[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address 54.161.130.145. Please contact the publisher to request reinstatement.
[Skip to Content Landing]
Download PDF
Figure 1.
Digitized image of an eye with a superimposed outline of the area of herpes simplex virus stromal keratitis and its geometric center (x). Geometric overlay was used to orient the location on the cornea.

Digitized image of an eye with a superimposed outline of the area of herpes simplex virus stromal keratitis and its geometric center (x). Geometric overlay was used to orient the location on the cornea.

Figure 2.
Calibration of photographic slitlamp cameras at 9 clinical centers. Magnification for each center (black bars) is the average increase in the linear dimension on reference photographs taken at the instrument's 6× setting (solid horizontal line). Pixel intensity at each center (gray bars) is the average 8-bit linear gray value when imaging the 18% gray card (dashed horizontal line). Error bars indicate SD.

Calibration of photographic slitlamp cameras at 9 clinical centers. Magnification for each center (black bars) is the average increase in the linear dimension on reference photographs taken at the instrument's 6× setting (solid horizontal line). Pixel intensity at each center (gray bars) is the average 8-bit linear gray value when imaging the 18% gray card (dashed horizontal line). Error bars indicate SD.

Figure 3.
Comparison of the elliptical area of herpes simplex virus stromal keratitis estimated by clinical measurements and the area obtained by image analysis. Four atypical cases (x) were not used to estimate the linear regression line and 95% confidence band.

Comparison of the elliptical area of herpes simplex virus stromal keratitis estimated by clinical measurements and the area obtained by image analysis. Four atypical cases (x) were not used to estimate the linear regression line and 95% confidence band.

Figure 4.
Distribution of the geometric shape of herpes simplex virus stromal keratitis. The shape factor measures relative roundness, from a line approaching 0 to a perfect circle at 1.

Distribution of the geometric shape of herpes simplex virus stromal keratitis. The shape factor measures relative roundness, from a line approaching 0 to a perfect circle at 1.

Figure 5.
Plot of geometric centers of stromal keratitis onto a corneal template. The nasal direction is mapped at 0° for right and left eyes.

Plot of geometric centers of stromal keratitis onto a corneal template. The nasal direction is mapped at 0° for right and left eyes.

Table. 
Relative Effects of Baseline Factors on Larger or Denser Stromal Keratitis Examined by Image Analysis
Relative Effects of Baseline Factors on Larger or Denser Stromal Keratitis Examined by Image Analysis
1.
Dawson  CRJones  DBKaufman  HEBarron  BAHauck  WWWilhelmus  KR Design and organization of the Herpetic Eye Disease Study (HEDS). Curr Eye Res 1991;10 ((suppl)) 105- 110
PubMedArticle
2.
Sudesh  SLaibson  PR The impact of the Herpetic Eye Disease Studies on the management of herpes simplex virus ocular infections. Curr Opin Ophthalmol 1999;10 (4) 230- 233
PubMedArticle
3.
Wilhelmus  KRGee  LHauck  WW  et al.  A controlled trial of topical corticosteroids for herpes simplex stromal keratitis. Ophthalmology 1994;101 (12) 1883- 1895
PubMedArticle
4.
Barron  BAGee  LHauck  WW  et al.  A controlled trial of oral acyclovir for herpes simplex stromal keratitis. Ophthalmology 1994;101 (12) 1871- 1882
PubMedArticle
5.
Wilhelmus  KR Diagnosis and management of herpes simplex stromal keratitis. Cornea 1987;6 (4) 286- 291
PubMedArticle
6.
Williams  HPFalcon  MGJones  BR Corticosteroids in the management of herpetic eye disease. Trans Ophthalmol Soc U K 1977;97 (2) 341- 344
PubMed
7.
Merin  LMWilhelmus  KRHerpetic Eye Disease Study Group, Anterior segment photography in the Herpetic Eye Disease Study. J Ophthalmic Photogr 1992;14 (1) 34- 38
8.
Herpetic Eye Disease Study Group, Manual of Operations.  Springfield, VA National Technical Information Service1990;NTIS publication PB93175453
9.
Mukerji  NVajpayee  RBSharma  N Technique of area measurement of epithelial defects. Cornea 2003;22 (6) 549- 551
PubMedArticle
10.
Committee on Vision, Assembly of Behavioral and Social Sciences, National Research Council, National Academy of Sciences, Recommended standard procedures for the clinical measurement and specification of visual acuity: report of working group 39. Adv Ophthalmol 1980;41103- 148
PubMed
11.
Wilhelmus  KRGay  CAMerin  LM  et al.  Image analysis of herpes simplex stromal keratitis.  Paper presented at: International Conference on Herpetic Eye Diseases July 26, 1990 Oxford, England
12.
Wolfe  JKChylack  LT  Jr Objective measurement of cortical and subcapsular opacification in retroillumination photographs. Ophthalmic Res 1990;22 ((suppl 1)) 62- 67
PubMedArticle
13.
Klein  BEKKlein  RLinton  KLPMagli  YLNeider  MW Assessment of cataracts from photographs in the Beaver Dam Eye Study. Ophthalmology 1990;97 (11) 1428- 1433
PubMedArticle
14.
Bender  LSpalton  DJUyanonvara  B  et al.  POCOman: new system for quantifying posterior capsule opacification. J Cataract Refract Surg 2004;30 (10) 2058- 2063
PubMedArticle
15.
Age-Related Eye Disease Study Research Group, The Age-Related Eye Disease Study (AREDS) system for classifying cataracts from photographs: AREDS report No. 4. Am J Ophthalmol 2001;131 (2) 167- 175
PubMedArticle
16.
Constable  IJ Foreword. Yogesan  KKumar  SGoldschmidt  LCuadros  JTeleophthalmology. Berlin, Germany Springer-Verlag2006;vii
17.
Solomon  AWBowman  RJCYorston  D  et al.  Operational evaluation of the use of photographs for grading active trachoma. Am J Trop Med Hyg 2006;74 (3) 505- 508
PubMed
18.
Gilchrist  J Computer processing of ocular photographs: a review. Ophthalmic Physiol Opt 1987;7 (4) 379- 386
PubMed
19.
Bartlett  HEperjesi  F Use of fundus imaging in quantification of age-related macular change. Surv Ophthalmol 2007;52 (6) 655- 671
PubMedArticle
20.
Patton  NAslam  TMMacGillivray  T  et al.  Retinal image analysis: concepts, applications and potential. Prog Retin Eye Res 2006;25 (1) 99- 127
PubMedArticle
21.
Niemeijer  Mvan Ginneken  BRussell  SRSuttorp-Schulten  MSAAbràmoff  MD Automated detection and differentiation of drusen, exudates, and cotton-wool spots in digital color fundus photographs for diabetic retinopathy diagnosis. Invest Ophthalmol Vis Sci 2007;48 (5) 2260- 2267
PubMedArticle
22.
Fieguth  PSimpson  T Automated measurement of bulbar redness. Invest Ophthalmol Vis Sci 2002;43 (2) 340- 347
PubMed
23.
Zegans  METoutain  CSaine  P  et al.  Analysis of digital images of corneal ulcers in the pilot study for the Steroid Corneal Ulcer Trial (SCUT).  Poster presented at: Annual Meeting of the Association for Research in Vision and Ophthalmology May 1, 2006 Fort Lauderdale, FL
24.
VanRoekel  RCBower  KSBurka  JMHoward  RS Anterior segment measurements using digital photography: a simple technique. Optom Vis Sci 2006;83 (6) 391- 395
PubMedArticle
25.
Chan  JWEdwards  MHWoo  GCWoo  VC Objective method to measure corneal clarity before and after laser in situ keratomileusis. J Cataract Refract Surg 2003;29 (1) 118- 124
PubMedArticle
26.
Barr  JTWilson  BSGordon  MO  et al. CLEK Study Group, Estimation of the incidence and factors predictive of corneal scarring in the Collaborative Longitudinal Evaluation of Keratoconus (CLEK) Study. Cornea 2006;25 (1) 16- 25
PubMedArticle
27.
Shin  DSKaiser  RSLee  MSBerger  JW Fundus image change analysis: geometric and radiometric normalization. Proc SPIE 1999;3591129- 136
28.
Bailey  ILBullimore  MARaasch  TWTaylor  HR Clinical grading and the effects of scaling. Invest Ophthalmol Vis Sci 1991;32 (2) 422- 432
PubMed
29.
Chong  ESimpson  TFonn  D The repeatability of discrete and continuous anterior segment grading scales. Optom Vis Sci 2000;77 (5) 244- 251
PubMedArticle
30.
Chylack  LT  JrRosner  BWhite  OTung  WHSher  LD Standardization and analysis of digitized photographic data in the longitudinal documentation of cataractous growth. Curr Eye Res 1988;7 (3) 223- 235
PubMedArticle
31.
Musadiq  MPatsoura  EHughes  SYang  YC Measurements of linear dimensions on fundus photographs: comparison between photographic film and digital systems. Eye 2003;17 (5) 619- 622
PubMedArticle
32.
Wolffsohn  JS Incremental nature of anterior eye grading scales determined by objective image analysis. Br J Ophthalmol 2004;88 (11) 1434- 1438
PubMedArticle
Clinical Sciences
February 9, 2009

Slitlamp Biomicroscopy and Photographic Image Analysis of Herpes Simplex Virus Stromal Keratitis

Author Affiliations

Author Affiliations: Cullen Eye Institute, Baylor College of Medicine, Houston, Texas (Drs Wilhelmus, Mitchell, and Jones); Francis I. Proctor Foundation, University of California, San Francisco (Dr Dawson); LSU Eye Center, Louisiana State University Health Sciences Center, New Orleans (Drs Barron and Kaufman); Illinois Eye and Ear Infirmary, University of Illinois at Chicago (Dr Sugar); Eye Institute, Medical College of Wisconsin, Milwaukee (Dr Hyndiuk); Wills Eye Institute, Jefferson Medical College, Thomas Jefferson University, Philadelphia, Pennsylvania (Dr Laibson); Emory Eye Center, Emory University, Atlanta, Georgia (Dr Stulting); and Mount Sinai Medical Center, New York, New York (Dr Asbell). Drs Barron and Hyndiuk are now retired from academic practice.

Arch Ophthalmol. 2009;127(2):161-166. doi:10.1001/archophthalmol.2008.577
Abstract

Objective  To validate photographic bioimaging for evaluating the severity of herpes simplex virus keratitis.

Methods  Stromal keratitis of patients in the Herpetic Eye Disease Study was clinically measured with a slitbeam micrometer and then photographed at trial entry. Calibrated images of 169 eyes were analyzed for the size, location, and density of stromal keratitis and endotheliitis, with shape factor as a function of area and perimeter. Validity was assessed by comparing clinical and computerized measurements and by correlating the keratitis area with visual acuity. Logistic regression explored characteristics associated with larger or denser corneal inflammation.

Results  Stromal keratitis had a median area of 22.4 mm2 (interquartile range, 12.8-31.6 mm2) with a median shape factor of 0.69 (interquartile range, 0.56-0.79); 126 eyes (75%) had their midpoint within 2 mm of the cornea's geometric center. Photoanalytical area estimates of herpetic stromal keratitis correlated closely with clinical measurements (correlation coefficient, 0.83). Eyes with larger stromal keratitis had worse vision (correlation coefficient, 0.32) and were more likely to have iritis (P = .01). Necrotizing stromal keratitis was significantly whiter (P = .02).

Conclusions  Image analysis validly assesses the disciform geometry of herpetic stromal keratitis and confirms that increased severity is associated with uveitis and reduced vision.

The Herpetic Eye Disease Study is a set of multicenter, randomized, controlled clinical trials on the treatment and prevention of ocular infection and inflammation due to herpes simplex virus.1,2 In 2 trials,3,4 the effects of corticosteroid and antiviral agents were studied in patients with herpetic stromal keratitis and edema. Besides categorizing inflammatory severity,5,6 clinical examiners systematically measured and categorized the degree of stromal keratitis and obtained anterior segment photographs.7 We used these biomicroscopic measurements and calibrated photographs to examine the validity of digital image analysis in assessing corneal inflammation during herpetic eye disease. We also applied image morphometry to describe the geometric pattern of stromal keratitis and explored clinical features associated with severity.

METHODS
PARTICIPANTS

Patients with herpes simplex virus stromal keratitis or endotheliitis who met eligibility criteria and gave informed consent were enrolled in 1 of 2 multicenter clinical trials at 9 clinical centers where the study had been approved by institutional review boards. One hundred six patients entered into a placebo-controlled trial of a topical corticosteroid (49 were assigned to placebo and 57 to prednisolone phosphate),3 and 104 corticosteroid-treated patients entered into a placebo-controlled trial of an oral antiviral agent (53 were assigned to placebo and 51 to acyclovir).4 All of the patients received trifluridine 1% solution.

CLINICAL ASSESSMENT

Certified investigators performed corneal examinations at each study visit using a slitlamp biomicroscope (Haag-Streit BM 900; Haag-Streit USA, Inc, Mason, Ohio) fit with 10× eyepieces.8 With the transformer switched to 6 V, biomicroscopic examination began with the instrument's illumination arm set upright with a fully opened filter and reticule. A fixation light in front of the contralateral eye helped to position the study eye so as to align the center of the region of stromal keratitis with the midpoint of the slitbeam. A sketch was made by outlining areas of stromal keratitis onto a printed diagram and adding reference lines to indicate length as the greatest diameter and width as its broadest perpendicular.

Beginning with a slitbeam positioned at 0° on the copivotal axis, the illumination housing was rotated to align a 10-unit-wide beam along the greatest length of stromal inflammation, and a continuous micrometer measured the longest diameter of stromal keratitis to the nearest 0.1 mm. For an axis longer than 8.0 mm, the eye was repositioned and the beam was adjusted to connect the center and border of stromal keratitis so that doubling this reading gave the lengthwise measurement. The illumination housing was then rotated 90° to measure the orthogonal dimension along the widest span of stromal keratitis. Length and width measurements were taken in triplicate and averaged for each region of nonnecrotizing or necrotizing stromal keratitis. Because a rectangular area overestimates an oval's size,9 the total area of stromal keratitis used in the trial protocol was transformed to an elliptical area by applying a factor of π/4. The density of the inflammatory zone was categorized by aligning the illuminating slitbeam onto the center of the area of greatest density to distinguish whether the visibility of iris details was slightly blurred or was partially or completely obscured.

Best-corrected visual acuity was measured by a certified examiner who used subjective refraction to determine the optimal lens correction.8 Visual acuity of each eye was measured at 4 m with modified Bailey-Lovie charts retroilluminated at 970 to 1075 lux10 and recorded as the smallest line for which at least 3 letters were correctly read. If visual acuity was 20/100 or worse, testing was repeated at 1 m after adding +0.75 sphere to the refraction. Snellen visual acuity was transformed to logMAR for analysis. For patients unable to discern letters at 1 m, the ability to count fingers, to detect hand motion, or to perceive light was evaluated and recorded as the logMAR equivalent of 2.0, 2.3, or 2.7, respectively.

PHOTODOCUMENTATION

Anterior segment photomicrography was performed with optical equipment (Zeiss Photo Slit Lamp; Carl Zeiss, Inc, New York, New York) consisting of a 35-mm single-lens reflex camera, 70/30 beam splitter, camera adapter with aperture diaphragm, 2× magnifier, and eyepiece graticule. The photographic protocol specified exposure parameters, image composition, and magnification settings.7 After attending a training seminar, photographers submitted acceptable slitlamp images from 2 patients to become certified.8

Each session filmed a combined metric and contrast scale with a 1.5-mm black band, 2.2-mm white band, and matte gray background having 18% reflectance (Kodak, Rochester, New York). Ocular images were obtained at galilean settings of 6× and 10× to yield photographic magnifications of 1.4× and 2.2×, respectively, using dual diffuse illumination. Paired photographs were obtained with the eye gazing in primary position and then with the eye realigned to center the corneal inflammatory lesion within the image frame. Additional photographs were taken with a thin slitbeam, sclerotic scatter, and retroillumination.

Slide frames were labeled with the date, eye, protocol, and numeric code. Photographs were sent to a coordinating center that received slides at a median of 17 days after patient enrollment. Collated sets were subsequently forwarded to a reading center masked to treatment allocation where slides were checked for appropriate labeling, focus, composition, centration, and control of artifacts. If a center's photograph sets were judged to be not acceptable or evaluable, corrective actions including photographer retraining, temporary center suspension of trial photography, and photographer recertification or replacement were taken. Photography was optional during the final 6 months of trial enrollment.

IMAGE PROCESSING

Diapositives archived in dark storage were digitized with a fixed-film scanning system (Super Coolscan 5000 ED; Nikon, Tokyo, Japan) having a charge-coupled linear sensor with optical resolution of 4000 pixels per inch. Images were obtained at a scan range of 3946 × 5782 pixels using neutralized correction and enhancement settings, imported in TIFFs, and cached on a 1-TB server (PowerEdge 4600; Dell, Round Rock, Texas). Image analysis software (SigmaScan Pro 5.0; Systat, Richmond, California) converted uncompressed images to linear gray-scale equivalents at a bit depth of 256 shade levels per pixel. Endpoints of the luminance scale were set to 0 and 255, and 18% gray was mapped to an intensity value of 128.

A pilot study exploring the feasibility of image analysis of biomicroscopic photographs11 captured transparencies with a mounted Vidicon camera (Dage-MTI, Michigan City, Indiana) and used a computerized image processor (1024XM; MegaVision, Inc, Santa Barbara, California). The interinstrument reproducibility of the SigmaScan and MegaVision systems for estimating the total area of stromal keratitis was examined in a subset of 52 patients photographed at study entry, using the Lin concordance correlation coefficient (ρC) with a 95% confidence interval (CI). Because results correlated (ρC = 0.71; 95% CI, 0.57-0.84), the SigmaScan system was used for analysis.

MORPHOMETRY

The width in pixels per millimeter was evaluated in triplicate for the mensural bands imaged at each photographic session. The luminance intensities at the upper and lower sections of the 18% gray background were recorded in triplicate and averaged, and similar readings were obtained from the white strip. Digitized clinical images were processed by boundary enhancement to accentuate the border of stromal infiltration and edema. Crispening was implemented through comparison of adjacent pixels, with the use of sharpening filters to enhance relative differences in neighboring pixel intensity values. Guided interactively by photographs taken with sclerotic scatter illumination, adjacent pixels were joined to outline a single area and a contrast histogram of pixels within the perimeter was plotted. Planimetric measurements and the averaged intensity of stromal keratitis were normalized to corresponding calibration scales.

An image analyst (B.M.M.) masked to clinical data recorded the area, perimeter, and axis dimensions. The maximal Feret diameter was the longest distance between parallel tangents. A shape factor estimating the relative circularity of the border of stromal keratitis was calculated as the product of 4π and the area divided by the perimeter squared. An overlay was modified to plot corneal disease,1215 and this template was centered onto the midpoint of the corneal image by encircling the limbus (Figure 1). The geometric center of the area of stromal keratitis was mapped onto a composite chart by polar coordinates.

STATISTICAL ANALYSIS

Criterion validity was assessed by the Spearman rank correlation coefficient (ρS) with a 95% CI to compare the area of stromal keratitis estimated clinically with the area determined by computerized image analysis. Simple linear regression further compared the relationship between areas estimated by clinical and planimetric methods, and model robustness to observed data points was assessed with Bland-Altman plots and influence diagnostics using standardized delta-beta (dfbeta) analysis. Validity assessment compared the averaged intensity of stromal keratitis with the density as categorized clinically using the Kendall rank correlation coefficient (τ) and Wilcoxon rank sum test. Face validity was examined by comparing initial area and intensity estimates with best-corrected visual acuity through the use of ρS. Internal consistency was examined by ρC to compare images obtained at identical magnification settings but different gaze positions. Logistic regression was used to explore univariate associations of demographic and other baseline variables with categorized area and density estimates of stromal keratitis. Statistical analyses were performed with Intercooled Stata version 10 statistical software (Stata Corp, College Station, Texas). All P values were 2-tailed, and P < .05 indicated statistical significance.

RESULTS

Of 210 patients enrolled in 1 of 2 clinical trials, 175 underwent slitlamp photography and 169 had 6× photographs of sufficient quality for analysis.

IMAGE CALIBRATION

Compared with 220 pixels/mm for 6× magnification, 7 centers used photographic equipment that slightly exceeded expected calibration by an average of 6% to yield actual mean magnification ranging from 6.3× to 6.5× (Figure 1). One center was not significantly different from 6×, while another center's photographic slitlamp camera resulted in average image magnification of 5.4×. Center values of pixel intensity for imaging pure white at the 6× setting ranged from a mean (SD) of 207 (11) units to 231 (5) units. For imaging the 18% gray background, values for each clinical center ranged from a mean (SD) of 79 (16) units to 157 (24) units (Figure 2), including 3 centers that did not significantly vary from a projected value of 128 gray-value units.

VALIDITY AND CONSISTENCY

The corneal inflammatory area estimated by photographic image analysis significantly correlated (P < .001) with the elliptical area estimated by clinical measurements (ρS = 0.83; 95% CI, 0.78-0.88). However, 4 cases with clinically measured areas of 76, 96, 102, and 133 mm2 and corresponding planimetric areas of 12, 30, 48, and 30 mm2, respectively, exceeded 95% limits of agreement and showed unduly influential dfbetas. After excluding these outliers for this comparison only, the slope of the regression line intersecting the origin (Figure 3) changed from 0.83 (95% CI, 0.78-0.89) to 0.98 (95% CI, 0.94-1.02). The clinical categorization of the density of stromal keratitis correlated (τ = 0.17; 95% CI, 0.10-0.25) with the averaged gray-scale intensity found by image analysis. The median averaged intensity value of the zone of stromal keratitis was 81 units (interquartile range [IQR], 59-103 units). The median intensity value of 64 units (IQR, 54-92 units) among 58 eyes (34%) clinically categorized as having mild haze was significantly lower (P < .001) than the median value of 95 units (IQR, 63-116 units) among 111 eyes (66%) judged to have moderate to marked opacification. Visual acuity correlated with the area of stromal keratitis (ρS = 0.32; 95% CI, 0.17-0.45) and with its averaged gray-level intensity (ρS = 0.35; 95% CI, 0.21-0.47). Comparison of 14 paired images in which one photograph was aligned with the eye's visual axis and the other photograph had the area of stromal keratitis at its center showed that gaze position did not significantly alter estimates of area (ρC = 0.99; 95% CI, 0.98-1.00) or intensity (ρC = 0.93; 95% CI, 0.86-1.00).

GEOMETRIC MAPPING

The median areas of stromal keratitis estimated by clinicians and by image analysis were 19.8 mm2 (IQR, 10.2-29.2 mm2) and 22.4 mm2 (IQR, 12.8-31.6 mm2), respectively. The mean (SD) dimension of the longest axis of the primary lesion was 5.1 (2.4) mm when measured clinically and 6.8 (2.1) mm when measured photographically, and the corresponding mean (SD) lengths of the orthogonal axis were 5.0 (2.8) mm and 5.0 (1.9) mm. The mean (SD) Feret diameter by image analysis was 5.4 (1.8) mm. The median perimeter of stromal keratitis was 21.5 mm (IQR, 16.5-26.0 mm). Stromal keratitis varied from a fusiform oval to near-circularity (Figure 4). The median shape factor was 0.69 (IQR, 0.56-0.79), and 40 eyes (24%) had a shape factor greater than 0.80. The geometric center of the zone of stromal keratitis was located at a median of 1.3 mm (IQR, 0.8-2.0 mm) from the corneal midpoint, and 126 eyes (75%) had their midpoint within the central 4-mm zone (Figure 5). Of 169 keratitis lesions among the 4 corneal quadrants, 71 (42%) were centered temporally, 48 (28%) inferiorly, 33 (20%) nasally, and 17 (10%) superiorly.

ATTRIBUTES OF SEVERITY

Eyes with an area of stromal keratitis larger than 20 mm2 had significantly increased odds of iritis (Table) by biomicroscopy (P = .03) and photogrammetry (P = .01). Necrotizing keratitis was more often associated with moderate to complete obscuration of the iris details (Table) on clinical examination (P = .03) and significantly exceeded the median intensity value of 81 gray-value units on densitometry (P = .02). Best-corrected visual acuity worse than 20/40 was found for 91 photographed eyes (54%). The odds of reduced vision occurred 3.08 (95% CI, 1.62-5.84) times as often when the stromal keratitis area was larger than 20 mm2 and 3.17 (95% CI, 1.69-5.95) times as often when stromal keratitis intensity exceeded 81 gray-value units on digitized images. Eyes with stromal keratitis whose geometric center was located within 2 mm of the corneal center were 3.75 (95% CI, 1.78-7.89) times as likely to have visual acuity worse than 20/40 compared with eyes with more peripheral keratitis.

COMMENT

Ophthalmic examination and photographic imaging make ocular abnormalities visible.16 By capturing a pictorial record of eye disease, photography is an integral part of clinical practice and research. As grading systems of ocular photographs can be unreliable,17 computerized evaluation is emerging as an objective approach for quantifying the severity of ocular disorders.18

Biometric analysis has been successfully used to assess posterior lens capsule opacification, diabetic retinopathy, and age-related macular degeneration.14,1921 Imaging technology also offers a way to evaluate conditions affecting the anterior segment such as conjunctivitis,22 corneal epithelial defects, and bacterial keratitis.23 Recent advances allow the size, shape, and haze of corneal disorders to be determined, nearly as accurately as with slitlamp biomicroscopy.9,2426 Image analysis systematizes the representation of ophthalmic disease.

This cross-sectional study evaluated the performance of photographic image acquisition and processing in the initial assessment of herpes simplex virus keratitis. To assimilate clinical images taken on different days and at various centers, sessions included measurement and density calibration.27 Photographic slitlamp cameras varied slightly in magnification and lighting, so image standardization helped to compensate for technical variations in optics and luminance. Among the different adjustments in magnification, illumination, and angle of view, photographs taken at higher magnification or with a thin slitbeam proved less informative than those taken at a 6× setting with diffuse illumination. Sclerotic scatter exaggerated the reflectance of stromal keratitis but highlighted the margins of corneal infiltration and edema and helped to guide planimetry. Shifting the eye position to bring the zone of stromal keratitis straight ahead gave results similar to those when having the eye in primary position, indicating that biomicroscopic measurements can be made without redirecting the patient's gaze.

Bioimaging revealed the geometry of herpes simplex virus stromal keratitis. With the use of an interactive edge-detection algorithm, the zone of corneal inflammatory opacification on digitized images was circumscribed by a curvilinear perimeter to yield a summed area of enclosed pixels. Mapping showed that stromal keratitis tended to cluster near the geometric center of the cornea, often in the temporal and inferior quadrants. Contour was evaluated by the shape factor, a geometric parameter that gauges the relative circularity of an object regardless of its border conformation. Despite an irregular perimeter for some eyes, half of the eyes in this series had stromal keratitis with a shape factor of 0.56 to 0.79, indicating an oval outline. Most could be morphometrically described as disciform. Larger keratitis became rounder and more likely to be associated with anterior uveitis. This study demonstrated that the size of herpes simplex virus stromal keratitis estimated by image analysis correlated closely with elliptically transformed clinical area measurements.

The severity of stromal keratitis was also judged by its relative whiteness. Although a finer scale of clinical classification could improve precision and reliability,28,29 the majority of stromal keratitis was at least moderately hazy. After converting color images into gray tones, the averaged densitometric value of stromal keratitis provided a numerical measure of relative opacification30 and showed that densely white inflammation was a feature of necrotizing inflammation. Visual acuity worse than 20/40 occurred more often when stromal inflammation was larger, more opaque, or centrally located.

This validation study is subject to some technical limitations. We used scanned images of biomicroscopic photographs rather than direct digital imaging, although measurements should not depend on the image-capturing method.31 Image analysis was problematic if the margins of stromal inflammation and edema could not be clearly distinguished or if residual opacification remained from a prior episode. Light reflections from the tear film and variations of the iris surface also affected the ability to discriminate the transition zone at the border of stromal keratitis. Because clinical scales of ocular inflammation are not necessarily linear,32 comparing categorical disease severity with gray-level intensity was imprecise.

Further improvements in image resolution and microdensitometry offer prospects for using ophthalmic metadata in practice and teleresearch. Computerized analysis extends the clinical utility of anterior segment imaging from photodocumentation to quantitative description, comparison, and follow-up of corneal disease.

Back to top
Article Information

Correspondence: Kirk R. Wilhelmus, MD, PhD, Sid W. Richardson Ocular Microbiology Laboratory, Department of Ophthalmology, Cullen Eye Institute, Baylor College of Medicine, 6565 Fannin St, Houston, TX 77030 (kirkw@bcm.tmc.edu).

Submitted for Publication: April 7, 2008; final revision received September 2, 2008; accepted September 12, 2008.

Financial Disclosure: None reported.

Funding/Support: This study was supported by the National Eye Institute through cooperative agreements EY07479, EY07480, EY07482, EY07483, EY07486, EY07487, EY07488, EY07489, and EY07496, by core grant EY02520 from the National Institutes of Health, and by Research to Prevent Blindness, Inc, and the Sid W. Richardson Foundation.

Box Section Ref ID

Herpetic Eye Disease Study Group Members

Clinical Centers

Personnel are listed as investigator (I), coordinator (C), and photographer (P).

Francis I. Proctor Foundation, University of California, San Francisco: Chandler R. Dawson (I), Todd P. Margolis (I), Robert A. Nozik (I), H. Bruce Ostler (I), Suzanne Banuvar (C, P); LSU Eye Center, Louisiana State University Health Sciences Center, New Orleans: Bruce A. Barron (I), Michael S. Insler (I), Herbert E. Kaufman (I), Dale Williams (C), Dale O. Pennell (P); Cullen Eye Institute, Baylor College of Medicine, Houston, Texas: Dan B. Jones (I), Alice Y. Matoba (I), Kirk R. Wilhelmus (I), Loretta A. Todaro (C), Lawrence M. Merin (P); Emory Eye Center, Emory University, Atlanta, Georgia: R. Doyle Stulting (I), George O. Waring III (I), Louis A. Wilson (I), Diana S. Coffman (C), Lindy G. DuBois (C), James P. Gilman (P); Eye Institute, Medical College of Wisconsin, Milwaukee: Robert A. Hyndiuk (I), Steven B. Koenig (I), Bruce M. Massaro (I), Mary Ann Fitzgerald (C), Linda Speich (C), Walter Wipplinger (P); Mount Sinai Medical Center, New York, New York: Penny A. Asbell (I), Andrew P. Davis (I), Michael J. Newton (I), Norma Justin (C), F. Robert Masini (P); Illinois Eye and Ear Infirmary, University of Illinois at Chicago: Joel Sugar (I), Sheridan Lam (I), Jeffrey B. Robin (I), Howard H. Tessler (I), Kay Rosheim (C), Marty Lindeman (C), Norbert J. Jednock (P); Wills Eye Hospital, Jefferson Medical College, Thomas Jefferson University, Philadelphia, Pennsylvania: Peter R. Laibson (I), Elisabeth J. Cohen (I), Kent G. Leavitt (I), Christopher J. Rapuano (I), Cari Montanez (C), Nubia Richman (C), Jamie Nicholl (P); and W. K. Kellogg Eye Institute, University of Michigan, Ann Arbor: Alan Sugar (I), Roger F. Meyer (I), H. Kaz Soong (I), Cindy Pope (C), Csaba L. Martonyi (P).

Resource Centers and Committees

Coordinating Center (1992-1994): University of California, San Francisco: Walter W. Hauck (director), Lauren Gee; Coordinating Center (1994-1999): Jaeb Center for Health Research, Inc, Tampa, Florida: Roy W. Beck (director), Pamela S. Moke; Photography Reading Center: Cullen Eye Institute, Baylor College of Medicine, Houston, Texas: Kirk R. Wilhelmus (director), Lawrence M. Merin, Loretta A. Todaro, Sandra J. Woodside; Steering Committee: Chandler R. Dawson (chair), Walter W. Hauck, Dan B. Jones, Herbert E. Kaufman, Natalie Kurinij; and Data and Safety Monitoring Committee: Shrikant Bangdiwala (chair), William E. Barlow, John W. Chandler, Lawrence J. Nelson, Anthony B. Nesburn.

References
1.
Dawson  CRJones  DBKaufman  HEBarron  BAHauck  WWWilhelmus  KR Design and organization of the Herpetic Eye Disease Study (HEDS). Curr Eye Res 1991;10 ((suppl)) 105- 110
PubMedArticle
2.
Sudesh  SLaibson  PR The impact of the Herpetic Eye Disease Studies on the management of herpes simplex virus ocular infections. Curr Opin Ophthalmol 1999;10 (4) 230- 233
PubMedArticle
3.
Wilhelmus  KRGee  LHauck  WW  et al.  A controlled trial of topical corticosteroids for herpes simplex stromal keratitis. Ophthalmology 1994;101 (12) 1883- 1895
PubMedArticle
4.
Barron  BAGee  LHauck  WW  et al.  A controlled trial of oral acyclovir for herpes simplex stromal keratitis. Ophthalmology 1994;101 (12) 1871- 1882
PubMedArticle
5.
Wilhelmus  KR Diagnosis and management of herpes simplex stromal keratitis. Cornea 1987;6 (4) 286- 291
PubMedArticle
6.
Williams  HPFalcon  MGJones  BR Corticosteroids in the management of herpetic eye disease. Trans Ophthalmol Soc U K 1977;97 (2) 341- 344
PubMed
7.
Merin  LMWilhelmus  KRHerpetic Eye Disease Study Group, Anterior segment photography in the Herpetic Eye Disease Study. J Ophthalmic Photogr 1992;14 (1) 34- 38
8.
Herpetic Eye Disease Study Group, Manual of Operations.  Springfield, VA National Technical Information Service1990;NTIS publication PB93175453
9.
Mukerji  NVajpayee  RBSharma  N Technique of area measurement of epithelial defects. Cornea 2003;22 (6) 549- 551
PubMedArticle
10.
Committee on Vision, Assembly of Behavioral and Social Sciences, National Research Council, National Academy of Sciences, Recommended standard procedures for the clinical measurement and specification of visual acuity: report of working group 39. Adv Ophthalmol 1980;41103- 148
PubMed
11.
Wilhelmus  KRGay  CAMerin  LM  et al.  Image analysis of herpes simplex stromal keratitis.  Paper presented at: International Conference on Herpetic Eye Diseases July 26, 1990 Oxford, England
12.
Wolfe  JKChylack  LT  Jr Objective measurement of cortical and subcapsular opacification in retroillumination photographs. Ophthalmic Res 1990;22 ((suppl 1)) 62- 67
PubMedArticle
13.
Klein  BEKKlein  RLinton  KLPMagli  YLNeider  MW Assessment of cataracts from photographs in the Beaver Dam Eye Study. Ophthalmology 1990;97 (11) 1428- 1433
PubMedArticle
14.
Bender  LSpalton  DJUyanonvara  B  et al.  POCOman: new system for quantifying posterior capsule opacification. J Cataract Refract Surg 2004;30 (10) 2058- 2063
PubMedArticle
15.
Age-Related Eye Disease Study Research Group, The Age-Related Eye Disease Study (AREDS) system for classifying cataracts from photographs: AREDS report No. 4. Am J Ophthalmol 2001;131 (2) 167- 175
PubMedArticle
16.
Constable  IJ Foreword. Yogesan  KKumar  SGoldschmidt  LCuadros  JTeleophthalmology. Berlin, Germany Springer-Verlag2006;vii
17.
Solomon  AWBowman  RJCYorston  D  et al.  Operational evaluation of the use of photographs for grading active trachoma. Am J Trop Med Hyg 2006;74 (3) 505- 508
PubMed
18.
Gilchrist  J Computer processing of ocular photographs: a review. Ophthalmic Physiol Opt 1987;7 (4) 379- 386
PubMed
19.
Bartlett  HEperjesi  F Use of fundus imaging in quantification of age-related macular change. Surv Ophthalmol 2007;52 (6) 655- 671
PubMedArticle
20.
Patton  NAslam  TMMacGillivray  T  et al.  Retinal image analysis: concepts, applications and potential. Prog Retin Eye Res 2006;25 (1) 99- 127
PubMedArticle
21.
Niemeijer  Mvan Ginneken  BRussell  SRSuttorp-Schulten  MSAAbràmoff  MD Automated detection and differentiation of drusen, exudates, and cotton-wool spots in digital color fundus photographs for diabetic retinopathy diagnosis. Invest Ophthalmol Vis Sci 2007;48 (5) 2260- 2267
PubMedArticle
22.
Fieguth  PSimpson  T Automated measurement of bulbar redness. Invest Ophthalmol Vis Sci 2002;43 (2) 340- 347
PubMed
23.
Zegans  METoutain  CSaine  P  et al.  Analysis of digital images of corneal ulcers in the pilot study for the Steroid Corneal Ulcer Trial (SCUT).  Poster presented at: Annual Meeting of the Association for Research in Vision and Ophthalmology May 1, 2006 Fort Lauderdale, FL
24.
VanRoekel  RCBower  KSBurka  JMHoward  RS Anterior segment measurements using digital photography: a simple technique. Optom Vis Sci 2006;83 (6) 391- 395
PubMedArticle
25.
Chan  JWEdwards  MHWoo  GCWoo  VC Objective method to measure corneal clarity before and after laser in situ keratomileusis. J Cataract Refract Surg 2003;29 (1) 118- 124
PubMedArticle
26.
Barr  JTWilson  BSGordon  MO  et al. CLEK Study Group, Estimation of the incidence and factors predictive of corneal scarring in the Collaborative Longitudinal Evaluation of Keratoconus (CLEK) Study. Cornea 2006;25 (1) 16- 25
PubMedArticle
27.
Shin  DSKaiser  RSLee  MSBerger  JW Fundus image change analysis: geometric and radiometric normalization. Proc SPIE 1999;3591129- 136
28.
Bailey  ILBullimore  MARaasch  TWTaylor  HR Clinical grading and the effects of scaling. Invest Ophthalmol Vis Sci 1991;32 (2) 422- 432
PubMed
29.
Chong  ESimpson  TFonn  D The repeatability of discrete and continuous anterior segment grading scales. Optom Vis Sci 2000;77 (5) 244- 251
PubMedArticle
30.
Chylack  LT  JrRosner  BWhite  OTung  WHSher  LD Standardization and analysis of digitized photographic data in the longitudinal documentation of cataractous growth. Curr Eye Res 1988;7 (3) 223- 235
PubMedArticle
31.
Musadiq  MPatsoura  EHughes  SYang  YC Measurements of linear dimensions on fundus photographs: comparison between photographic film and digital systems. Eye 2003;17 (5) 619- 622
PubMedArticle
32.
Wolffsohn  JS Incremental nature of anterior eye grading scales determined by objective image analysis. Br J Ophthalmol 2004;88 (11) 1434- 1438
PubMedArticle
×