Open Perimeter Interface (OPI)

The Open Perimetry Interface (OPI):Glasnost and Perestroika in the Visual Field

Andrew Turpin, MelbournePaul Artes, HalifaxNorth American Perimetric Society (NAPS), New York, 2011

In 1986, the Soviet Union was facing economic, social, and moral ruin.

Michael Gorbatchov came along and proposed

“glasnost” - openness, transparency “perestroika” - rebuilding, reconstruction

By 2010, perimetry has seriously fallen behind imaging in terms of innovation and excitement.

There has been no major innovation since SITA(1997).

The field is isolated from mainstream science (physics, psychology, medicine)

Commercial players do their own thing. Clinicians and scientists work largely within their own groups using laboratory equipment. A vast resource of creativity lies unharnessed and frustrated.

It feels very different from Tübingen in 1986...

We need to:

Creatively seek new ways to examine visual fields.

The center as well as the peripheryWith static as well as kinetic stimuliWith threshold as well as suprathreshold strategiesWith spots as well as with new types of stimuli

Efficiently collaborate with each other, and with industry.

Quickly translate new ideas into experiments.

Quickly translate experiments into clinical tools.

Many current problems of perimetry have long been addressed in other disciplines. Perimetry is at least two to three decades behind the state-of-the-art.

Many talented scientists would like to work on clinically relevant problems, but lack the tools.

Industry makes beautiful hardware, but this is inaccessible for basic experiments.

Accessing clinical data is often like breaking into the Pentagon. There are laudable exceptions.

Variability and Reaction Times in Normals and in Patients With Glaucoma 879

perimetrically normal areas.8 In a study using conven-tional automated perimetry, Flammer and colleagues9

reported a significant relationship of a prolonged RTwith an increase in threshold.

If reaction time is to be used as an outcome mea-sure in conventional automated perimetry, the rela-tionship of RT to threshold must be defined. If thereis a constant relationship between reaction time andthreshold in health and disease, one would hope thisrelationship could be used to help predict thresholdin a more accurate and less variable way. Our goal wasto determine this relationship of reaction time to thepsychometric function in normal subjects, normal sen-sitivity test locations in patients with glaucoma, andtest locations with 10 to 20 dB loss in patients withglaucoma, as well as to investigate the relationshipsuggested by Flammer and coworkers9 of a prolongedRT associated with an increase in threshold in subjectswith glaucoma.

METHODS

SubjectsTen patients with well-established primary open angleglaucoma and 10 normal volunteers gave informedconsent to participate in the study. The protocol wasapproved by the University of Iowa Institutional Re-view Board. The tenets of the Declaration of Helsinkiwere followed. The normals were paid volunteers whowere hospital employees or friends or family membersof eye clinic patients. The normal subjects werematched pairwise to patients by age within 5 years.Normals were included if they had no history of eyedisease except refractive error and normal results ofophthalmologic examination. They all had normal au-tomated perimetry results using the Humphrey VisualField Analyzer (HVFA; Humphrey Instruments, SanLeandro, CA), program 24-2. If a potential normalsubject had three or more adjacent points with a totaldeviation score at the P < 0.05 level or two adjacentpoints with a total deviation score at the P < 0.05 levelor two adjacent points with one at the P < 0.01 levelwith STATPAC (Humphrey Instruments),10 they wereexcluded. Subjects also were excluded if the meandeviation index was outside the 95% confidencebound for normals. All normals had normal resultson the Glaucoma Hemifield Test.''

Patients with glaucoma were recruited from previ-ous studies in which they had taken HVFA tests thatshowed test locations in one eye with visual field dam-age in the 10- to 20-dB range (P < 0.005 for pointwisetotal deviation with STATPAC analysis) and test loca-tions with normal results (P > 0.05) with the totaldeviation plot. All patients with glaucoma had under-gone threshold automated perimetry within the pre-ceding 2 months. All patients with the clinical diagno-

sis of primary open angle glaucoma had open angleson gonioscopy. They had an intraocular pressuregreater than 21 mm Hg during their course, alongwith glaucomatous optic disc changes, glaucomatousvisual field defects, and no other apparent mechanismof glaucoma. All patients were receiving treatment forintraocular pressure control, but none were using amiotic. Patients with primary open angle glaucomawere excluded if they had any other disease known tocause visual loss. All subjects had corrected visual acu-ity of 20/25 or better, pupil diameter of at least 3 mmwhen tested, spectacle correction not exceeding 6.00D (equivalent sphere), and previous experience withautomated perimetric examinations.

Testing StrategyConventional automated perimetry was first per-formed with the HVFA with program 24-2 using themanufacturer's recommendations (except for threesubjects in whom program 30-2 was performed).These programs test the central 21° or 27° of the visualfield with stimuli spaced 6° apart; in addition, the 24-2 program tests one stimulus above and one stimulusbelow the nasal horizontal at 27°. We used a Gold-mann size III (4 mm2) object on a 31.5-asb back-ground. The size of the stimulus was fixed, and thethreshold to differential light intensity was estimatedat each test point with a staircase procedure. Eachpatient's appropriate near correction was used. Restbreaks were given when requested.

Frequency of seeing curves were measured by con-trolling the HVFA with a custom program run by apersonal computer (Hewlett Packard [Palo Alto, CA]Vectra 486, 33 mHz). At two different test locations,stimuli were presented in 2-dB intervals up to andincluding 10 dB from either side of the estimated(HVFA program 24-2) threshold, with 15 repetitionsat each stimulus intensity. All presentations of stimulusintensity and location were randomized. To determinefalse-positive and false-negative responses, 60-dB and0-dB stimuli were presented 20 times at each location.Therefore, each location was tested a total of 205times. To prevent subjects from concentrating theirentire attention on the two tested points, eight ran-dom additional locations of normal or near normalvision were tested with three repetitions of the 0-dBstimulus for a total of 24 extra trials. These 434 trialsproduced approximately a 30-minute test. This con-straint was added to simulate usual clinical testing timeof two 30-2 tests. All subjects were asked to respondas quickly as possible to the stimuli and were remindedtwo more times during testing to respond to the stim-uli as quickly as possible. Reaction time for each trialwas measured as the time between the beginning ofthe 200-msec stimulus presentation to the subject'sbutton-pressing response. If the subject did not re-spond within 2 seconds, the stimulus was considered

The Psychometric Function and Reaction Times ofAutomated Perimetry in Normal and Abnormal Areas ofthe Visual Field in Patients With Glaucoma

Michael Wall,* RichardJ. Maw* Kim E. Stanek* and Balwantray C. Chauhanf

Purpose. To study the relationship of reaction time to the psychometric function in normalsubjects, normal sensitivity test locations in patients with glaucoma, and test locations with 10to 20 dB loss in patients with glaucoma.Methods. The authors tested 10 patients with glaucoma and 10 age-matched normal volunteerswith the Humphrey perimeter, first with program 24-2 and then with the method of constantstimuli to generate frequency of seeing curves. At two widely separated visual field locationson the program 24-2 grid, they presented stimuli in 2-dB intervals, 10 dB either side of theprogram 24-2 threshold, at 0 dB and 60 dB (15 repetitions per intensity). For the patientswith glaucoma, they chose a visual field location with normal sensitivity and a location in anarea of 10 to 20 dB loss.Results. Analysis of variance with post hoc <-tests showed that reaction time (RT) at the 0-dBintensity was prolonged by approximately 90 msec in the abnormal sensitivity test location ofpatients with glaucoma compared to the control and the glaucoma normal sensitivity groups(P < 0.0001). However, this difference was accounted for by only 4 of the 10 patients withglaucoma, reaching 100% of stimuli seen with the brightest stimulus at the moderately dam-aged test location. Reaction time at the frequency of seeing 50% estimated threshold showedno significant differences among the groups. Prolongation of RT from the 0-dB value wasanalyzed as a function of increasing attenuation of stimulus intensity. The results fit theequation RT = a + b(Intensityf for all groups.Conclusions. There is no significant difference in RT between normal subjects and patientswith glaucoma either at threshold or to suprathreshold stimuli. Reaction time increases aftera power function with increasing attenuation of stimulus intensity up to the threshold. InvestOphthalmol Vis Sci. 1996; 37:878-885.

Axi analysis of the psychometric function or the "fre-quency of seeing curve" has been used in three studiesin patients with glaucoma.1"3 Weber and Rau1 showedthat the transition zone between not seen and seen ofthe frequency of seeing curve varied by as much as afactor of 17 in patients with glaucoma.1 Chauhan andcolleagues,2 in 70 subjects at four test locations, usedfive repetitions per intensity to investigate variabilityin patients with glaucoma. They found that with a

From the * Veterans Administration Hospital and the Departments of Neurology andOphthalmology, University of Iowa, Iowa City, and the f Department ofOphthalmology, Dalhousie University, Halifax, Canada.Supported by VA Merit Review, an unrestricted grant to the Department ofOphthalmology from Research to Prevent Blindness (New York, NY) (MW), andMedical Research Council of Canada grant #MT-11357 (BCC)Submitted for publication June 30, 1995; revised November 6, 1995; acceptedDecember 15, 1995.Proprietary interest category: N.Reprint requests: Michael Wall, Department of Neurology, College of Medicine,University of Iowa, 200 Hawkins Drive #2007 RCP, Iowa City, IA 52242-1053.

decrease in sensitivity there was an increase in variabil-ity, flattening of the slope of the frequency of seeingcurve (increase in standard deviation), and worseningof the goodness of fit. Lynn and coworkers3 also usedfrequency of seeing curves in patients with glaucoma.They focused their analysis on internal inconsistenciesof the automated perimetry test procedure.3

Unlike variability, reaction time (RT) in perimetryhas received little attention. In nonperimetric psycho-physical studies, RT has been shown to increase witheccentricity and with decreasing light intensity.4"7 Re-gan and coworkers8 tested the RT in patients withoptic neuritis by stimulating corresponding retinal lo-cations and varying the delay of onset of a light stimu-lus in one of the eyes. They reported prolongation ofRT in the eye with optic neuritis. In addition to theabnormal areas being complimentary to the deficitsfound with kinetic perimetry, they found RT delays in

878Investigative Ophthalmology & Visual Science, April 1996, Vol. 37, No. 5Copyright © Association for Research in Vision and Ophthalmology

Characteristics of Frequency-of-Seeing Curves in NormalSubjects, Patients With Suspected Glaucoma,and Patients With GlaucomaBalwantray C. Chauhan* James D. Tompkins,-\ Raymond P. LeBlanc*and Terry A. McCormick*

Purpose. The authors performed this study to determine factors that affect the characteristicsof frequency-of-seeing curves in normal subjects, patients with suspected glaucoma, and pa-tients with glaucoma.Methods. The sample consisted of 70 subjects (22 normal subjects, 12 patients with suspectedglaucoma, and 36 patients with glaucoma). A program was written to interface with theHumphrey Field Analyzer (Humphrey Instruments, San Leandro, CA) to measure frequency-of-seeing curves. The authors presented stimuli 8 dB either side of the estimated threshold in1-dB intervals with five repetitions at each stimulus intensity. The authors tested four to sixlocations in each subject, with randomization of the stimulus intensity and location. Fixationwas monitored with the Heijl-Krakau method. Using a probit program, the authors calculatedthe threshold and slope (estimated by the interquartile range) of each curve.Results. The authors obtained 124 curves from the normal subjects, 71 from the patients withsuspected glaucoma, and 183 from the patients with glaucoma. In all three groups, the slope ofthe frequency-of-seeing curve was correlated highly with the threshold or threshold deviation,although the correlation was significantly higher in the normal subjects compared with thepatients with suspected glaucoma and patients with glaucoma, even after controlling for therange of the threshold and threshold deviation. For this reason, the authors found consider-ably different frequency-of-seeing curves, between subject groups and also within the group ofpatients with glaucoma, in locations with the same threshold .Conclusions. There may be fundamental differences in areas of normal subjects and patientswith glaucoma with similar thresholds or threshold deviations. These differences also may existwithin patients with glaucoma. Invest Ophthalmol Vis Sci. 1993;34:3534-3540.

Jb requency-of-seeing curves describe the relationshipbetween the probability of seeing a stimulus and a stim-ulus property such as contrast or size. They can bethought of as cumulative Gaussian functions that de-pict local threshold variability.1 Frequency-of-seeingcurves measured with stimuli used in conventional pe-rimetry have been described previously in normal sub-

From the ^Department of Ophthalmology, Dalhousie University, and ^Departmentof Electrical Engineering, Technical University of Nova Scotia, Halifax, NovaScotia, Canada.Supported in part by research grant MT-J1357 from the Medical Research Councilof Canada (BCC).Submitted for publication November 25, 1992; accepted May 19, 1993.Proprietary interest category: N.Reprint requests: Balwantray C. Chauhan, Nova Scotia Eye Centre, Camp HillMedical Centre, 1335 Queen Street, Halifax, Nova Scotia, Canada B3J 2H6.

jects2'3 and patients with glaucoma,34 but the numberof subjects generally has been small.

The characteristics of frequency-of-seeing curveshave implications not only for perimetric thresholdsand their variability, but also for modeling perimetricresponses in simulation experiments. Previous studiesdescribing such experiments have assumed that thethreshold and threshold variability are related lin-early,56 with lower thresholds (or higher sensitivities)yielding lower variability. In frequency-of-seeingcurves, a steep slope reflects low threshold variability,whereas a more shallow slope reflects higher variabil-ity. Perimetric responses in simulation experimentsgenerally are determined with the slope of the fre-quency-of-seeing curve because the probability of see-ing a stimulus of a given intensity is known. Perimetric

3534Investigative Ophthalmology & Visual Science, December 1993, Vol. 34, No. 13Copyright © Association for Research in Vision and Ophthalmology

Characteristics of Frequency-of-Seeing Curves 3535

responses in normal subjects can be modeled theoreti-cally with considerable accuracy because the slope ofthe frequency-of-seeing curve can be described by thelocal threshold, visual field location, and age.2 It is notclear whether the same applies to responses in patientswith glaucoma.

We wanted to perform a comprehensive study ofthe characteristics of frequency-of-seeing curves innormal subjects, patients with suspected glaucoma,and patients with glaucoma. We also planned to deter-mine whether the same type of theoretic modeling asthat used to simulate normal visual fields could be ap-plied to glaucomatous fields. We particularly wantedto compare frequency-of-seeing curves obtained at lo-cations in patients with glaucoma and normal subjectswith similar thresholds and similar threshold devia-tions from age-corrected normal values.

SUBJECTS AND METHODS

SubjectsThe study sample consisted of 22 normal subjects(mean age, 54.27 years; range, 34 to 76 years), 12 pa-tients with suspected glaucoma (mean age, 54.83years; range, 36 to 77 years), and 36 patients withopen-angle glaucoma (mean age, 66.78 years; range,36 to 86 years). The normal subjects were recruitedfrom a group of subjects undergoing other psycho-physical studies. All normal subjects had normal visualfields defined by the Statpac program of theHumphrey Field Analyzer (Humphrey Instruments,San Leandro, CA)7 and a negative family history ofglaucoma. The patients with suspected glaucoma andthose with glaucoma were recruited consecutivelyfrom the clinical practice of one of us (RPL). Patientswith suspected glaucoma had normal visual fields asdefined above; however, 10 of 12 of these patients hadconsistently increased intraocular pressures (> 22 mmHg), whereas the remaining 2 had suspect optic diskson the grounds of asymmetry with the fellow eye. Pa-tients with glaucoma had glaucomatous visual fields(with early to advanced damage) and optic disks. Allsubjects had an ophthalmic examination, visual acuityof 6/9 or better, previous experience with perimetricexaminations, pupil diameter of at least 3 mm whentested, and a spectacle correction not exceeding 4.00diopters (equivalent sphere). The study was approvedby the Camp Hill Medical Centre Research EthicsCommittee and followed the tenets of the Declarationof Helsinki. Informed consent was obtained from eachsubject.

Testing Methods

The Humphrey Field Analyzer can be controlled ex-ternally by a personal computer through a series of

commands. We have written a computer program thatuses these commands and allows us to conduct manycustomized tests, including the measurement of fre-quency-of-seeing curves.

In normal subjects and patients with suspectedglaucoma, we tested locations along the 45, 135, 225,or 315 meridian, chosen randomly. In patients withglaucoma, we tested locations in scotomas and areaswith normal thresholds. Only one eye of each subjectwas tested. In most subjects we tested six locations,although in a small number of patients with glaucomawe tested four or five locations. All locations weretested with a white Goldmann size III stimulus (visualangle = 0.43) on a white background of 31.5 asb.

We first used the computer program to carefullydetermine the center of the blind spot. The thresholdsat the chosen test locations then were estimated with astandard 4-2 staircase procedure.8 We then presentedstimuli within a range of 8 dB either side of the esti-mated threshold in 1-dB steps, with five trials at eachstimulus intensity. Therefore, there were 85 trials perlocation. The location and stimulus intensity were ran-domized during testing. Fixation was monitored withthe Heijl-Krakau technique,9 with a maximum lumi-nosity stimulus (10,000 asb) presented in the mea-sured center of the blind spot every fifth presentationduring the first 50 presentations, and every twenty-fifth presentation thereafter. Numerically, the rate offixation losses was the percentage of blind spot presen-tations reported by the subject over the total numberof blind spot presentations.

Data AnalysisThe frequency-of-seeing at each stimulus intensity wascomputed. These data were subjected to a probit analy-sis, which fitted a frequency-of-seeing curve to thedata. The threshold was taken as the stimulus intensitycorresponding to the 50% frequency-of-seeing of thefitted curve. The threshold deviation was computed bysubtracting the threshold from the age-corrected nor-mal value for that specific location. The slope of thecurve was estimated by calculating the interquartilerange (stimulus intensity interval corresponding to25% to 75% frequency-of-seeing) of the fitted curve.Figure 1 shows an example of a frequency-of-seeingcurve with the calculated parameters.

Correlations between variables were examinedwith the Spearman correlation. Differences betweencorrelation coefficients were determined with stan-dard tests.10 An analysis of covariance was used to ana-lyze the pooled data. Residuals of the regressions be-tween the dependent variables and covariates weretested for normality and homoscedasticity. Where nec-essary, log transforms were performed. All statisticaltests were two tailed, and statistical significance was setat P< 0.05.

Characteristics of Frequency-of-Seeing Curves 3535

responses in normal subjects can be modeled theoreti-cally with considerable accuracy because the slope ofthe frequency-of-seeing curve can be described by thelocal threshold, visual field location, and age.2 It is notclear whether the same applies to responses in patientswith glaucoma.

We wanted to perform a comprehensive study ofthe characteristics of frequency-of-seeing curves innormal subjects, patients with suspected glaucoma,and patients with glaucoma. We also planned to deter-mine whether the same type of theoretic modeling asthat used to simulate normal visual fields could be ap-plied to glaucomatous fields. We particularly wantedto compare frequency-of-seeing curves obtained at lo-cations in patients with glaucoma and normal subjectswith similar thresholds and similar threshold devia-tions from age-corrected normal values.

SUBJECTS AND METHODS

SubjectsThe study sample consisted of 22 normal subjects(mean age, 54.27 years; range, 34 to 76 years), 12 pa-tients with suspected glaucoma (mean age, 54.83years; range, 36 to 77 years), and 36 patients withopen-angle glaucoma (mean age, 66.78 years; range,36 to 86 years). The normal subjects were recruitedfrom a group of subjects undergoing other psycho-physical studies. All normal subjects had normal visualfields defined by the Statpac program of theHumphrey Field Analyzer (Humphrey Instruments,San Leandro, CA)7 and a negative family history ofglaucoma. The patients with suspected glaucoma andthose with glaucoma were recruited consecutivelyfrom the clinical practice of one of us (RPL). Patientswith suspected glaucoma had normal visual fields asdefined above; however, 10 of 12 of these patients hadconsistently increased intraocular pressures (> 22 mmHg), whereas the remaining 2 had suspect optic diskson the grounds of asymmetry with the fellow eye. Pa-tients with glaucoma had glaucomatous visual fields(with early to advanced damage) and optic disks. Allsubjects had an ophthalmic examination, visual acuityof 6/9 or better, previous experience with perimetricexaminations, pupil diameter of at least 3 mm whentested, and a spectacle correction not exceeding 4.00diopters (equivalent sphere). The study was approvedby the Camp Hill Medical Centre Research EthicsCommittee and followed the tenets of the Declarationof Helsinki. Informed consent was obtained from eachsubject.

Testing Methods

The Humphrey Field Analyzer can be controlled ex-ternally by a personal computer through a series of

commands. We have written a computer program thatuses these commands and allows us to conduct manycustomized tests, including the measurement of fre-quency-of-seeing curves.

In normal subjects and patients with suspectedglaucoma, we tested locations along the 45, 135, 225,or 315 meridian, chosen randomly. In patients withglaucoma, we tested locations in scotomas and areaswith normal thresholds. Only one eye of each subjectwas tested. In most subjects we tested six locations,although in a small number of patients with glaucomawe tested four or five locations. All locations weretested with a white Goldmann size III stimulus (visualangle = 0.43) on a white background of 31.5 asb.

We first used the computer program to carefullydetermine the center of the blind spot. The thresholdsat the chosen test locations then were estimated with astandard 4-2 staircase procedure.8 We then presentedstimuli within a range of 8 dB either side of the esti-mated threshold in 1-dB steps, with five trials at eachstimulus intensity. Therefore, there were 85 trials perlocation. The location and stimulus intensity were ran-domized during testing. Fixation was monitored withthe Heijl-Krakau technique,9 with a maximum lumi-nosity stimulus (10,000 asb) presented in the mea-sured center of the blind spot every fifth presentationduring the first 50 presentations, and every twenty-fifth presentation thereafter. Numerically, the rate offixation losses was the percentage of blind spot presen-tations reported by the subject over the total numberof blind spot presentations.

Data AnalysisThe frequency-of-seeing at each stimulus intensity wascomputed. These data were subjected to a probit analy-sis, which fitted a frequency-of-seeing curve to thedata. The threshold was taken as the stimulus intensitycorresponding to the 50% frequency-of-seeing of thefitted curve. The threshold deviation was computed bysubtracting the threshold from the age-corrected nor-mal value for that specific location. The slope of thecurve was estimated by calculating the interquartilerange (stimulus intensity interval corresponding to25% to 75% frequency-of-seeing) of the fitted curve.Figure 1 shows an example of a frequency-of-seeingcurve with the calculated parameters.

Correlations between variables were examinedwith the Spearman correlation. Differences betweencorrelation coefficients were determined with stan-dard tests.10 An analysis of covariance was used to ana-lyze the pooled data. Residuals of the regressions be-tween the dependent variables and covariates weretested for normality and homoscedasticity. Where nec-essary, log transforms were performed. All statisticaltests were two tailed, and statistical significance was setat P< 0.05.

1990: The Humphrey Field Analyzer can be controlled from an external PC (“Gateway”).

1993 1996

Open Perimeter Interface

Open Perimeter Interface2006: At ARVO, Turpin, Artes, McKendrick discuss the idea of an “open-source perimeter”:

A new instrument that can serve as platform for experiments and clinical studies, and is open to all.

But: excellent hardware already exists (Zeiss, Haag-Streit, Heidelberg, Oculus, Tinsley, Medmont, etc).

2010: Octopus Research Group: Uli Schiefer (Tübingen) & Matthias Monhart (Haag-Streit)First draft of OPI by Turpin (with Dietzsch and Demirel)

Open Perimeter Interface

OPI – Open Perimeter Interface

Version 0.3

1 Preamble

This document describes a standard set of R functions for interfacing with a perimeter (an instru-ment for examining visual fields). It began existence at the First Octopus Research Meeting heldin Tubingen in July 2010, which was hosted by Prof. Ulrich Schiefer (University of Tubingen),and Matthias Monhart (Haag-Streit). R code that implements this interface should provide theset of functions described.

1.1 Document History

0.0 3 Jul 2010 Began in Hotel Hospiz, Tubingen by Andrew Turpin.0.1 7 Jul 2010 Complete rewrite with feedback from Janko Dietzsch and Shaban

Demirel.0.2 17 May 2011 Redraft by Andrew Turpin based on extensive feedback over a few rounds

from Paul Artes and Shaban Demirel.0.3 9 July 2011 image added to stimuli and a few errors in examples fixed by Turpin.

1.2 Document Future

It is expected that this document will be revised at meetings of the Imaging and Perimetry Society.

1.3 Conventions

All (x, y) coordinates are Cartesian relative to the fixation point (0, 0) in degrees of visual angle(not radians). Positive y-coordinates refer to stimuli in the superior field. Positive x-coordinatesrefer to locations “east” of fixation (temporal for the Right eye, nasal for the Left).

For “traditional” perimetry, where stimuli are projected spots, the diameter of the spot is indegrees of visual angle, hence a Goldmann Size III would have size = 0.43. The intensity of thespot is in cd/m2, thus intensity is the luminance of the background plus the stimulus, not anincrement, and not a “dB value”. When the stimuli is an image, size becomes a scaling parameterand intensity is specified in the image itself.

All times are in milliseconds.

1.4 Usage

An implementation of this interface should support the basic types and functions below. For moresophisticated stimuli (eg ramping of a static stimulus), subclass the existing types and write yourown presentation methods.

1

Open Perimeter InterfaceJune 2011: PHA and AT (via skype) meet with engineers at Haag-Streit in Berne, Switzerland, to discuss the implementation

It finally becomes clear how exactly to do this:

1) The OPI specifies a small set of low-level commands2) Manufacturers implement in any way they like (eg. Java, C++, Python), and make available (free, or for $)

3) Researchers call the OPI functions from any programming language they like. (We like )

Open Perimeter InterfaceOpen Perimeter Interface2 Data Types

opiStaticStimulus

stim <- list(x, y, image=NA, level, size=0.43, color="white",duration=200, responseWindow=1500, ...)

class(stim) <- "opiStaticStimulus"

x x coordinate of the center of stimulus in degrees relative to fixation

y y coordinate of the center of stimulus in degrees relative to fixation

image an image to display in a machine specific format

level stimulus level in cd/m2(ignored if !is.na(image))

size diameter of target in degrees, or scaling factor for image if specified

color machine specific stimulus color settings (ignored if !is.na(image))duration total stimulus duration in milliseconds

responseWindow maximum time (>= 0) in milliseconds to wait for a response from the

onset of the stimulus presentation

... machine specific parameters

opiTemporalStimulus

stim <- list(x, y, image=NA, lut, size=0.43, color="white", rate, duration,responseWindow=1500, ...)

class(stim) <- "opiTemporalStimulus"

x x coordinate of the center of stimulus in degrees

y y coordinate of the center of stimulus in degrees

image an image to display in a machine specific format

lut if is.na(image) then this is a lookup table (vector) for stimulus level

at each step of rate Hz in cd/m2. If image is specified, then this is a

list of images, in the same format as image, that is stepped through at

rate Hz.

size diameter of target in degrees, or scaling factor for image if specified

color machine specific stimulus color settings (ignored if !is.na(image))rate frequency with which lut is processed in Hz

duration total length of stimulus flash in milliseconds. There is no guarantee that

duration mod |lut|/rate == 0. That is, the onus is on the user to

ensure the duration is a multiple of the period of the stimuli.

responseWindow maximum time (>= 0) in milliseconds to wait for a response from the

onset of the stimulus presentation

... machine specific parameters

2

opiInitialize (“Oct900”)opiSetBackground (10, “white”)stim <- list(list (x=-3, y=-3, level=318, size=0.43, color="white"))class (stim) <- "opiStaticStimulus"result <- opiPresent (stim)

Fifty-six years ago the Allies liberated Europe.

Now, the OPI can liberate visual field research (and more). Join us. There is much space on this truck.

eg: Psychometric functions for static perimetry (stimulus size, contrast, duration).

What we can do now...

stimulus area, mm^2

asb

1/16 1/4 1 4 16 64 256

110

100

1000

10000

Gol

dman

n I

Gol

dman

n II

Gol

dman

n III

Gol

dman

n IV

Gol

dman

n V

0.25

1 4 16 640.0625

0.125

0.5

1.41

2 2.83

5.66

8 11.3

22.6

32 128

256

mm^2

Thank you!

The way forward...