Comparison between Semiautomated Kinetic Perimetry and

Comparison between Semiautomated Kinetic
Perimetry and Conventional Goldmann
Manual Kinetic Perimetry in Advanced
Visual Field Loss
Katarzyna Nowomiejska, Dr med,1,2 Reinhard Vonthein, Dr rer pol,3 Jens Paetzold, Dr rer nat,1
Zbigniew Zagorski, Prof Dr med,2 Randy Kardon, MD, PhD,4 Ulrich Schiefer, Prof Dr med1
Purpose: To compare quantitatively visual field (VF) results obtained using a new standardized semiautomated kinetic perimetry (SKP) with those obtained by conventional Goldmann manual kinetic perimetry (MKP) in
patients with advanced VF loss.
Design: Prospective, single-center, observational comparative case series.
Subjects and Methods: Seventy-seven eligible patients (36 suffering from advanced retinal nerve fiber layer
loss, 20 with concentric constriction of the VF, and 21 with hemianopia) were included in the study. One eye of
each patient was examined on the same day with MKP and SKP. Three isopters, identical in both tests, were
chosen to assess the extent of the VF loss. To compare the location and size of the corresponding isopters
obtained with MKP and SKP, intersection areas of superimposed isopters were expressed as a percentage of
union areas.
Main Outcome Measures: The area and position of isopters for a defined stimulus condition obtained with
both methods were compared. Test duration and patients’ preference were also evaluated.
Results: Isopters obtained with Goldmann MKP enclosed areas smaller by 20% (confidence interval [CI],
12%–27%). The mean intersection area of Goldmann and SKP VFs was 1763.1 square degrees (CI, 1558.6 –
1967.7) smaller than the union for stimulus III4e over all groups of patients. Semiautomated kinetic perimetry was
preferred by 60% of patients with concentric constriction of the VF. Median duration of the examination was 15
minutes and did not differ significantly between the 2 methods.
Conclusions: Our results indicate that SKP isopter shape and size were very comparable to those obtained
on the same eyes with MKP. Semiautomated kinetic perimetry may represent a more standardized method of
kinetic perimetry, which still takes advantage of perimetrist–patient interaction to diagnose and monitor advanced VF loss in clinical practice. Ophthalmology 2005;112:1343–1354 © 2005 by the American Academy of
Ophthalmology.
Since its introduction almost 60 years ago,1,2 manual kinetic
perimetry (MKP) using a Goldmann perimeter has become
the method of choice for the edge detection of advanced
steeply bordered absolute scotomas. It allows the entire
visual field (VF) to be surveyed in a comparatively short
time and quantifies the extent and shape of profound VF
defects. This is why it is still a well-accepted method of
monitoring progression, stability, or improvement of pro-
found VF defects in patients suffering from concentric
constriction of the VF,3 advanced retinal nerve fiber layer
(RNFL) loss,4 and hemianopia.5 Kinetic perimetry is often
a better alternative to static perimetry when the VF defects
are dense and large and when the shape characteristics are
important in the diagnosis and monitoring of change in
many patients. In addition, some patients are better suited
for kinetic perimetry based on fatigue artifacts that can
Originally received: September 23, 2004.
Accepted: December 30, 2004.
Manuscript no. 2004-127.
1
Department of Pathophysiology of Vision and Neuro-ophthalmology,
University Eye Hospital, Tuebingen, Germany.
2
First Department of Ophthalmology, Eye Hospital, Medical University,
Lublin, Poland.
3
Department of Medical Biometry, University of Tuebingen, Tuebingen,
Germany.
4
Neuro-ophthalmology Division, Department of Ophthalmology and Visual Science, University of Iowa Hospital and Clinics and VA Hospital,
Iowa City, Iowa.
Drs Nowomiejska and Vonthein contributed equally to this work.
This study was supported by the European Union, Brussels, Belgium,
under the FP5 Marie Curie Training Site “Fighting Blindness” (contract
no.: QLG5-CT-2001-60034 [KN]). Consulting fees and lecture fees paid
by Haag-Streit Inc., Bern, Switzerland: Dr Paetzold (about 2500 Euro/year)
and Dr Schiefer (about 2000 Euro/year). Grant support from Haag-Streit
Inc.: Dr Schiefer (about 10 000 Euros [2001-2004]).
Correspondence to Prof Dr med Ulrich Schiefer, University Eye Hospital,
Department of Pathophysiology of Vision and Neuro-ophthalmology,
Schleichstrasse 12-16, D-72076 Tuebingen, Germany. E-mail:
[email protected].
© 2005 by the American Academy of Ophthalmology
Published by Elsevier Inc.
ISSN 0161-6420/05/$–see front matter
doi:10.1016/j.ophtha.2004.12.047
1343
Ophthalmology Volume 112, Number 8, August 2005
Figure 1. Perimetric results in a glaucomatous optic neuropathy. Left, Results of conventional Goldmann perimetry. Right, Results using the the Octopus
101 instrument’s semiautomated kinetic perimetry (SKP). The arrows represent the origin, direction, and end of the stimulus movement. Double-headed
arrows represent so-called reaction time vectors, which are presented within intact regions of the visual field. Filled arrowheads symbolize moved stimuli,
not seen by the patient. Broad shaded lines show the age-correlated normative isopter for each used kinetic stimulus. Isolated circles without a vector
symbolize static stimulus presentation. Filled circles symbolize static stimuli, not perceived by the patient. Transparent circles symbolize perceived stimuli.
OS ⫽ left eye.
occur with static perimetry,6 the need for interaction with a
perimetrist, and cases in which static– kinetic dissociation of
VF defects is present. Additionally, MKP is more comfortable for some patients relative to static perimetry,7 and
fixation may also be easier to perform.8 Another advantage
is that the perimetrist–patient interaction during the test can
guide the kinetic testing, with the perimetrist querying the
patient and rejecting spurious responses based on poor
attention or anticipation.9 Because a motion detection task
may also better relate to real-world visual performance than
static perimetry, it is also understandable that this method
may be especially important when there is a need for assessment of visual performance and disability in relation to
work and driving or assessment for other medical–legal
purposes. However, there are many disadvantages of the
Goldmann instrument used for these purposes. Direction
and speed of stimulus movement are guided by the examiners’ hand and, therefore, are difficult to standardize. Thus,
the results depend on the examiner’s skills and may be
confounded by examiner bias.9 Examiner dependence can
be associated with inaccuracy, which results in a limited
capacity to detect defects and poor reproducibility of re-
1344
sults.10 Kinetic VF results, as being subjectively obtained,
are notoriously difficult to quantify, and this is made even
more difficult by the lack of standardization of equipment
and method.11 Because of the pantograph mechanism in the
Goldmann instrument, the spatial resolution decreases with
increasing eccentricity, which can give rise to a poor cartographic accuracy.12 There are also other shortcomings of
the Goldmann perimeter, such as lack of autocalibration,
lack of permanent documentation of the test procedure used
to determine individual VF borders, and the inability to
examine the area of 2° around the fixation point with the
standard setting due to the telescope used for fixation control. In principle, this problem can be solved by an alternate
fixation target, but it is still a difficult process.
To overcome disadvantages of Goldmann perimetry, a
new software-based technique (Invest Ophthalmol Vis Sci
41:295, 2000) called semiautomated kinetic perimetry
(SKP) was designed to perform interactive kinetic VF examination using the Octopus 101 instrument (Haag-Streit
Inc., Bern, Switzerland).13–17 Semiautomated kinetic perimetry allows the examination of almost the entire field of
vision, with stimulus movements in any direction, including
Nowomiejska et al 䡠 Comparison between Semiautomated and Manual Kinetic Perimetry
Figure 2. Perimetric results in a homonymous quadrantonopia due to an astrocytoma of the right cerebral hemisphere. Top, Results of conventional Goldmann
perimetry. Bottom, Results using the Octopus 101 instrument’s semiautomated kinetic perimetry (SKP). For description of symbols, see Figure 1. OD ⫽ right eye;
OS ⫽ left eye.
1345
Ophthalmology Volume 112, Number 8, August 2005
Figure 3. Perimetric results in a concentric constriction due to retinitis pigmentosa. Left, Results of conventional Goldmann perimetry. Right, Results
using the Octopus 101 instrument’s semiautomated kinetic perimetry (SKP). For description of symbols, see Figure 1. OS ⫽ left eye.
the periphery (up to 85% horizontally). It takes advantage of
standardization of testing conditions by using computercontrolled presentation of a stimulus at a specified constant
velocity for any chosen Goldmann stimulus size and intensity. In addition, the stimulus vector starting point, ending
point, and direction can be prespecified by the user, and the
delay between stimulus presentation and patient response
can be recorded.
The purpose of this study was to compare conventional
Goldmann MKP with SKP implemented in an Octopus 101
perimeter in patients with advanced VF loss. The 2 methods
of kinetic perimetry were compared quantitatively in the
same eye by assessing differences in the extent and location
of VF defects, differences in test duration, and patient
preference for either MPK or SKP.
Materials and Methods
Patients
Eighty-one patients suffering from advanced VF loss were recruited from the outpatient clinic of the University Eye Hospital in
Tuebingen, Germany. Sample size was planned to give a level 5%
1346
equivalence test power 95% to detect differences in VF size of
20%. The study was approved by the independent ethics committee and performed in accordance with the ethical standards laid
down in the 1964 Declaration of Helsinki. Written informed consent was obtained from each individual after an explanation of the
nature of the study. One eye of each individual was included in the
study; if both eyes met the inclusion criteria, the more advanced
eye was selected. Patients have been experienced in taking at least
one VF examination. All subjects underwent a full ophthalmologic
examination consisting of visual acuity (VA) testing, slit-lamp
examination, applanation tonometry, and a fundus examination.
The patients had to meet the following inclusion criteria: at least
18 years of age, refractive error of ⬍8 diopters (the authors
excluded higher ametropias to avoid artifacts due to correction
lenses), best refracted VA better than 10/20, clear central optic
media, pupil diameter of ⬎3 mm, no other diseases affecting the
VF (e.g., no macular degeneration, retinal vein or artery occlusion,
retinal detachment, diabetic retinopathy, or drusen of the optic
nerve head). Enrolled patients were classified into 3 categories
according to their VF defects: (1) 36 cases of advanced VF loss
(greater than Aulhorn stage III [as defined by Aulhorn and Karmeyer18]) with an arcuate defect extending toward and including the
blind spot constituting RNFL loss due to glaucoma or anterior
ischemic optic neuropathy, (2) 20 cases of concentric constriction
of the VF due to retinitis pigmentosa, and (3) 21 cases of hemi-
Nowomiejska et al 䡠 Comparison between Semiautomated and Manual Kinetic Perimetry
Figure 4. Semiautomated kinetic perimetry delineating a circumscribed inferior arcuate visual field defect due to glaucomatous optic neuropathy.
Repetitive (4 times) presentation along each vector reveals local intrasubject intrasession scatter, which is symbolized by the grey error bars (mean of local
kinetic threshold ⫾ standard deviation). For description of symbols, see Figure 1.
anopic VF defects due to a chiasmal or postchiasmal lesions
(tumor, stroke, or trauma). Four patients were excluded because
their VF defect was smaller than expected (i.e., less than Aulhorn
stage III), so the total consisted of 77 patients who were analyzed.
The mean age of the 77 patients was 54 years. The median age of
patients with retinal nerve fiber layer loss was 67 years (range,
37– 86), whereas median ages of patients with concentric and hemianopic VF defects were 49 years (22–78) and 41 years (18 – 62),
respectively. Thirty-eight women and 39 men were enrolled.
Examination Procedure
Perimetric examinations for this study were performed in random
sequence by the same skilled examiner (KN) on the same day
using a Goldmann perimeter (Haag-Streit) and a modified Octopus
101 instrument. A rest break of at least 5 minutes was given
between tests. The background luminance was standardized at the
level of 10 candelas per square meter (31.5 apostilb) in both
instruments. An opaque occluder covered the untested eye during
the perimetric session. Appropriate near refraction (adjusted for a
33-cm Goldmann bowl radius and a 45-cm Octopus 101 bowl
radius) with adjustment for age was provided for each patient
inside the central 30° VF and was removed for plotting peripheral
isopters in both methods. The near correction was carefully
checked with each subject before starting the examination. The
examiner monitored fixation in both instruments. Additional rest
breaks were given when requested.
Manual kinetic perimetry with the Goldmann instrument, calibrated according to the manufacturer’s instructions, was performed on all patients in the standard fashion19 with continuous
telescopic observation of the position of the eye. The hemispherical bowl of the Octopus 101 instrument was connected to and
1347
Ophthalmology Volume 112, Number 8, August 2005
Figure 5. Follow-up (3 sessions) with semiautomated kinetic perimetry (SKP) of a patient suffering from glaucomatous optic neuropathy, revealing the
intrasubject intersession variability. For description of symbols, see Figure 1. OD ⫽ right eye; OS ⫽ left eye.
controlled by an external personal computer with the SKP module
implemented. Semiautomated kinetic perimetry provides continuous movement of stimuli of selected size and luminance according
to the Goldmann classification along user-defined vectors having a
constant angular velocity chosen from 1° to 25° per second.
Vectors are drawn manually using a computer mouse or an electronic pen directly on the computer touch screen either before the
perimetric test begins or also during the test. Origin, direction, and
length of the vectors are individually selected, so the stimulus is
moved almost perpendicularly towards the presumed scotoma border from nonseeing towards seeing areas of the VF. Fixation is
monitored by a digital infrared camera, which provides a highly
magnified image of the tested eye. The stimulus movement along
each vector is terminated by the response of the patient, who is
instructed to look straight ahead at the fixation point (green cross)
and press a button as soon as the stimulus is perceived. The
respective stimulus location is marked on the screen automatically
by the software with a size- and intensity-specific symbol, and
after several repetitions with different vectors, the symbols are
connected, and enabling isopters are drawn in selectable colour.
Additionally, it is possible to record the individual reaction time
(RT) of each patient by presenting RT vectors within the seeing
regions of the VF (RT vectors are, by definition, always located
within the seeing area of an isopter). The area enclosed by an
isopter was automatically quantified by the software using triangulation in square degrees of eccentricity.
In the present study, stimulus angular velocity was selected to
be 3° per second in SKP and was kept approximately at this speed
during the manual presentation of stimuli in Goldmann MKP over
the entire VF. Identical stimulus size and intensity (same isopters)
were chosen for SKP and MKP, respectively. Because III4e stimuli are standard for opinion regarding driving and disability, and
I4e is used for assessment of the blind spot, targets III4e and I4e
were obligatory for all 77 examinations. One other stimulus (V4e
in 31 VFs, I3e in 21 VFs, or I2e in 25 VFs) was individually
chosen depending on the VA and the depth of VF defect of each
patient. All isopters were tested in an almost equidistant manner on
the individual hill of vision. Stimuli (vectors) were presented about
1348
every 15° in random order along the presumed VF border. Inside
the central 30° VF, at least 12 stimuli (vectors) were presented, and
outside 30° of eccentricity, at least 24 stimuli (vectors) were
presented. The origin of the vector was shifted outwards when not
seen. The blind spot was always mapped using the I4e test object
(at least 5 vectors starting from its presumed center). For scotoma
testing, which also included the assessment of the blind spot, the
test object was presented inside the region of the field loss and
moved outwards perpendicularly to the border of the scotoma until
it was perceived. Local kinetic threshold results were linearly
connected after examination (manually with pencil on a chart in
MKP and electronically on a screen in SKP).Three typical results
are shown in Figures 1 to 3. We also show an example of
intrasubject intrasession scatter (Fig 4). The retest reliability additionally will be analyzed in an ongoing study. In the meantime,
a participant of this study was examined 3 times (Fig 5).
Test duration was measured in minutes automatically during
SKP and by the examiner during Goldmann perimetry by stopwatch as overall examination time (with breaks) and net examination time (without breaks). After both examinations, patients
were asked to fill in a questionnaire of subjective self-assessment.
The first question was “Which examination procedure was more
comfortable in your opinion?” The next questions addressed how
patients rated the course of the examination and the contact with
the examiner. The rating scales ranged from 1, the best, to 5, the
worst.
Data Analysis
Because one eye of each of 77 patients was tested on 2 perimeters
with 3 stimuli, a total of 462 isopters were obtained. Analysis of
VFs consisted of comparing the area and position of isopters
obtained with both instruments. The area of each isopter in SKP
was measured automatically in square degrees without RT correction. The area (square degrees) of the blind spot or other defined
scotomas was subtracted from the total corresponding isopter area.
When separate islands of the VF occurred within the same isopter,
their areas were added. Goldmann VF results were digitized using
Nowomiejska et al 䡠 Comparison between Semiautomated and Manual Kinetic Perimetry
Figure 6. Union area and intersection area (shaded region) obtained by superimposing III4e isopters with the Goldmann and Octopus instruments in one
representative patient.
the touch screen of the enhanced SKP workstation, and the area of
isopters was electronically measured in square degrees. Additionally, both Goldmann and SKP VF results were superimposed in
SKP software by the same examiner to compare the position of
corresponding isopters (Fig 6). The areas of intersection and union
for each isopter were determined to quantify the similarity between
SKP and MKP. For example, if the isopter shape and size were
exactly the same in an eye examined with the 2 methods of testing,
then the intersection area of the isopters would be the same as the
union area (see “Statistical Analysis”).
Statistical Analysis
Computations were carried out with JMP statistical software (version 5.0.1.2, SAS institute Inc., Cary, NC). The areas of isopters
obtained with both methods were compared according to Bland
and Altman.20 We extended the approach by calculating areas of
unions and intersections obtained by superimposing corresponding
isopters. Intersection areas were expressed as percentage of union
areas to compare the location of the VFs as well as their size.
Confidence intervals (CIs) for the median were estimated via the
logit transformation. Another approach assuming a mean meridian
model was carried out by multiple regression from the one VF area
on the other VF area and its square root. If radii only were to differ
systematically by d, coefficients of that regression should be 1, and
proportional to d, respectively, and a constant proportional to d 2.
Individual differences of examination duration are described by
medians and 95% CIs for ratios of geometric means from an
analysis of variance (ANOVA), with order as an additional factor.
For indicators of preference generated by the patient survey, fre-
1349
Ophthalmology Volume 112, Number 8, August 2005
Table 1. Comparison of Mean Visual Field (VF) Areas of Particular Isopters Obtained with 2 Methods
Isopter
Cases
VF area
Goldmann (square degrees)
Octopus (square degrees)
Mean difference (square degrees)
95% confidence interval of mean difference (square degrees)
I2e
I3e
25
21
1075
1560
⫺484
⫺708 to ⫺261
quency of each response was reported by VF defect type. Patients’
ratings were summarized by odds ratios (ORs) of ordinal logistic
regressions.
860
1311
⫺451
⫺665 to ⫺236
I4e
77
3391
4207
⫺815
⫺1133 to ⫺497
III4e
V4e
77
31
7343
7730
⫺387
⫺666 to ⫺108
6247
6391
⫺144
⫺668 to 380
Goldmann perimetry with stimulus I4e resulted in a far larger VF
defect than SKP.
Comparison of the Position of Isopters
Results
Comparison of the Area of Isopters
The overall median of 77 Goldmann VF areas inside the III4e
isopter was 7942.4 square degrees (range, 17.2–16 033.4), and that
for the I4e isopter was 3282.4 square degrees (range,
5.4 –11 622.8). Overall, the Goldmann VFs were smaller than the
SKP VF by 20% (CI, 12%–27%). The mean VF area of the
isopters obtained with each method, mean difference, and 95% CI
are presented in Table 1. The mean difference in areas between the
2 perimetric methods was the least for target V4e (⫺144 square
degrees), which is the largest and brightest stimulus. The difference was the greatest for target I4e (⫺815 square degrees), followed by I2e and I3e, which are small, dim stimuli. This is shown
in Figure 7 for target III4e and in Figure 8 for target I4e, which
were used with all patients. Relative to the mean area inside an
isopter, differences in areas for the 2 perimetric tests were smaller
than 25%. In only one patient who had hemianopic VF loss,
The intersection area within III4e isopters obtained by superimposing Goldmann and SKP results must be less than the
union isopter area of these 2 VFs, assuming some measurement
differences between the 2 perimeters. The proportions
amounted to 97% (CI, 97%–98%) of the union of the III4e VFs
in patients with an RNFL defect, 98% (CI, 97%–98%) in those
with hemianopic, and 94% (CI, 91%–96%) in those with a
concentric VF defect, and differed by 1763.1 square degrees
(CI, 1558.6 –1967.7) over all groups.
Approximating VFs by circles, as in a mean meridian model,
the radius is estimated by the square root of the area divided by ␲.
Accordingly, the scatter plot of union and intersection areas was
described by a multiple linear regression from the larger union area
on the smaller intersection area and the square root of the intersection area. As the lines run almost parallel to the line of equality,
a Bland–Altman diagram of the mean radii gives an estimate of the
difference in degrees eccentricity (Fig 9). When subtracting the
mean meridian values obtained with MKP from those obtained
with SKP, systematic shifts induced by the examiner are estimated
Figure 7. Bland–Altman plot to compare stimulus III4e visual field (VF) areas obtained with the Goldmann perimeter and the semiautomated kinetic
perimetry software on the Octopus instrument as a function of the mean VF area at that isopter. The 77 patients suffered from retinal nerve fiber layer
(RNFL), concentric, and hemianopic VF defects. Mean difference (solid horizontal line) is estimated into 95% confidence and reference intervals (broken
lines). deg ⫽ degrees.
1350
Nowomiejska et al 䡠 Comparison between Semiautomated and Manual Kinetic Perimetry
Figure 8. Bland–Altman plot to compare stimulus I4e visual field (VF) areas obtained with the Goldmann perimeter and the semiautomated kinetic
perimetry software on the Octopus instrument. The 77 patients suffered from retinal nerve fiber layer (RNFL), concentric, and hemianopic VF defects.
Mean difference (solid horizontal line) is estimated into 95% confidence and reference intervals (broken lines). deg ⫽ degrees.
to 1.4° (CI, 0.5°–2.3°) with stimulus III4e and 3.7° (CI, 2.4°–5.1°)
with stimulus I4e. The standard deviations of such mean measurements’ differences were 4.03° and 6.1°.
Comparison of the Test Duration
Median overall times were 15 minutes (range, 5–28) per eye on
the Octopus perimeter and 15 minutes (range, 5–39) per eye on
the Goldmann perimeter. Median net test durations were 14
minutes (range, 5–25) and 13 minutes (range, 5–37), respectively. The estimated Box–Cox transformation suggested logarithms to ensure the assumptions of normality and equal variances. The ANOVA of logarithms of examination durations by
patient (random), perimeter type, diagnosis group, order, and
the interaction of type and order showed no interaction between
diagnosis group and perimeter type, so it was eliminated from
the subsequent calculations. Durations did not differ much
between Goldmann perimetry and SKP, the latter being faster
Figure 9. Estimated difference of radius (computed as the square root of the area inside isopter III4e divided by ␲) for RNFL, concentric, and hemianopic visual
field defects. Mean difference (solid line) and 95% confidence and reference intervals (broken lines) are shown. deg ⫽ degrees.
1351
Ophthalmology Volume 112, Number 8, August 2005
Figure 10. Examination duration including breaks by type of perimetry
and sequence of examinations. 1, Examination performed with the device
named below as the first one. 2, Examination performed with the device
named below as the second one. Strong lines show geometric means, and
thin ones indicate confidence intervals. Semiautomatic examination (Octopus) is quick from the start, whereas manual examination (Goldmann)
is quick only when conducted as the second one. min ⫽ minutes.
overall by 6% (CI, ⫺3% to 13%) and 8% (CI, ⫺0% to 14%)
without breaks. Second examinations were performed faster by
22% overall (CI, 7%–34%) and by 47% with the Goldmann
perimeter (CI, 22%–78%) (Fig 10). Examination duration was
arguably longer for patients suffering from RNFL defects than
for hemianopic patients, by 24% (CI, ⫺1% to 55%) overall and
by 25% (CI, 2%–53%) without breaks, and was perhaps a bit
shorter for patients with concentric constriction of the VF than
for those with RNFL defects, by 17% (CI, ⫺7% to 35%) overall
and by 16% (CI, ⫺6% to 33%) without breaks. From these
results, one can conclude that all expected differences were
detected but the one between perimetry procedures.
Comparison of Patients’ Preferences
Octopus instrument examination was preferred by 40 patients
(52%), Goldmann was preferred by 25 patients (32%), and 12
patients (16%) had no preferrence. Semiautomated kinetic perimetry examination by the Octopus instrument was preferred the
most among patients with concentric constriction of the VF (60%),
whereas kinetic examination with the Goldmann instrument scored
highest in hemianopic patients (47%). The OR for contact with the
examiner was 0.75 (CI, 0.55–1.02) for the Octopus examination
versus the Goldmann examination, indicating a tendency in favor
of the Goldmann instrument, but no significant difference. The
Octopus instrument got significantly higher ratings in the assessment of the examination course (OR, 1.5; CI, 1.1–2.1).
Discussion
The development of automated perimetry has been concerned mainly with static perimetry because automated test-
1352
ing with kinetic targets is more difficult to run with a
computer program. Some attempts have been made to automate this procedure,21,22 but none of them has become a
standard procedure in clinical practice due to the technical
limitations. For example, Twinfield perimeters (Oculus,
Wetzlar, Germany) allowed examination between 2° and
70° eccentricity only with stimulus angular velocity between 1° and 10° per second.23 Fieldmaster 5000 (Synemed,
Benicia, CA) perimeters offered the possibility to combine
static (within 30°) and kinetic (by plotting isopters outside
of 30°) VF testing.24 An automated kinetic mode was also
added to the Humphrey Field Analyzer 630 (Humphrey
Instruments, Dublin, CA), but it also was not adopted for
wide use in a clinical setting.25 None of them offered the
option to quantify precisely the extent of VF loss progression and to move stimuli in other than a centripetal direction; therefore, they could not have been applied for testing
defects other than concentric constriction of VFs, such as in
hemianopia or in RNFL defects.
Quantification of VF loss in MKP is especially important
for observing scotoma progression, but significant progression is difficult to establish8,26 and eventually requires complicated planimetrical measurement techniques,11 which
seem to be unrealistic in the case of irregularly shaped
scotomas. Drance et al27 determined the VF areas of normal
and glaucoma patients in square millimeters using a
planimeter, but the areas of glaucomatous and normal VFs
were not compared. Because SKP offers the possibility of
measuring the area of isopters in square degrees, an attempt
was made to compare VF areas on Goldmann perimetry and
SKP in this study. Goldmann VF data were entered manually into the Octopus system using the touch screen by the
one examiner (KN).
We observed that VF isopter areas obtained with
Goldmann were smaller than those obtained with the
Octopus 101 perimeter. We presume that constant stimulus velocity and elimination of the examiner’s RT resulted in larger VFs obtained with the Octopus perimeter.
If only the VF area would be taken into consideration, it
would be possible to obtain the same size of a different
shape at a different location. To compare the isopter
shape, location, and area using the 2 kinetic perimetric
methods, the isopters were superimposed to derive the
intersection area and union area. For example, if the same
isopters differed in shape between Goldmann and SKP
methods but had the same areas, then the intersection
area shared by both isopters would be smaller than the
union area, resulting in a proportion of ⬍1 (⬍100%). If
the 2 isopters were in perfect agreement, the proportion
would be 1 (100%). For the III4e isopter, the average
correspondence area was 73% between the 2 perimetric
methods used in this study. The Goldmann MKP’s
isopter area generally was smaller than that obtained with
SKP. This is most probably due to the fact that, for
technical reasons, in the manual mode with the Goldmann instrument stimuli cannot be presented at an exactly constant angular velocity. Furthermore, angular velocity was most probably higher than the intended level
(i.e., 3° per second), which was achieved exactly by SKP.
Both mechanisms could result in a systematic shift of
Nowomiejska et al 䡠 Comparison between Semiautomated and Manual Kinetic Perimetry
scotoma borders in the stimulus direction, resulting in an
inward shift of the VF borders, which also leads to a
reduction of the area within. Insufficient standardization
of the Goldmann instrument examination results in loss
of precision. Therefore, MKP may not represent an adequate gold standard. The examination duration of a 1:1
mix of new and known VFs, as shown in the dot plot (Fig
10), was nearly the same in the 2 instruments. Semiautomatic examination is quick from the start, whereas
manual examination is quick only when conducted as a
second examination. It is implicit that Goldmann perimetry depends on the examiner’s knowledge of the scotoma pattern more than SKP does and requires more
active involvement of the examiner.
In kinetic perimetry, the influence of target speed on
the location of a border of a defect is of fundamental
importance. The slower the movement, the smaller the
defects that can be detected; 3° per second seems to be
the optimal velocity— between the 4° per second recommended by Johnson and Keltner28 and Wabbels and
Kolling (Der Ophthalmologe K20[suppl 1]:5, 1999) and
the 1° to 2° per second proposed by Lachenmayr and
Vivell.29 It is also known that only a slight difference in
the threshold location can be demonstrated comparing
speeds of 1° per second and 5° per second.10 We have
observed that the order of the 2 types of kinetic examination performed influenced testing time. The second test
performed was faster, which is probably due to the learning effect of both the examiner and the patient. It already
has been observed that normal and neurological VFs take
slightly less time and the glaucoma VFs more time compared with static perimetry,8 but exact durations of the
Goldmann examinations were not reported.
Half of the patients preferred the new SKP method over
Goldmann kinetic perimetry, and 16% found no difference
between the methods. It seems that SKP is acceptable to
patients with advanced VF loss who are experienced in
perimetric examinations.
This is the first study comparing the variety of advanced
VF defects obtained with SKP and classic Goldmann MKP.
Results indicate that SKP VFs are comparable to those of
conventional MKP techniques. Semiautomated kinetic perimetry represents a more standardized method of kinetic
perimetry, which enables quantitative assessment of the VF
area. The high agreement between MKP and SKP does not
indicate superiority of one method. Because the main methodological difference between the procedures is stimulus
control, the lack of agreement is likely to be caused by the
imprecision of the manual stimulus movement in the Goldmann instrument.
We believe that SKP may be used in clinical practice in
diagnosing and monitoring advanced VF loss as a standardized form of kinetic perimetry and may provide a new and
improved assessment of profound VF defects alternative to
traditional Goldmann MKP. Because it gives the advantages
of eliminating the examiner’s RT and providing quantitative
assessment of the VF loss, it may be used for quantitative
follow-up of VF defects. Further studies comparing the
repeatability of MKP and SKP may be helpful in the final
evaluation of both methods.
References
1. Goldmann H. Demonstration unseres neuen Projektionskugelperimeters samt theoretischen und klinischen Bemerkungen über
Perimetrie. Ophthalmologica 1946;11:187–92.
2. Goldmann H. Ein selbstregistrierendes Projektionskugelperimeter. Ophthalmologica 1945;109:71–9.
3. Grover S, Fishman GA, Brown J Jr. Patterns of visual field
progression in patients with retinitis pigmentosa. Ophthalmology 1998;105:1069 –75.
4. Chauhan BC, Drance SM. The relationship between intraocular pressure and visual field progression in glaucoma.
Graefes Arch Clin Exp Ophthalmol 1992;230:521– 6.
5. Gloor B, Vollrath C. Hemianope Gesichtsfeldausfälle—Untersuchungsmethoden und Lokalisationsprobleme. Klin
Monatsbl Augenheilkd 1988;192:507–17.
6. Heijl A, Drance SM. Changes in differential threshold in
patients with glaucoma during prolonged perimetry. Br J
Ophthalmol 1983;67:512– 6.
7. Szatmáry G, Biousse V, Newman NJ. Can Swedish Interactive
Thresholding Algorithm fast perimetry be used as an alternative to Goldmann perimetry in neuro-ophthalmic practice?
Arch Ophthalmol 2002;120:1162–73.
8. Beck RW, Bergstrom TJ, Lichter PR. A clinical comparison of
visual field testing with a new automated perimeter, the Humphrey Field Analyzer, and the Goldmann perimeter. Ophthalmology 1985;92:77– 82.
9. Trobe JD. The Neurology of Vision. New York: Oxford
University Press; 2001.
10. Greve EL. Single and multiple stimulus static perimetry in
glaucoma; the two phases of perimetry. Doc Ophthalmol
1973;36:1–346.
11. Berry V, Drance SM, Wiggins RL. An evaluation of differences between two observers plotting and measuring visual
fields. Can J Ophthalmol 1966;1:297–300.
12. Fankhauser F, Giger H, Lotmar W. A new attachment for the
Goldmann perimeter. Its precision, limitations and clinical
applications. Am J Ophthalmol 1966;62:885–92.
13. Schiefer U, Schiller J, Dietrich TJ, et al. Evaluation of advanced visual field loss with computer-assisted kinetic perimetry. In: Wall M, Mills RP, eds. Perimetry Update 2000/2001.
The Hague: Kugler Publishers; 2001:131– 6.
14. Schiefer U, Schiller J, Paetzold J, et al. Evaluation ausgedehnter Gesichtsfelddefekte mittels computerassistierter kinetischer Perimetrie. Klin Monatsbl Augenheilkd 2001;218:13–20.
15. Schiefer U, Strasburger H, Becker ST, et al. Reaction time in
automated kinetic perimetry: effects of stimulus luminance,
eccentricity, and movement direction. Vision Res 2001;41:
2157– 64.
16. Schiefer U, Rauscher S, Paetzold J, Schiller J. Realisation of
semi-automated kinetic perimetry (SKP) with Interzeag 101
instrument. In: Wall M, Mills RP, eds. Perimetry Update
2002/2003. The Hague: Kugler Publishers; 2003:233– 8.
17. Paetzold J, Schiller J, Rauscher S, Schiefer U. A computer
application for training kinetic perimetry. In: Wall M, Mills
RP, eds. Perimetry Update 2002/2003. The Hague: Kugler
Publishers; 2003:69 –73.
18. Aulhorn E, Karmeyer H. Frequency distribution in early glaucomatous visual field defects. Doc Ophthal Proc Series 1977;
14:17– 83.
19. Anderson DR. Testing the Field of Vision. St. Louis: Mosby;
1982.
20. Bland JM, Altman DG. Statistical methods for assessing
agreement between two methods of clinical measurement.
Lancet 1986;1:307–10.
1353
Ophthalmology Volume 112, Number 8, August 2005
21. Zingirian M, Gandolfo E, Orciuolo M. Automation of the Goldmann perimeter. Doc Ophthalmol Proc Series 1983;42:103–7.
22. Zingirian M, Gandolfo E, Capris P, Mattioli R. Computerized
system for static and kinetic automatic perimetry. Eur J Ophthalmol 1991;1:181– 6.
23. Wabbels B, Kolling G. Automatische kinetische Perimetrie
mit dem Twinfield Perimeter. Z Prakt Augenheilkd 1999;20:
401– 6.
24. Miller KN, Shields MB, Ollie AR. Automated kinetic perimetry with two peripheral isopters in glaucoma. Arch Ophthalmol 1989;107:1316 –20.
25. Ballon BJ, Echelman DA, Shields MB, Ollie AR. Peripheral
1354
26.
27.
28.
29.
visual field testing in glaucoma by automated kinetic
perimetry with the Humphrey Field Analyzer. Arch Ophthalmol 1992;110:1730 –2.
Esterman B. Grid for scoring visual fields—II. Perimeter.
Arch Ophthalmol 1968;79:400 – 6.
Drance SM, Berry V, Hughes A. Studies in the reproducibility
of visual field areas in normal and glaucomatous subjects. Can
J Ophthalmol 1966;1:14 –23.
Johnson CA, Keltner JL. Optimal rates of movement for
kinetic perimetry. Arch Ophthalmol 1987;105:73–5.
Lachenmayr BJ, Vivell PMO. Perimetrie. Stuttgart: Thieme;
1992.