As a corollary, 5% of normals will automatically be labeled as abnormal. This is the built-in logic of the machine. However, it is important to remember that abnormal is not the same as diseased. Whether an abnormal field is due to disease depends on the testing situation.

For example, suppose we decide to screen the general population for glaucoma. With a prevalence of glaucoma of 1%, we would expect to find 1% of the population to have glaucomatous field defects and 99% to have normal fields. If we performed automated perimetry on 100 randomly chosen people in this general population we would expect to find six abnormal fields. One field due to glaucoma and 5 persons who are labeled as abnormal because they fall in the lower 5% of normal.

Let us now consider a clinic where the ophthalmologist has an interest in glaucoma. The referral pattern from doctors aware of this interest, as well as the clinical examination will ensure that the probability of the patient having glaucoma (before the field is ordered) is higher - at least 30%. In other words 30% of the patients are likely to have glaucomatous field defects. If we test 100 patients in this clinic, we are likely to obtain 33 abnormal fields: 30 belonging to patients with glaucoma; and 3% who have been labeled as abnormal because they fall in the lower 5% of normal.

Notice how the usefulness or the "diagnostic yield" of the test has increased so dramatically with the increase in prevalence of the disease from 1 to 33 %. Most tests provide the most useful results when the prevalence of the disease is between 30-70%. The machine, however, does not know the testing situation. We know it since we have all the clinical information to interpret the field. We hope it is clear from the above that your expensive automated perimeter may have an expensive computer built into it, but it still needs your interpretation. Automated fields too need interpretation and clinical correlation and this is why no smart computer can replace human experience!

 A few hints

Before we begin to interpret fields, here are a few hints: as far as possible rely on the full threshold tests. They provide the first real evidence of glaucoma and may actually be able to predict visual field loss. Reserve "screening" for those situations where you are actually screening or in specific situations like the detection of hemianopia and in those who are so fatigued that they can't last out the time required for threshold or SITA test.

Interpreting the decibel, value is just half the challenge of interpreting the visual field printout. False positives, false negatives and fixation losses of more than a certain percentage (the printout highlights this information) may negate any comparison that the machine makes with stored normative data. Fluctuations, long term or short term, may interfere with diagnosis. The perimetry strategy (programme) one selects determines the type of information obtained. The patient's experience with the perimeter also determines what sort of information is obtained; that is, there is a learning curve involved.

The "technician" who operates the machine is supposed to properly instruct and encourage the patient; to some extent this can determine how many fields will actually be reliable. The technician can also help avoid artifacts.

Artifacts can occur with automated perimetry. [Figure:1] shows a constricted field. When we look at the optical correction that has been used, it is a very high plus lens. When the field was repeated with an appropriate contact lens, this field defect disappeared [Figure:2].

Usually we do not use the gray-scale printout for diagnosis, but fields with false-positive and false-negative errors may produce characteristic changes in the gray-scale print out. In the gray-scale of [Figure:3] multiple "white areas", so-called "white" scotomas are seen. Some of the actual decibel values are in the high 30s and 40s, even away from the centre. A perimetrically trained youngster may have a foveal threshold of 40 decibel; the thresholds we are seeing in the periphery are not physiological. The white scotomas draw your attention to high false-positive responses. The gray scale in [Figure:4] shows the typical "clover leaf" pattern fairly characteristic of a fatigue field with high false negatives. Here, initially the patient performs well and then become progressively less responsive.

Without an obvious defect that can be correlated with clinical features, it may be difficult to make a decision on the first field. In case of doubt, as with other investigations in clinical medicine, it is a good practice to repeat the field test to get more reliable information.

 Interpretation of automated perimetry

One last task before we go on to interpreting printouts: we have to change our mindset. Till recently we were used to kinetic perimetry - Bjerrum screens, Goldmann Perimeters and Isopters. In other words we were used to looking at curves. With automated perimetry we do not allow the curves (the gray scale) to disturb us; we concentrate our attention on the figures and patterns.

We would recommend that you interpret the visual field systematically, in the 8 sections shown in [Figure:5].

Zone 1: This documents patient's data. It tells us which test has been performed (in this case of 30-2) and on whom (the patient's name) It is important to look because the technician may not have done the test you ordered. In this area we also look at the type of target used, the strategy (SITA or Full threshold), whether the fixation target was central or a large diamond (as is usually done when there is a central scotoma). We also check whether the patient's birth date has been entered correctly. If not, the patient's threshold will be compared to the normals of the wrong age. We look at the correction that has been used to do the field test and make sure it was appropriate. Should a contact lens have been used? Check the pupil size. Was it at least 2.5 to 3mm in diameter and was it the same as in the patient's earlier fields? We then look at the patient's visual acuity: in this case 20/20 (or 6/6). This machine does have the option to depict vision in the metric system.

Zone 2: Having looked at the recorded vision, we suggest that you look at the bottom of Zone 2 and correlate the visual acuity with the foveal threshold. Recording the foveal threshold is optional, but it is an option that is preferably turned on. It serves as an internal validation for the visual acuity; the two should correspond. If the visual acuity is good but the foveal threshold is low, there may be early damage to the fovea. On the other hand if the foveal threshold is good and the visual acuity is low, perhaps the patient needs refraction before the visual field test is repeated.

Zone 2 also provides the reliability criteria: the machine will flag fixation losses, false-positive and false-negative errors above a certain percentage, indicating that the patient has low reliability criteria. This does not necessarily imply that the field will provide no useful information; it is just that such fields were not included in the database. Hence such fields must be interpreted with more caution.

Zone 3: Zone 3 of the grey scale is meant to give an impression of the dangerous curves we spoke of. It provides the actual perimetry results that we use for interpretation and clinical decision making. It is useful in some instances where they highlight areas to look at it in detail; it is also useful when there are gross false-positive and false-negative errors. But in general, we do not make a diagnosis based on the gray scale. So we quickly glance at zone 3 and move on to zones 4, 5, 6, 7 and 8 (actual threshold values). These form the "guts" of the machine. The truth is that Zones 4, 5, 6 and 7 provide the statistical help we have paid for.

Zone 4: This is the Total Deviation plot. It is a point by point difference of the patient's threshold from those expected in age corrected normals. In [Figure:6] the yellow line is the expected normal hill of vision (in normals of the same age as the patient). The shaded area is the depression of the patient's hill of vision. The total deviation plot highlights this deviation. In other words, the total deviation plot draws our attention to any "overall sinking" of the hill of vision. This overall sinking of the hill of is usually caused by media opacities, such as cataracts, corneal opacities, refractive error, and miosis, (rarely this can be the sole manifestation of glaucoma). The shaded area in the total deviation plot highlights an overall sinking of the patient's vision, as compared to age-related normals. It is depicted as both a numerical plot and probability plot. The probability plot predicts the chances of such an abnormality occurring in the normal population. A scale is provided for ease of interpretation: the blackest dots indicate that less than 0.5% of the normal population would be expected to have such a depression in those areas. The total deviation plot also highlights any scotoma that may be present, involving a large area of the visual field.

Generally the diagnosis of glaucoma depends on the identification of localized field loss. The total deviation plot in zone 4 highlights any overall depression of visual field of the patient compared to age related normals. However, it does not reveal any hidden scotoma that may be present in this depressed field.

In [Figure:7] the normal hill of vision is marked out in yellow. The patient's depressed hill of vision is shaded to show that it has sunken; this is the total deviation plot, but the total deviation plot does not reveal the black scotoma hidden within the depressed field. This sort of scotoma can occur in a patient with glaucoma who also has a sunken hill of vision due to cataract, miosis, or refractive error or other causes. In glaucomas, it is the localized scotomas that are diagnostic. The total deviation plot is not capable of revealing these hidden scotomas.

Zone 5. To produce this zone, the machine adjusts for overall depression of the visual field, due to cataract or some other reason. It adjusts for any overall sinking of the hill of vision and in the Pattern Deviation Plot, draws our attention to any localized scotomas that may have been hidden inside this depressed visual field. The Pattern Deviation plot is also provided as a numerical plot as well as a probability plot. We generally look at the probability plot. We try to look for abnormal points in a cluster. The low probability symbols don't matter as much as abnormal points in a cluster in an expected area.

In [Figure:5], the total deviation plot has several points that are depressed, but the pattern deviation plot is normal. We are looking at something that has caused an overall depression of the visual field, something like cataract, miosis, media opacities, or even something like the patient having a bad day. On the other hand, if there are abnormal points in the total deviation plot that persist in the pattern deviation plot, we are looking at scotomas that have been revealed after adjusting for any depression of the hill of vision.

Zone 6. This shows the global indices. What are the global indices? The statistical manipulations providing the total and pattern deviation plot are point by point calculations. In the global indices, all these points are reduced to a single value that provides us information about our patient's visual field. There are four such global indices. The Mean Deviation (MD), the Pattern Standard Deviation (PSD), the Short Term Fluctuation (SF), and the Corrected Pattern Standard Deviation (CPSD).

The Mean Deviation is derived from the total deviation plot. Like the total deviation plot the mean deviation indicates any overall depression (or elevation) of the patient's hill of vision. A positive number indicates a better than normal field (elevation of the hill of vision). A negative number indicates a depression of hill of vision. This is likely to be found in cases of media opacities such as cataract, corneal opacity, refractive error, or miosis. A large scotoma can also produce a negative mean deviation.

To use the analogy of a road we are travelling on, the total deviation plot shows how far above or below sea level that road is.

The next global index is the Pattern Standard Deviation. The pattern standard deviation is also derived from the total deviation plot; but it gives us different information. It tells us how different the numbers, in the total deviation plot, are from each other. In other words, if the Mean Deviation told us how far below sea level the road that we are travelling on was, the Pattern Standard deviation draws our attention to pot holes (meaning scotomas) on that road. It highlights any irregularity in the visual field, irrespective of any overall depression in the hill of vision.. Thus the Pattern Standard Deviation, like the Pattern Deviation plot, highlights any scotomas that may be hidden in a depressed hill of vision.

The Short-term Fluctuation is the intra-test variation in threshold. It is essentially the error in threshold determination. Threshold values at 10 predetermined points in the visual field are obtained twice. The standard deviation of these values is the short-term fluctuation. The short-term fluctuation is an indicator of reliability; but it could also be an indicator of pathology. Ten fixed points are used to determine short-term fluctuation. Diseased points have a greater variability. If any of these predetermined fixed points were pathologic, the variability would be greater. In that case the short-term fluctuationwould reflect pathology. If all these fixed points were tested in normal areas of the visual field, a high short term fluctuation would indicate low reliability.

We have said that the pattern standard deviation suggests irregularity (scotomas) in the visual fields after correcting for any depression of hill of vision. We have also said that there is an error in threshold determination reflected by the short-term fluctuation. In that case, it is entirely possible that the pattern standard deviation is affected by this error in threshold determination. In order to correct for this we have another global index, the Corrected Pattern Standard Deviation (CPSD). The CPSD draws our attention to any irregularities in the visual field (that is localized scotomas) irrespective of any overall depression due to media opacities as well as after adjusting for errors of threshold determination (as reflected by the short term fluctuation).

The double thresholding of 10 points required for testing for short-term fluctuation takes a lot of time. It is generally felt that the pain was not worth the gain. Hence the newer, faster programs like SITA do not determine short-term fluctuation. In the SITA program the short-term fluctuation and the CPSD are no longer available. The pattern standard deviation is substituted as criteria in making the diagnosis. We will discuss this in more detail later.

In Summary, Zone 5 and 6 help us make a diagnosis in the following manner. The mean deviation global index and the total deviation plot draw our attention to a sunken hill of vision or a scotoma involving a large proportion of the visual field. The pattern deviation plot and the pattern standard deviation draw our attention to any localized irregularity or scotoma in the visual field irrespective of any overall depression of the hill of vision (due to media opacities, etc). The corrected pattern standard deviation draws attention to localized scotomas in the visual field irrespective of overall depression of the hill of vision as well as after adjustment for errors in threshold determination.

Zone 7. This is a very important zone called the Glaucoma Hemifield test. In the glaucoma Hemifield Test, five sectors in the upper field are compared to five mirror images in the lower [Figure:8]. If the values between any sector in the upper, and lower zone differ to an extent found in less than 1% of the population, the glaucoma hemifield test is considered "outside normal limits".If any one pair of sectors is depressed to the extent that would be expected in less than 0.5% of the population, it is again considered "outside normal limits".

If both these conditions do not apply, but the difference between any one of the upper and lower mirror zones is what might be expected in less than 3% of the population, the glaucoma hemifield test is considered "borderline".

If the best part of the visual field is depressed to an extent expected in less than 0.5% of the population, the field test is considered to have "abnormally low sensitivity". On the other hand, if the best part of the visual field is such as would be found less than 0.5% of the population, it is considered to have "abnormally high sensitivity".

The glaucoma hemifield test is not infallible. One can see from [Figure:8] (from the way the sectors are placed), that the glaucoma hemifield test is not designed to detect a temporal wedge defect. Fortunately, such defects are rare.

Finally, even if all the zones are normal, but the clinical features are very suspicious, we would suggest that the actual threshold values in the given patient be inspected for any pattern or scotoma. These values are shown in zone 8. Remember, a scotoma, as defined, is not compared to normals, but to the surround. By concentrating on the actual threshold values, one may pick up a suggestion of a scotoma. If such defects are repeatable and correlate with the clinical picture we may elect to treat. It is also of value to look at the thresholds in the upper arcuate area compare it with the lower part in the arcuate area, and see if you can find any clinical correlation.

We would like to emphasize once again that no diagnosis should be based on the gray scale alone. The gray scale is meant to give an impression and draw one's attention to an area that needs detailed examination. While its false positive and negatives do have some value, it should never be the sole basis for diagnosis. One might consider the gray scale as artistic license for the computer. The perimeter tests 77 points and then the artist in the computer extrapolates the data to points that have not been tested and draws out hundreds of points to construct the gray scale. In the process, small diagnostically important scotomas may be missed.

The gray scale can be perilous in the opposite manner too. To illustrate this point let us look at [Figure:9]. Here the gray scale shows the suggestion of a scotoma. The total deviation plot indicates an overall sinking of the hill of vision; or a scotoma involving a large area of the visual field. When the machine corrects for any overall depression of the visual field, the total deviation seems to indicate a central scotoma. However, we know that the pattern deviation plot compares findings to normal, while a scotoma is defined compared to the surround. In this figure when we look at the foveal threshold, it is much higher than the surrounding areas. It is therefore unlikely that there is a central scotoma. Of course, to be absolutely sure, one might want to look at points that are closer to the fovea. To do this we have to obtain a program'that tests points closer to the fovea such as the 10-2 program. So what is happening here? Why is the machine showing a central scotoma when none exists? This patient had a cataract, causing an overall sinking of the hill of vision. With cataract all points are depressed. Points in the periphery of the visual field vary more than those in the center. When the pattern deviation plot adjusts for any depression of the hill of vision, the points in the center of the visual field that fluctuate less than the points in the periphery exceed the boundaries of "normal" (at least as far as the computer is concerned). These central points are therefore marked out as abnormal. The smart computer inside the perimeter has created a central scotoma where none exists. This serves to reemphasize that the grey scale should never be the only parameter for diagnosis.

The question that follows the differentiation of a field defect from a normal field the is: "Is this due to glaucoma"?

 Confirmation of glaucoma field defect

There are three criteria by which we would label a field defect as early glaucoma.

Criterion 1: First in the pattern deviation plot, if it is a 30-2 program (i.e. testing out to 30 degrees), one has to ignore the outside edge points. This is not needed for a 24-2 program. Then let us see if there are three or more non-edge points that are depressed to an extent that would be found in less than 5% of the population; one of those points should be depressed to an extent found expected in 1% of the population. These points should be clustered in the arcuate area.

In [Figure:10] one can see that there are more than three such points in the pattern deviation plot, and all of them are depressed to an extent that has been expected in less than 0.5% of the population.[1] That is criterion 1.

Criterion 2. The Corrected Pattern Standard Deviation (CPSD) (or the PSD) should be depressed to an extent found in less than 5% of the population.[1] In [Figure:10] the CPSD is depressed to an extent found in less than 0.5% of the population.

Criterion 3. Is the Glaucoma hemifield test outside normal limits?[2] In [Figure:10] it is.

So there are three criteria by which we could say a field is glaucomatous or not. Do we always need all three criteria to establish a diagnosis of glaucoma? Can we rely on one positive criterion? This depends on the testing situation. If one wants to detect all the glaucomas and is not unhappy dealing with false positives, then even one positive criterion will suffice. On the other hand, if one is willing to miss very early glaucomas but wishes to minimize the false positives, then all three criteria must be positive before one considers a diagnosis of glaucoma. In [Figure:10], all three criteria are fulfilled.

In the clinical context, if the clinical features strongly indicate glaucoma (raised IOP, and suspicious disc with glaucomatous features), even one criterion is good enough to make a diagnosis. On the other hand, if the clinical features are not suspicious at all, ("normal" IOP and healthy neuroretinal rims that maintain the ISNT, (i.e, Inferior rim > Superior rim > Nasal rim > Temporal rim rule), all three criteria must be positive before one even begins to consider a diagnosis. Also we would repeat the field and go back and look at clinical features (especially the disc) before making a diagnosis.

Now that we know how to identify a field defect and also know how to decide whether it is due to glaucoma, let us interpret the field we looked at earlier [Figure:5]. Starting with zone 1, of course.

Zone 1: 30-2 program. First confirm that the said field you are interpreting actually belongs to the patient you are assessing.

The fixation target used was routine; not the large diamond. The birth date was entered correctly. The correction used was appropriate. The papillary diameter was 2.5mm. The visual acuity was 20/20. We immediately validate this by looking at the foveal threshold at the bottom of zone 2 - the foveal threshold correlates with the visual acuity. If it did not, we'd have pursued that further. The reliability criteria in zone 2 are okay.

We glance at zone 3, the gray scale. Let us refuse to be led astray and move on to the statistical analysis.

In Zone 4, the total deviation plot shows several points scattered around the field which are depressed to an extent expected in less than 5% of the population. This could be a generalized sinking of the hill of vision, but could be a scotoma too.

When we look at Zone 5 (or the Pattern deviation plot) where the machine has adjusted for any overall sinking of the hill of vision, (caused by whatever reason), no localized scotomas are revealed.

Next we look at the global indices in Zone 6. Except for the mean deviation, all the global indices are normal. The mean deviation has been flagged as abnormal; this again reflects an overall sinking of the hill of vision.

Zone 7: The Glaucoma Hemifield test finding states "general reduction sensitivity".

So all three statistical aids confirm that there is a generalized depression, a sinking of the hill of vision. The patient we discussed was actually a glaucoma suspect. A look at the actual threshold showed several points in the upper half of the visual field that were depressed compared to the lower half. However, since this patient is a very reliable and compliant person, we elected to follow him without any treatment.

 What does a typical cataract look like on automated perimetry ? [Figure:11]

The gray scale is depressed all over, but we won't tarry there. The total deviation plot shows several points that are depressed to an extent that would be found in a minority of the normal population. But once the machine adjusts for any depression of the hill of vision, the pattern deviation plot reveals no abnormality. To an automated perimeter, that's what a typical cataract looks like.

What about a typical glaucomatous Defect ? [Figure:10]

The gray scale shows something going on in the upper nasal quadrant. The total deviation plot shows several points depressed in the upper nasal quadrant. But when the machine adjusts for any sinking of the hill of vision, these points persist in the pattern deviation plot. All the criteria for a glaucomatous field defect are fulfilled.

Let's Look At Two More Cases [Figure:12a] [Figure:12b]

Are these two cases due to glaucoma?

Let's assume zones 1 to 3 are okay. Zone 4 shows several points that are depressed. Once the machine adjusts for any depression of the hill of vision, there remain points in the pattern deviation plots that fulfill the criteria for an early glaucomatous field defect. The CPSD has been turned off (unfortunately); the GHT is borderline. Could these fields be due to glaucoma?

That depends on the clinical picture. For example, if these fields belong to a patient with a suspicious looking disc, inferior rim thinner than the superior, mild pallor, 0.7 cup and pressure of 25 mmHg; then yes, these are due to glaucoma. On the other hand, if the inferior neuroretinal rim is thicker than the superior, with healthy rims, no pallor and the pressure were never beyond 18 mmHg, then we would not be too convinced. We might repeat the field. If these defects persisted, we would re-examine the disc more carefully on a slitlamp using a 60D (or similar lens, but sterobiomicroscopically) to see if there is something one has missed. But if the clinical findings remain similar, we would probably not pay attention to the visual field. A clinical correlation is of paramount importance. One should never interpret a visual field in isolation.

We now know there is a field defect. And we know that it is due to glaucoma. Time to try and answer the third question:

 Is the defect progressing ?

Before we determine whether this defect is progressing we must ensure that we have baseline visual fields to compare. The selected baseline fields must be after the learning curve of the patient. (Generally the learning curve is 2-3 fields in automated perimetry.) Most patients have a learning curve for automated perimetry. Generally this takes two to three fields but could face longer. Several factors may interfere with our ability to determine progression: they include something self explanatory called the long-term fluctuation; artifacts which may be present only in one field and simulate progression; patient factors (a patient might do badly on a particular day), the pupil size, etc.

The field of the patient in [Figure:13] was deemed to have progressed (compared to the earlier one). When you look at the pupil size, the pupil size was small. The patient had been recently started on pilocarpine. When the field was repeated after dilating the pupil [Figure:14], it began to resemble the earlier field and no further intervention was deemed necessary.

The goal of follow up is to detect a difference. The question is: has anything changed? Has the patient's scotoma grown in size or depth, or has it decreased? To determine this, the machine provides several statistical packages like "Change Analysis", "Overview" and the "Glaucoma Change Probability". Generally, for purposes of follow up we prefer to use the Overview and the Glaucoma Change Probability.

The Overview program is a sequential series of fields of same patient printed out on a single piece paper that contains all the data that a single field analysis provides. Up to 16 fields can be printed on a single piece of paper Looking at the Overview program, one can get a feel for what is happening and also why. For example, the fields in [Figure:15] show that the total deviation plot is getting worse over time; but once the machine has adjusted for any sinking of the hill of vision, the pattern deviation plot has also become worse. The patient probably has glaucoma that is getting worse.

The Overview program also helps us to determine whether progression is due to cataract or glaucoma. For example, [Figure:16] shows a patient whose visual field is getting worse. The worsening involves not just the total deviation, but also the pattern deviation plot. This patient's glaucoma has gotten worse. A surgery should have been performed; last year, perhaps.

[Figure:17] is again an Overview program where the field is getting worse. We can appreciate that till the second last field, the total deviation plot has gotten worse, but the pattern deviation plot has hardly been affected. This patient probably had cataract. If we take out the cataract, the Total deviation plot should improve. As one can see from the last field, that is exactly what happened. The cataract surgery was done. The Total Deviation plot improved and the pattern deviation plot remained unaffected.

While the Overview program provides a general idea of the visual field changes glaucoma change probability program provides the necessary statistical help. In the glaucoma change probability program, we select two fields as the baseline [Figure:18]. These are not necessarily the first two fields that the patient has performed; rather they are the fields done after the learning curve of the patient. The baseline may need to be changed. For example, if the patient has undergone intervention (such as a trabeculectomy), we would prefer to obtain a new baseline.

Either way, two fields are obtained as baseline. Subsequent fields are compared to the baseline, and are compared to a set of stable glaucoma patients. In the extreme right of the printout, the closed triangles indicate points that have worsened; open triangles indicate points that have improved; dots indicate points where enough data is not available to make a comment. One can see in [Figure:18] a lot of black triangles indicating that this patient's field is worsening. The machine also provides a message stating that less than 2.5% of stable glaucoma patients would be expected to show such worsening.

[Figure:19] shows another glaucoma change probability. The first two fields are the baseline; subsequent fields are compared to the baseline as well as to a stable set of glaucoma patients. One can see that when the comparison is made for the first field, there are several black triangles, points that have worsened. The machine indicates that less than 2.5% of stable glaucoma patients would be expected to show such worsening. However, as with any field where we suspect worsening, we always repeat the field. And when we do that, the next field shows that there are only 3 points that have worsened.

[Figure:20a] shows the baseline fields for a patient with glaucoma. Subsequent fields are compared to this baseline and show worsening [Figure:20b]. In the subsequent fields worsening continues. It is obvious that this person should have probably had surgery.

When the visual field is very advanced, a 30-2, (or 24-2) single field analysis is no longer very helpful. [Figure:21] shows the field of a patient with advanced open angle glaucoma. The total deviation plot is depressed, but the pattern deviation plot is clear. This is due to the advanced nature of the field - most of the points have a zero threshold and very few points have any sensitivity. Under these conditions, the statistical assumptions underlying the pattern deviation plot are no longer fulfilled. This results in the blank pattern deviation plot. It no longer provides any information about whether there is indeed an arcuate type scotoma lurking somewhere.

When the patient has such advanced glaucomatous field defects, we have to fall back on other strategies. The first thing we can do is to test the central 10 degrees. The central 10 degrees are blanketed with points to give us a two degree resolution. Once we test the 10 degrees in this manner, [Figure:22], enough points have been thresholded to fulfill the statistical assumptions required for the pattern deviation plot'. In [Figure:22], the pattern deviation plot returns and shows a double arcuate scotoma. In the newer versions of the software releases, global indices are also available with the 10-2 program.

If the field defect is so advanced that most points have sensitivity less than 10 - 15 decibels, it is then preferable to use the size 5 target [Figure:23]. With the size 5 target we do not have the usual total deviation pattern deviation plot. Instead, we get a defect depth and actual threshold values with which to follow up the patient.

In some patients we may elect to use a macular program [Figure:24]. The macular program can be done with the standard size 3 target or, if the sensitivities are low, with the size 5 target. We find the macular program useful to determine if the macula is split. This has a bearing on patient prognosis - a test showing a split macula with the size 3 target should be repeated using the size 5 stimulus [Figure:25], the absence of a split macula on the size 5 macular program indicates a better visual prognosis. The macular program is generally used only when the central five degrees of field remains. If a defect impinges on fixation, but other points in the field are preserved, the macular program is used in addition to, not in lieu of the standard programs.

A newer thresholding algorithm is now available in SITA. This algorithm shows more abnormal points in the pattern deviation than the full threshold program. The detected defects are shallower, but achieve significance probably because of the narrower confidence limits.

A SITA print out is shown in [Figure:26]. In order to make the test shorter, determination of short term fluctuation has been eliminated; we also lose the corrected pattern standard deviation. Hence in the printout there are no SF and the CPSD.

The interpretation of the SITA program is similar to the full threshold program already discussed. We interpret them in the same 8 zones. Zone 6 does not contain SF and CPSD. It does not matter: In the diagnosis of glaucoma the criteria of CPSD less than 5% can be replaced by the criteria of PSD less than 5%.

Accordingly, in [Figure:27], we would say that the criteria for the pattern deviation plot (3 non edge points depressed to an extent found in less than 5% of the population, one of which is depressed to an extent found in less than 1% of the population) is fulfilled. In the global indices we see that pattern standard deviation is depressed and has a value expected in less than 5% in the population. Thirdly, the Glaucoma Hemifield Test is abnormal. This field fulfills all three of Anderson and Patella's criteria[1] for a glaucomatous field defect. So the interpretation of the SITA program remains exactly the same as the other programs.

The line tracing at the bottom of the page represents the gaze monitor tracing available on the newer HFA models, The details are beyond scope of this article. In brief, the machine records every movement of the eye off fixation as an upward spike on the tracing; a blink is recorded as a downward spike. The gaze tracker records eye movements for the entire duration of the test and provides a better knowledge of fixation losses compared to the standard method. If the formal testing of fixation losses is turned off, the testing time is further shortened. The tracing however, does not provide details of the point tested at the time of fixation loss, or blink.

 Applying the skills

Having learned this much, one should be ready to apply these interpretive skills - so here in an opportunity. [Figure:28] shows a field incorporating all the principles we read about.

For the interests of space, let's stipulate that Zone 1 and 2 are normal. Zone 3 shows an arcuate defect but we won't tarry there. In Zone 4, the total deviation plot shows several points that are depressed well below the expected normals. Once the machine adjusts for the hill of vision, the Pattern deviation plot shows an arcuate scotoma. The global indices are depressed (including the PSD and CPSD at the 5% level). Finally the glaucoma hemifield test is outside normal limits.

All would agree that this field fits all the criteria for a glaucomatous field defect. The main criteria are: (a) the three non-edge point criteria for the pattern deviation plot; (b) the Corrected Pattern Standard Deviation (or in SITA the PSD) depressed to a level found in less than 5 % of the normal population; and (c) the glaucoma hemifield test outside normal limits. We all agree that is consistent with a typical glaucomatous field defect and that this patient has glaucoma.

[Figure:29] shows the optic disc of the same patient. It is important to remember however, that the field defects must always be correlated with clinical findings. This particular field defect was explained by this inferior coloboma.

In conclusion then, automated perimetry has several advantages and is the current state of the art in visual field testing. This is particularly true in diagnosis and management of glaucoma. However, as emphasized by Zalta,[2] sophisticated techniques and elaborate data printouts should not seduce us into a false sense of security or misplaced belief in the reliability of automated perimetry. Certainly, the field must never, ever be interpreted in isolation.


1Anderson DR, Patella VM. Automated Perimetry (2nd Edition). St. Louis: Mosby & Co; 1999.
2Zalta AH. Lens rim artifact in automated threshold perimetry. Ophthalmology 1989;96:1302-ll.