the reliability of a video-enhanced hirschberg test under clinical

The Reliability of a Video-Enhanced Hirschberg TestUnder Clinical Conditions

Satoshi Hasebe, Hiroshi Ohtsuki, Yasunori Tadokoro, Masaki Okano,and Takashi Furuse

Purpose. To examine the reliability and usefulness of a video-based Hirschberg test underclinical conditions.

Methods. The authors estimated ocular deviation in 87 patients with strabismus through auto-mated analysis of corneal reflex displacement using a video refractor. The reproducibility ofmeasurement, the comparison with the prism and alternate cover test (PACT), and thedistribution of the Hirschberg ratio were investigated.

Results. The 95% limits of agreement of the video-based Hirschberg test evaluated by repeatedmeasurements were ±0.18 mm (equivalent to ±2.2° or ±3.8 prism diopters [PD] of calculatedstrabismic deviation) for the horizontal deviation and ±0.28 mm (equivalent to ±3.4° or ±5.9PD) for the vertical deviation. The 95% limits of agreement between the Hirschberg measuresand the PACT were within ±7.8° or ±13.7 PD. The average (± SD) Hirschberg ratio was 12.3± 1.2°/mm or 21.8 ± 2.1 PD/mm.

Conclusions. The video-enhanced Hirschberg measurement shows good reproducibility andease of application, even in the testing of infants. In quantitative analysis, however, systematicmeasurement error resulting from intersubject variance of the Hirschberg ratio should betaken into consideration. Invest Ophthalmol Vis Sci. 1995;36:2678-2685.

1 he Hirschberg test consists of the estimation of thestrabismic angle based on the decentration of the cor-neal light reflex produced by a small point source oflight, the first Purkinje image. This simple test is usedfrequently when inadequate cooperation or reducedvision precludes the use of other methods.12

To obtain the strabismic angle (in degrees orprism diopters [PD]), the observer multiplies the esti-mated decentration (in millimeters) of the cornealreflex by some suitable conversion factor termed theHirschberg ratio. Recent investigations have indicatedthat this ratio is 12°/mm or 21 PD/mm.3"" Accord-ingly, the displacement of the corneal reflex per de-gree is too small to determine precisely with naked eyeobservation. Putnam12 also noted that the precision in

From the Department of Ophthalmology, Okayama University Medical School,Okayamtt, Japan.Supported Iry a Cranl-in-Aid. for Scientific Research from the Japanese Ministry ofEducation, Science and Culture (no. 07671920).Submitted for publication November 8, 1994; revised August 21, 1995; acceptedAugust 22, 1995.Proprietaiy interest category: N.Reprint requests: Saloshi Hasebe, Department of Ophthalmology, OkayamaUniversity Medical School, 2-5-1, Shikata-cho, Okayama 700, Japan.

assessment of the corneal reflex position was greaterthan 0.5 to 1 mm. As a consequence, the Hirschbergtest generally has been acknowledged to provide onlya gross approximation of the strabismic deviation andnot to be suitable for quantitative purpose.12I2

In some types of strabismus, early surgical align-ment of the eyes within the first or second year of lifeis thought to be prerequisite for normal developmentof binocular visual function.13'14 Therefore, an easy toapply but more reliable technique, on the basis ofwhich surgical dosage can be decided, is now required.

Recently, the strategy of finding the corneal reflexposition on photographic3"9'15'"' or videographic10'17

images of the eyes was proposed. In experimental stud-ies, the precision of these methods was found to bewithin a few degrees,6"8'10'15"17 and they are thus nota-bly more precise than the traditional Hirschberg test.12

For application in clinical examination, we were par-ticularly interested in the development of the Hirsch-berg test through video image processing because itprovides immediate results.

Some geometric analyses51118 have suggested thatthere is variation among subjects for the Hirschberg

2678Investigative Ophthalmology & Visual Science, December 1995, Vol. 36, No. 13Copyright © Association for Research in Vision and Ophthalmology

Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/journals/iovs/933408/ on 03/26/2018

Video-enhanced Hirschberg Test 2679

TABLE l. Diagnoses of Patients Using aVideo-Enhanced Hirschberg Test

Diagnosis Number (%)

Intermittent exotropia 32 (36.8)Microtropia 13 (14.9)Infantile esotropia 10 (11.5)Acquired esotropia 8 (9.2)Constant exotropia 7 (8.0)Accommodative esotropia 7 (8.0)Abducens nerve palsy 3 (3.4)Divergent palsy 2 (2.3)Duane's syndrome 2 (2.3)Overaction of inferior oblique muscles 2 (2.3)Consecutive esotropia 1 (1.1)Total 87

ratio because it depends on the two biometric parame-ters of the corneal curvature and anterior chamberdepth. Because the intersubject variance of theHirschberg ratio can be a source of systematic mea-surement error, we cannot overlook this problem totake full advantage of the precision provided by theabove-mentioned image analysis.

In the current study, we carried out the video-based Hirschberg test under clinical conditions usinga commercially available video refractor. We then veri-fied the reliability of this test by assessing the repro-ducibility of the measurement, the degree ofagreement between this test and the prism and alter-nate cover test (PACT), and the distribution of theHirschberg ratio.

METHODS

Subjects

Our subjects consisted of 87 patients with strabismus(Table 1; mean age, 17.5 years; standard deviation,22.9 years; median age, 17 years; age range, 4 monthsto 75 years) recruited randomly at the StrabismusClinic at Okayama University. We excluded patientswith very large deviations (>45 PD) from this studybecause we sometimes did not obtain any result forthem with the video-based Hirschberg test. The rangeof horizontal deviation (as measured by the PACT orthe Krimsky test) was 4 to 45 PD of exotropia and 4to 40 PD of esotropia. Vertical ocular misalignmentalso was found in some patients; however, we couldnot identify the range because some patients had verti-cal deviation that could not be measured satisfactorilyby any technique. The noncycloplegic refraction,which was measured by using the video refractor de-scribed below, ranged from —4.25 to +3.50 dioptersspherical equivalents. No patient was cognitively ordevelopmentally disabled. Informed consent was pro-vided by either the patient or the patient's legal guard-

ian after an explanation of the study purpose andprocedures. This research followed the tenets of theDeclaration of Helsinki.

Test Apparatus

For acquisition and processing of the video image, avideo refractor PR1000 (Topcon, Tokyo, Japan) wasused. Originally, it was designed for refractive screen-ing of infants, and it has the capacity to measure therefractive error of both eyes simultaneously at a dis-tance of 120 cm.19"21 During the refraction computa-tion, it also provides a measure of the decentration ofthe corneal light reflex relative to the entrance pupilas a byproduct.20'21 We thus could apply this apparatusdirectly to Hirschberg measurement without modifi-cation.

A CCD video camera and a coaxial infrared light-emitting diode (wavelength, 800 fim) with which thePR1000 is equipped were used to take the subject'sface image. Its optics show a narrow depth of focus(approximately ±1 cm) and, therefore, testing dis-tance was assured. A scale reference was not necessary.Because this apparatus was placed on an altazimuthmount that could be slid back and forth, the examinercould quickly adjust its alignment and focus. The ac-quisition time per frame was 0.033 seconds, and foursuccessive frames were acquired on the frame grab-bing hardware within 0.15 seconds for a single deter-mination. A 166 X 300 pixel image of each eye wasthen extracted from the video signal and processed.The pixel resolution of the image was 7.7 pixels/mmhorizontally and 9.1 pixels/mm vertically. The coordi-nates of the pupillary center were determined by com-puting the geometric centroid of a binary image ofthe pupil. The coordinates of the corneal light reflexwere determined by computing a centroid from thegray scale image of the eye. The distance betweenthese two centroids, which was calculated from foursuccessive frames on average, was displayed in unitsof 0.1 mm in both the horizontal and vertical direc-tions (Fig. 1). The calculation time was approximately8 seconds.

Test Procedure

The subject either stood or sat 120 cm away from thePR1000. To estimate the strabismic deviation in theprimary position, we used the fixation target of thePR1000 (a flashing red light-emitting diode with asmall cross as an accommodative stimulus). The dis-tance between this built-in target and the subject's eyewas 140 cm. While the subject was gazing at this target,the decentration of the corneal light reflex relative tothe pupil center was measured twice by the PR1000,with an interval of at least 10 minutes. The strabismicdeviation was obtained based on the corneal reflexasymmetry by subtracting the decentration of the right


2680 Investigative Ophthalmology & Visual Science, December 1995, Vol. 36, No. 13

FIGURE I. A video image of the bright pupils and the cornealreflections acquired by a video refractor. The lowest lineshows the decentration of the corneal light reflex. On theleft, L0.2 10.2 indicates that the pupil for die subject's righteye is located 0.2 mm lateral and 0.2 mm inferior to thecorneal reflex. Similarly, MO.5 10.2, on the right, indicatesthat the pupil for the left eye is located 0.5 mm medial and0.2 mm inferior to the corneal reflex.

corneal reflex from the left. Testing was binocular inall instances. Spectacle correction was not worn be-cause both reflections from the lens surface and dis-tortion through the lens can interfere with this mea-surement. We did not use any instrument to fix thesubject's head. Parents held their infants' heads toface straight ahead. We sometimes used sound, a syn-chronized repeated bleep, to attract the subject's at-tention.

After the Hirschberg measurement, we performedthe PACT using the same fixation target. Briefly, thecover was placed alternately over each eye, while theeye movement of redress was compensated for by us-ing prisms. In 58 patients who participated in thisprocedure, the agreement between the video-basedHirschberg test and the PACT value was examined.

The Hirschberg ratio was assessed in 72 patientswho allowed monocular occlusion with an eye patch.A horizontal tangent scale was placed at a distance of110 cm from the subject, concentric with and perpen-dicular to the optical axis of the PR1000 (Fig. 2).Thirty small fixation targets, 15 on either side, werealigned horizontally on the scale at regular 5-cm inter-vals. Preliminarily, we verified the linearity of the dis-placement (mm) of the corneal light reflex relativeto eye rotation (degrees) in three selected subjectswho had neiuSer refractive error nor fixation abnor-mality. Although the subject looked at each fixationtarget in turn, the horizontal location of the corneallight reflex was stored by the PR1000. To investigatethe frequency distribution of the Hirschberg ratio, weselected two fixation targets placed at 55 cm on either

side of the center. The width between the two targets,which was verified with a laser beam pointer mountedon a rotational positioning device, was equivalent to51.63° visual angle from the subject's eye position. Weused this value as the common denominator of dieHirschberg ratio. While the subject was changing fixa-tion between these two targets monocularly, the dis-placement of the corneal light reflex was measuredby the PR1000. We used this value as the numeratorof the individually estimated Hirschberg ratio.

Statistical Analysis

The obtained data were examined according to Blandand Altman's method22 to determine the reproducibil-ity of the video-based Hirschberg test. We took re-peated measurements on a series of the patients andthen calculated, for each subject, the numeric differ-ence between the values obtained on two differentoccasions. These data were plotted on a differenceversus mean plot2'2'2"'4 (we do not know the absolutevalue, and the mean of the two measurements is thebest estimate we have). The bias between the first andsecond measurement was assessed statistically as themean of the differences compared to zero (paired t-test). The test-retest reproducibility was evaluated bydetermining the interval within which 95% of thesedifferences lie, which was constructed as the mean ofdie differences ± 1.96 X SD of the differences. Fordetermination of the magnitude of the difference be-tween the video-based Hirschberg test and the PACT,the degree of measurement agreement was examinedin the same way.

[ • • • • • • • • • • * • • » *

PR1000

5cm

140cm 120cm 110cm

FIGURE 2. The setup (top) for the assessment of the Hirsch-berg ratio.



&

- 3 - 2 - 1 0 1 2 3Mean of two horizontal measurements (mm)

- 3 - 2 - 1 0 1 2 3B Mean of two vertical measurements (mm)

FIGURE 3. A difference versus mean plot23'24 for repeatedmeasures of the video-based Hirschberg test for the hori-zontal (A) and the vertical (B) measurement. The hori-zontal line indicates the mean difference between the re-sults, and the shaded area indicates the 95% limits ofagreement (= mean difference ± 1.96 X SD of the differ-ences). Negative values indicate exotropic deviations, andpositive values represent esotropic deviations. The 95% lim-its of agreement were ±0.18 mm for the horizontal deviationand ±0.28 mm for the vertical deviation.

RESULTS

We succeeded in measuring the strabismic deviationin all 87 subjects with the video-based Hirschberg test,although the measurements were incomplete in sixsubjects (7%). The values obtained on two differentoccasions were highly correlated for both the hori-zontal measurement (r= 0.997) and the vertical mea-surement (r = 0.936). Figure 3A shows difference ver-sus mean plots for the horizontal measurement. Norelationship was observed between the difference and

the mean, and the difference was 0.00 ± 0.09 mm(mean ± SD). The 95% limits of agreement for thereproducibility of horizontal measurement thus were±0.18 mm. The difference versus mean plots for thevertical measurement are shown in Figure 3B. Again,no relation was observed between the difference andthe mean, and the difference was 0.00 ± 0.14 mm(mean ± SD). The 95% limits of agreement for thereproducibility of vertical measurement were ±0.28mm, and they were wider than those for the reproduc-ibility of horizontal value (P < 0.001, F-test for homo-geneity of variance). The difference between the twomeasurements according to age is shown in Figure 4.

<D <DS E

.Q co

CD

100-0.5

1 10Age (years)

B1 10Age (years)

100

FIGURE 4. The differences between two measures obtainedon different occasions were plotted as a function of thepatient's age (in logarithmic scale). There is no significantrelationship to age for horizontal measurement (A), al-though outlying values were sometimes found in patientsyounger than 2 years of age for vertical measurement (B).



The subject's age did not influence the reproducibil-ity, at least for horizontal measurement

A conventional plot and a difference versus meanplot comparing horizontal measurements obtained bythe two different techniques of the video-basedHirschberg test and the PACT are presented in Figure5. The values obtained by the two measurement meth-ods were highly correlated (r = 0.956), and the differ-ence was 0.78° ± 3.58° (mean ± SD). The mean differ-ence compared to zero (i.e., the bias) was not signifi-cant. The 95% limits of agreement between the twomeasuring methods were ±7.8°, or ±13.7 PD. In theseplots, we found that the data points for five subjectswere located far away from the others (outliers):Three of the subjects had intermittent exotropia, andtwo had accommodative esotropia.

A linear relationship between horizontal eye rota-tion and displacement of the corneal light reflex upto ±30° (±58 PD) was established in all three subjects.The correlation coefficient was high (r = 0.998 to0.999). The slope of the regression lines ranged from0.073 to 0.077 mm/degree, and the root mean squareresidual ranged from 0.046 to 0.073 mm. The displace-ment of the corneal light reflex during alternate gaz-ing at the ±55 cm lateral targets was 3.67 ± 0.35 mm(mean ± SD).

As shown in Figure 6, the Hirschberg ratio wasdistributed normally (Gaussian). The mean (± SD)Hirschberg ratio was 12.3° ± 1.27mm, or 21.8 ± 2.1PD/ram (n = 143). The maximum value was 15.6°/mm, and the minimum was 9.9°/mm.

DISCUSSION

The video-based Hirschberg measurement offersdefinite advantages in testing infants and young chil-dren who are unable to cooperate. First, the videoimage required for the measurement is captured inonly 0.033 seconds and measures the horizontal andvertical ocular deviation simultaneously. Accordingly,to complete this test, it is necessary to get the subject'sattention for only a brief moment. Next, this systemevaluates the relative distance between the corneallight reflex and the pupillary center. Because this dis-tance is little influenced by the subject's head move-ment as long as the gaze is not altered,17 rigid headfixation of the subject with a chin rest or a head bandis not necessary. These features are thought to contrib-ute to the high success rate of this measurement. Next,it is not guaranteed diat die subject will look at thedesired target, although proper fixation is crucial inany method of quantifying the strabismic deviation.As Barry and colleagues15 mentioned, the fixation canbe checked only by repeated measurements and byverifying the consistency of the results. The video-based Hirschberg test provides immediate results and

CD

CD

o

-30 - 2 0 - 1 0 0 10 20 30PACT (degrees)

B

-30 -20 -10 0 10 20 30Mean Hirschberg test-PACT (degrees)

FIGURE 5. A conventional plot of video-based Hirschberg andthe prism and alternate cover test values (A) and a differ-ence versus mean plot23'24 for the same values (B). The hori-zontal line in (B) indicates the mean difference, that is, biasbetween the two methods, and the shaded area the 95%limits of agreement. The values obtained by the two differentmethods were correlated (r = 0.956). However, the 95%limits of agreement between the two methods were ±7.8°or ±13.7 PD, and oudying values were found in some pa-tients with intermittent exotropia or accommodative esotro-pia. • = intermittent XT; O = microtropia; V = infantileET; • = acquired ET; • = constant XT; -I- = accommoda-tive ET; X = others.

can be repeated. This allows us to confirm the fixationquality when proper fixation is doubtful. Finally, as-sessment of deviations at distance fixation is a disad-vantage of the conventional Hirschberg test and ofthe Krimsky test as well.1'212 For example, intermittent


Video-enhanced Hirschberg Test

30-,

25-

20-

3oO

15-

10-

5

9 10 11 12 13 14 15 16Hirschberg ratio (degrees/mm)

FIGURE 6. Frequency distribution of the Hirschberg ratioThe mean (±SD) ratio is 12.3° ± 1.2°/mm, demonstratinga large intersubject variation in the Hirschberg ratio.

exotropia is likely to be overlooked with the Hirsch-berg test because the ocular misalignment is easilycompensated for by the motor fusion at near fixation.In the video-based Hirschberg measurement, 27 of the32 patients (84%) with intermittent exotropia weredetermined to have significant ocular deviation. Thisresult suggests that the 27 patients were exotropic un-der this testing condition. This fairly good success ratein the detection of intermittent exotropia probablyresulted from the long testing distance, which was pro-vided by the remote optical system of the PR1000.

We found that the results obtained with the video-based Hirschberg test were highly reproducible. The95% limits of agreement were ±0.18 mm for the hori-zontal measurement and ±0.28 mm for the verticalmeasurement. These values correspond to ±2.2°(±3.8 PD) and ±3.4° (±5.9 PD), respectively, whenconverted to eye rotation with the average Hirschbergratio we found in this study. Any manifest strabismusor time-series fluctuation of a strabismic deviation thatis larger than these threshold values can be detectedin principle; we could identify 7 of the 13 patients(54%) with microtropia with the video-based Hirsch-berg test alone.

The resolution of the CCD image sensor is gener-ally poor compared with that of the photographic film.However, the coordinates of the corneal light reflexand the pupil can be determined with subpixel resolu-tion by mathematical computation of the center ofgravity (i.e., centroid) of their images.10 Moreover, theretroilluminated pupil, which was produced by arrang-ing optically coincident lighting and collection aper-tures,17 was well defined regardless of the iris color.We think that the centroid coordinates of the pupilare a more reliable standard than the scleral-corneallimbus, which was often used in the photographicmethod1"' but which is actually an indistinct bound-ary. Nevertheless, the measurement range with this

2683

technique is at present limited. When the eye is ro-tated over 25° or 45 PD, the intensity of the retroillu-mination of the entrance pupil is gradually decreased,and bright reflections are present on the conjunctiva.These phenomena sometimes disturb the image pro-cessing in the PR1000.

Cover tests generally are regarded as the best wayto diagnose strabismus accurately, although they dorequire that the patient fixate a target continuously.In these tests, the lower limit of observable change ofthe eye position with the naked eye has been reportedto be 2 PD or more under ideal conditions.2'1"' Yet,the limit can deteriorate even further when there iserror in prism positioning, misinterpretation of am-biguous fixation eye movements, or both. Consideringthat, we conclude that the precision of the video-basedHirschberg test is comparable to that of the PACTunder clinical conditions.

The reproducibility of the video-based Hirschbergmeasurement was excellent regardless of the subject'sage group, and we think we can apply this test forinfants and young children. Using this test, we candetermine whether the deviation is manifest. When amanifest strabismus is present, we can still examinethe binocular fixation pattern. These clinical assess-ments are important in evaluating and following pa-tients with strabismic amblyopia until they are oldenough to take the Snellen visual acuity measurement.

The difference of the reproducibility between thehorizontal and vertical measurements was probablyrelated to the differing resolution of the CCD imagesensor for these two planes. In addition, an obstruc-tion of the entrance pupil by the eyelashes or theeyelid may be a contributing factor, especially in in-fants and young children. The influence of this factormay be reflected by the oudying values observed inthe reproducibility assessment for the vertical mea-surement. Hence, attention should be paid to suchobstructions just before measurement.

The 95% limits of agreement between theHirschberg and PACT values were poor. In the differ-ence and mean plots of this comparative analysis, wefound outlying values for some patients with intermit-tent exotropia and with accommodative esotropia. Inthe former group, the discrepancy between the twomethods was perhaps caused by the intermittent na-ture of the strabismus; the Hirschberg test examinesmanifest deviation and the PACT manifest plus latentdeviation with covering one eye covered. In the lattergroup, the deviation in those patients with hyperopiawould be expected to vary significantly from one test-ing time to the next because spectacle correctioncould not be worn.

We found that the mean (± SD) Hirschberg ratiowas 12.3° ± 1.2°/mm. The average value is inagreement with that shown previously on the basis



of laboratory research or theoretical extrapola- Acknowledgments3-fi,9.10.18tion;1""*111'" In addition, Riddell and colleagues9 re-

ported that the standard deviation of the Hirschbergratio, which was obtained with the photographicmethod in 333 normal subjects, was 1.8°/mm. Consid-ered together, their result and ours suggest the exis-tence of considerable intersubject variation in theHirschberg ratio: The standard deviation was approxi-mately 10% of the mean value. If the individualHirschberg ratio is not determined, the potential foras much as 20% systematic measurement error (95%confidence limits) should be taken into consideration.In the patients with large-angle strabismus, the in-tersubject variance in Hirschberg ratio can be anothernotable reason that the Hirschberg test and PACTvalues conflict. When highly quantitative data are re-quired, such as in the preoperative planning for stra-bismic surgery, an individually estimated "personal"Hirschberg ratio should be used for the averagedHirschberg ratio.

The angle between the pupillary axis and the lineof sight is known as the angle lambda (\). The anglesX. differ among subjects and, hence, can bias the re-sults in the Hirschberg test. Although comparison ofbinocular and monocular fixating data is essential fordiscriminating between true strabismus and the angle\, there is a trade-off between this theoretical rigidityand the ease of clinical application. However, theangles \ have been reported to be similar for an indi-vidual, except in instances in which there are grossanatomic differences, such as marked anisometropiaor unilateral retinopathy of prematurity.51026 Accord-ingly, we regarded the corneal reflex asymmetry be-tween the two eyes as the strabismic deviation in thisstudy.

Finally, the apparatus described herein originallywas not designed for examination of strabismus, andit does not necessarily have mechanical specificationssuitable for this purpose. Accordingly, some improve-ments in hardware and software design should bemade to adapt it for use in various clinical applica-tions, such as optical correction and very large-anglestrabismus. The systematic measurement error causedby the intersubject variance in Hirschberg ratio is notinsignificant in terms of the quantitative analysis. How-ever, the current study verified that the video-en-hanced Hirschberg test affords good reproducibilityand ease of application under actual clinical condi-tions. As long as it is used for specialized purposes withconsideration of its advantages and disadvantages, thistest will continue to serve as an invaluable tool.

Key Words

Hirschberg test, image processing, infant, ocular alignment,strabismus

The authors thank Dr. Nobuhiko Matsuo, Professor andChairman, Department of Ophthalmology, Okayama Uni-versity Medical School, for his help. They also thank Mrs.Mie Tohi for her expert technical assistance.

References

1. Krimsky E. The Corneal Light Reflex. Springfield:Charles C Thomas; 1972:25-27.

2. von Noorden GK. Binocular Vision and Ocular Motility.4th ed. St. Louis: CV Mosby; 1990:178-179.

3. Carter AJ, Roth N. Axial length and the Hirschbergtest. Am] Optom Physiol Optics. 1978;55:361-364.

4. Eskridge JB. The Hirschberg test: A double-maskedclinical evaluation. Am J of Optom Physiol Optics.1988; 65:745-750.

5. Brodie SE. Photographic calibration of the Hirsch-berg test. Invest Ophthalmol Vis Sri. 1987;28:736-742.

6. DeRespinis PA, Naidu E, Brodie SE. Calibration ofHirschberg test photographs under clinical condi-tions. Ophthalmology. 1989; 96:944-949.

7. Quick MW, Boothe RG. Measurement of binocularalignment in normal monkeys and monkeys with stra-bismus. Invest Ophthalmol Vis Sci. 1989; 30:1159-1168.

8. Quick MW, Boothe RG. A photographic techniquefor measuring horizontal and vertical eye alignmentthroughout the field of gaze. Invest Ophthalmol Vis Sci.1992; 33:234-246.

9. Riddell PM, Hainline L, Abramov I. Calibration of theHirschberg test in human infants. Invest OphthalmolVis Sri. 1994; 35:538-543.

10. Miller JM, Mellinger M, Greivenkemp J, Simons K.Videographic Hirschberg measurement of simulatedstrabismic deviations. Invest Ophthalmol Vis Sri.1993; 34:3220-3229.

11. Brodie SE. Corneal topography and the Hirschbergtest. Appl Opt. 1992;31:3627-3631.

12. Putnam OA, Quereau JV. Precisional errors in mea-surement of squint and phoria. Arch Ophthalmol.1945;34:7-15.

13. Helveston EM, Ellis FD, Plager DA, Miller KK. Earlysurgery for essential infantile esotropia. J Pediatr Oph-thalmol Strabismus. 1990; 27:115 -129.

14. von Noorden GK. A reassessment of infantile esotro-pia: XLIV Edward Jackson Memorial Lecture. Am JOphthalmol. 1988; 105:1-10.

15. Barry JC, Effert R, Reim M, Meyer-Ebrecht D. Com-putational principles in Purkinje I and IV reflectionpattern evaluation for the assessment of ocular align-ment. Invest Ophthalmol Vis Sri. 1994;35:4205-4218.

16. Barry JC, Effert R, Kaupp A, Burhoff A. Measurementof ocular alignment with photographic Purkinje I andIV reflection pattern evaluation. Invest Ophthalmol VisSri. 1994; 35:4219-4235.

17. Merchant J, Morrissette R, PorterfieldJ. Remote mea-surement of eye direction allowing subject motionover one cubic foot of space. TF.F.F. Trans Biomed Eng.1974;21:3O9-317.

18. Uozato H, Loupe D, Guyton DL. Theoretical analysis



of the Hirschberg test. Jpn Rev Clin Ophthalmol.1987;81:1204-1207.

19. Uozato H, Hirai H, Saishin M, Fukuma Y. Screeningof refraction in infants with a new infrared video-re-fraction technique. / Rev Clin Ophthalmol. 1990;84:627-631.

20. Yamada M, Ohtake Y, Hiida Y, Tanaka Y. Screeningof refraction and eye position in children: Compari-son of photorefractor PR-1000 and photorefractioncamera. Jpn Rev Clin Ophthalmol. 1992;86:1468-1472.

21. KatoriJ, Komachi Y, Ukai K, Ishikawa S. Measurementof near triads using video refraction method. Jpn RevClin Ophthalmol. 1992; 86:1473-1477.

22. Bland JM, Altman DG. Statistical methods for as-

sessing agreement between two methods of clinicalmeasurement. Lancet. 1986; 1-8476:307-310.

23. Zadnik K, Mutti DO, Adams AJ. The repeatability ofmeasurement of the ocular components. Invest Oph-thalmol Vis Sci. 1992;33:2325-2333.

24. Ludvigh E. Amount of eye movement objectively per-ceptible to the unaided eye. Am J Ophthalmol.1949; 32:649-650.

25. Romano PE, von Noorden GK. Limitations of covertest in detecting strabismus. Am J Ophthalmol.1971;72:10-12.

26. Ferreri G, Giorgianni A, Siracusano B, Cusmano R.The angle kappa in ametropia. New Trends Ophthalmol.1988;3:27-33.


the reliability of a video-enhanced hirschberg test under clinical

Documents