- Email: [email protected]
Grading of corneal transparency
Contact Lens & Anterior Eye 27 (2004) 161–170 www.elsevier.com/locate/clae
Grading of corneal transparency Clare O’Donnella,*, James S Wolffsohnb a
Department of Optometry and Neuroscience, UMIST, PO Box 88, Manchester M60 1QD, UK b Neurosciences Research Institute, Aston, Birmingham, UK
Abstract Aim: To examine the academic literature on the grading of corneal transparency and to assess the potential use of objective image analysis. Method: Reference databases of academic literature were searched and relevant manuscripts reviewed. Annunziato, Efron (Millennium Edition) and Vistakon–Synoptik corneal oedema grading scale images were analysed objectively for relative intensity, edges detected, variation in intensity and maximum intensity. In addition, corneal oedema was induced in one subject using a low oxygen transmissibility (Dk/ t) hydrogel contact lens worn for 3 h under a light eye patch. Recovery from oedema was monitored over time using ultrasound pachymetry, high and low contrast visual acuity measures, bulbar hyperaemia grading and transparency image analysis of the test and control eyes. Results: Several methods for assessing corneal transparency are described in the academic literature, but none have gained widespread use in clinical practice. The change in objective image analysis with printed scale grade was best described by quadratic parametric or sigmoid 3parameter functions. ‘Pupil image scales’ (Annunziato and Vistakon–Synoptik) were best correlated to average intensity; however, the corneal section scale (Efron) was strongly correlated to variations in intensity. As expected, patching an eye wearing a low Dk/t hydrogel contact lens caused a significant (F = 119.2, p < 0.001) 14.3% increase in corneal thickness, which gradually recovered under open eye conditions. Corneal section image analysis was the most affected parameter and intensity variation across the slit width, in isolation, was the strongest correlate, accounting for 85.8% of the variance with time following patching, and 88.7% of the variance with corneal thickness. Conclusion: Corneal oedema is best determined objectively by the intensity variation across the width of a corneal section. This can be easily measured using a slit-lamp camera connected to a computer. Oedema due to soft contact lens wear is not easily determined over the pupil area by sclerotic scatter illumination techniques. # 2004 British Contact Lens Association. Published by Elsevier Ltd. All rights reserved. Keywords: Transparency; Oedema; Cornea; Grading; Image analysis
1. Introduction Corneal oedema and transparency is usually monitored clinically with the use of a slit-lamp biomicroscope. Although this instrument offers the advantages of variable magnification and illumination techniques, the subjective monitoring of oedema is limited to the observation of signs such as microcysts, striae and folds, and perhaps the comparison of corneal haze against grading scale images. The ability to monitor changes in corneal transparency non-invasively and objectively could assist in improving patient management. Measurement of corneal transparency could provide a useful benchmark when evaluating the ocular response to different types and modalities of contact * Corresponding author. Tel.: +44 161 200 3872; fax: +44 161 200 4442. E-mail address: [email protected] (C. O’Donnell).
lens wear. There would also be considerable benefits of being able to quantify the degree of sub-epithelial haze and stromal scarring in refractive surgery patients in the clinical setting. The transparency of the cornea results from the fact that the normal cornea does not absorb visible light and light scatter is minimal. Maurice [1] explained the transparency of the cornea on the basis of the uniform diameter and regular separation of stromal collagen. Maurice suggested that the collagen fibrils of the stroma were arranged regularly in a lattice and that scattered light is eliminated by destructive interference so that only forward travelling light is permitted [1]. Factors involved in maintaining collagen fibril size and spatial order are not fully understood. It has been proposed that collagen fibril diameters may be controlled by the incorporation of minor collagens [2] and that fibril spacing is
1367-0484/$ – see front matter # 2004 British Contact Lens Association. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.clae.2004.08.001
162
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
a function of proteoglycan–collagen interaction. Reduced corneal transparency occurs with age, due to the threedimensional growth of collagen fibrils in the stroma [3]. The state of corneal hydration is an important factor in corneal transparency. Physiologically, corneal hydration is maintained at approximately 78%. When the cornea swells, light scatter increases with a loss in transparency due to the disruption of the collagen matrix. The collagen fibrils themselves swell very little and most of the additional water enters the inter-fibrillar spaces. 1.1. Techniques for assessing corneal transparency The measurement of light scatter has been used to probe characteristics of the corneal fibrillar matrix, based on the assumption that fibrils are the dominant source of scattering within the living cornea. Other possible sources of scattering in the cornea are the epithelial and endothelial cells, Bowman’s layer, keratocytes and folds or striae (where present). 1.1.1. Confocal microscopy Discussions of corneal transparency centre on the ultrastructure and optical properties of the cornea, particu-
larly the stroma. Most of the information on corneal ultrastructure has come from electon microscopy [4]. However, in recent times in vivo confocal microscopy has been used to make direct observations on corneal tissue, avoiding the shrinkage and distortion associated with conventional processing and sectioning. Confocal microscopy enables visualisation of all corneal layers even weakly reflecting layers like the basal cells. Although stromal collagens and ground substance are not directly visible in the healthy cornea with confocal microscopy, the technique does enable the early detection of deposits and stromal matrix disruption (Fig. 1b). Confocal microscopy through focusing (CMTF) [5] has been shown to be a powerful tool that enables quantitative evaluation of refractive surgery outcomes over time by providing corneal thickness analysis, photoablation depth assessment, and unbiased haze measurement [6]. MollerPederson et al. [7] used CMTF to assess back scattered light from the anterior cornea as an objective estimate of corneal haze in rabbits after transepithelial photoablation. The magnitude of corneal wound repair and the development and duration of corneal haze increased proportionally with the volume of stromal tissue removed [7]. CMTF has also been
Fig. 1. Confocal images, prior to (left) and after (right) maximally induced oedema. (a) basal layer (b) anterior stroma (c) posterior stroma (d) endothelium. (e) Confocal microscopy through focussing (CMTF) scan showing reflected light intensity (y-axis) and distance into the cornea (x-axis) (courtesy of C. O’Donnell and C. Maldonado-Codina, 1999).
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
used on lumican-deficient mice to quantify epithelial and stromal thickness and corneal light-scattering [8,9]. Pisella et al. [10] carried out a confocal microscopy evaluation of corneal changes after laser assisted in situ keratomileusis (LASIK). The z-scan feature was used to evaluate haze after LASIK for myopia and the authors concluded that haze was negatively correlated with flap thickness [10]. 1.1.2. X-ray X-ray scattering patterns have been used to determine the biological structure of corneal collagen. When X-rays interact with biological tissues they are scattered through differing angles. Depending on the degree of order in the structure, different levels of interference between the scattered rays will occur. X-ray diffraction techniques have been applied to samples of diseased human corneas, as well as to corneas post-refractive surgery [8,11–13]. 1.1.3. Scheimpflug photography The Scheimpflug principle enables cross-sectional images of the anterior eye to be captured (Fig. 2). It can be used for the assessment of cataract [14–16] and for measuring corneal curvature and thickness [17]. Some researchers have examined the transparency of the cornea using this technique [18–21], but as the image covers a relatively large area, resolution can be an issue. The
163
objective analysis of tissue transparency involves processing the image, usually in terms of the average greyscale of a selected area. Smith et al. [18] found no correlation in healthy subjects between the amount of light scatter and age, but found that more light was scattered at the anterior and posterior surfaces of the cornea, rather than from the stroma. Recently the Nidek EAS-1000 Scheimpflug camera has been adapted and evaluated for use as an objective corneal ‘hazemeter’ [22,23]. 1.1.4. Polarised microscopy Evaluation of corneal birefringence provides an alternative way to monitor wound healing and tissue remodelling, both qualitatively and quantitatively. Huang and colleagues [24] attempted a quantitative measurement of birefringence, in terms of the optical path difference using a polarized microscopy technique based on the analysis of interference colours. The results showed that certain agents effectively prevented deep corneal ulceration in rabbit corneas injured with alkali. Some of these agents were found to delay the onset of ulceration and the treated corneas appeared to have better transparency than the control corneas [24]. 1.1.5. Slit-lamp biomicroscopy Slit-lamp biomicroscopy is the most popular clinical technique for assessing corneal transparency and integrity.
Fig. 2. Scheimpflug image of the anterior eye taken with a Pentacam (Oculus Inc).
164
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
Subjective viewing has been used in many applications, such as evaluations of the effects of excimer laser surgery on corneal transparency [25], the efficacy of agents to treat haze after PRK [26], corneal changes with contact lens wear in diabetics [27] and to assess the usefulness of collagen crosslinking for altering the progression of keratoconus [28]. Initial attempts to objectively grade anterior eye transparency using slit-lamp biomicroscopy involved the use of video camera attachments and image digitisers [29]. Darkfield illumination has been used to evaluate the transparency of donor corneal tissue for transplantation, but this technique cannot be applied to in vivo human corneas [30]. Virtually all other attempts to objectively grade corneal transparency have been related to refractive surgery studies although Lohmann and colleagues [31] compared corneal light scatter in myopic individuals with spectacles, RGP or soft contact lenses and excimer laser PRK, and Chan et al. [32] assessed the clarity of the cornea after soft contact lens wear as well as in LASIK patients. Typically an optic section of the cornea is imaged with a charged couple device (CCD) and the image analysed by a frame grabber [33] or directly with computer software [34,35]. Descriptions of the technique vary (with terms such as opacification index determination, corneal clarity index measurement, videography and microdensitometry used), but essentially the average greyscale of the area selected in an 8 bit (256 level) monochrome image is calculated [32,36], with edge detection algorithms applied initially in some studies [35,37]. Polarisers can also be inserted in front of the light-source to remove reflected light, allowing backscattered light to be assessed [32,33,38]. Corbett and colleagues [38] examined the cornea retro-illuminated in addition to using an optic section. The repeatability of objective image analysis techniques has been shown to be good [32,37]. Although several methods for assessing corneal transparency are described in the academic literature, none have gained widespread usage in clinical practice. The introduction of desktop computers into eye care practices and the decreasing costs of digital imaging technology offers the potential for corneal transparency (and other ocular parameters) to be assessed objectively. This paper examines the incremental nature of corneal oedema printed grading scale images and the most appropriate illumination technique and image analysis variable for quantifying oedema induced by a hydrogel contact lens.
the pupil, the Efron grading scale images are pictorial sections of the cornea including features such as striae and folds and the Vistakon–Synoptik images are photographic images of the pupil illuminated by sclerotic scatter. The other commonly used grading scale, from the CCLRU, only has representative images of oedematous signs, such as microcysts and striae, rather than incremental scale images of corneal oedema. The CCLRU scale was therefore not included in this work. Each of the grading scale images was converted into 8 bit greyscale (256 shades). The average intensity, edges detected by a 3 3 kernal, variation in intensity across the width of the corneal section (standard deviation) and maximum intensity of the corneal area was determined three times for the manually selected area of interest. The results were curve fitted using linear (y = mx + c), quadratic (y = ax2 + bx + c) and sigmoid 3-parameter (y = a/1 + e (x x0)/b) functions. The second aim of this work was to determine the most appropriate corneal illumination technique and image analysis variable for assessing corneal oedema in vivo. This work was approved by the Institutional Research Ethics Committee at UMIST, Manchester, UK and written informed consent was obtained from the subject (JSW) after the nature and possible consequences of the study were discussed. Corneal thickness (Allergan-Humphrey 850 ultrasonic pachometer, CA, USA) and logMAR distance high (90%) and low (10%) contrast visual acuity (normal room illumination) was measured for both eyes of the subject. Digital images of the bulbar conjunctiva (5 magnification, under diffuse illumination), the cornea over the pupil (pupils dilated with 50 ml 0.5% Tropicamide Hydrochloride, 8 magnification, sclerotic scatter illumination) and a corneal section (8 magnification, direct illumination 1 mm beam at 458 to the vertical) were taken (Fig. 3). To induce corneal oedema, a soft contact lens (nominal parameters +10.00DS, 8.60 base curve, 13.80 diameter
2. Methods Images from the Annunziato [39], Efron (Millenium Edition) [40] and Vistakon–Synoptik [41] grading scales were scanned at a resolution of 600 dpi, stored in lossless tagged image format (TIFF) and analysed using a purpose written program (Labview, National Instruments, USA). The Annunziato images are pictorial images of transparency over
Fig. 3. Corneal section viewed through a slit-lamp biomicroscope. (a) prior to lens wear (b) loss of corneal transparency with maximum swelling in the present study (c) stria and (d) endothelial fold.
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
Durasoft 2; Phemfilcon A 38% water content material, Dk 9, centre thickness 0.07 mm; CIBA Vision, USA) was placed onto the right cornea and the eye was lightly patched for 3 h. After 3 h, the lens and patch were removed and the clinical measures described above were repeated at 15–30 min intervals on both eyes, until corneal thickness in the test eye had returned to baseline values. Corneal confocal microscopy images (ConfoScan 3, NIDEK, Japan) were taken of the right eye at baseline and immediately after removing the contact lens and patch. 2.1. Statistical analysis Analysis of variance was used to examine overall effects with time and Tukey’s pairwise multiple comparison test was used to assess individual time differences. The variance accounted for by image analysis measures was assessed using best subset Pearson’s product moment correlation.
3. Results 3.1. Grading scale images Best-fit linear analysis using average intensity (A), edge detection (E), variation in intensity (V) and maximum intensity (M) to describe grading scale variation showed that greatest variance was described by: Annunziato: 5.8 + 0.4E + 0.1A (adj r2 = 0.94) Efron: 26.5 + 6.4E 0.4A + 0.1M (adj r2 = 0.96) Vistakon–Synoptik: 4.4 + 1.0E + 0.1A (adj r2 = 0.97). However, the change in scale grade was best described by quadratic parametric or sigmoid 3-parameter functions (Table 1; Fig. 4). The pictorial (Annunziato) and photographic (Vistakon–Synoptik) pupil image scales were best correlated to average intensity. However, the Efron grading scale showed a decrease in average intensity with increasing scale grade, due to the artistic rendition of oedematous features pictorial scale, with variation in intensity being the strongest correlate. 3.2. Oedema trial Patching an eye wearing a low oxygen transmissibility hydrogel contact lens caused a significant (F = 119.2, p <0.001) 14.3% increase in corneal thickness, which gradually thinned under open eye conditions (Fig. 5). Although both high and low contrast distance visual acuity were affected by the induced corneal oedema, the effect returned to baseline within 55 min (Fig. 5). Bulbar hyperaemia increased significantly from baseline (F = 12.0, p <0.001) for 20 min following removal of the contact lens and patch and transient contra-lateral, increased redness
165
Table 1 Regression variance (r2) with image analysis parameter and regression function for each of the transparency grading scales Annunziato Efron Vistakon–Synoptik Average intensity
Linear 0.95a Quadratic 0.95a Sigmoid 0.93a
0.89 0.90 0.87
0.96a 0.98a 0.94a
Edge detection
Linear 0.63a Quadratic 0.67a Sigmoid 0.80a
0.61 0.98 0.98
0.04 0.76 0.53
Intensity variation
Linear 0.07 Quadratic 0.11 Sigmoid 0.13
0.86a 0.88a 0.88a
0.67a 0.76a 0.88a
0.44a 0.53a 0.00a
0.89a 0.91a 0.90a
Maximum intensity Linear 0.94a Quadratic 0.95a Sigmoid 0.93a
a Indicates variables showing an incremental increase between at least three out of four increments.
was also evident (F = 3.4, p <0.01; Fig. 6). The average standard deviation of three repeated measures of each corneal section image relative to the change with maximal oedema (compared to baseline) was similar for maximum intensity (2.8%), intensity variation (3.1%) and maximum intensity (2.8%) and was lower in the control eye (1.0; 1.4; 1.2%, respectively). However, the variation between images was larger for edge detection (7.7%) and was similar in the control eye (7.8%). Image analysis of a corneal section was able to determine changes in oedema by increases in average intensity (F = 167.0, p <0.001); detected edges (F = 91.0, p <0.001; only at 0, 85 and 100 min following lens removal); intensity variation (F = 168.9, p <0.001); and maximum intensity (F = 190.0, p <0.001; Fig. 7). Best-fit linear analysis identified that intensity variation across the slit width, in isolation, was the strongest correlate, accounting for 85.8% of the variance with recovery time and 88.7% of the variance with corneal thickness. Image analysis over the pupil zone showed significant changes with time (average intensity: F = 25.1, p <0.001; detected edges: F = 22.1, p <0.001; intensity variation: F = 35.7, p <0.001; maximum intensity: F = 20.5, p <0.001), but only average intensity showed an increase from baseline values immediately following removal of the lens and patch and this had dissipated within 20 min of open eye recovery (Fig. 8). Slit-lamp observations of striae and folds were present until the corneal oedema had reduced to less than 4 and 7%, respectively (Fig. 3). Confocal microscopy images of the oedematous eye, compared to baseline, showed increased clarity of the basal cell borders, decreased definition of the anterior stromal keratocyte nuclei and the presence of horizontal and vertical lines in the posterior corneal stroma. The corneal endothelium appeared to be relatively unaffected (Fig. 1).
166
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
Fig. 4. Average intensity, edge detection, intensity variation and maximum intensity for Annunziato, Efron and Vistakon–Synoptik grading scale images. Error bars = 1 S.D. of three repeat readings.
4. Discussion Corneal clarity is an important indicator of corneal physiology. Although contact lenses and refractive surgery techniques are constantly being improved, more sensitive measures of corneal transparency have the potential to identify future complications with contact lens wear and to allow more sensitive monitoring of corneal integrity following surgery, trauma and disease. The grading scale images were best described by quadratic parametric or sigmoid functions. This could be considered appropriate for clinical human grading as the majority of eyes have minimal oedema, so a smaller increment between low grades will make this part of the scale more sensitive. However, this could lead to errors if clinicians interpolate between scale grade images to improve sensitivity [42–44] or if a 0.1 improvement at one end of the scale is considered equivalent to a similar change at the other end. The strong correlation with a
3-parameter sigmoid function suggests that the quadratic increase in increments between grades is not maintained for the top grades, perhaps in photographic scales because of the limited availability of severe oedematous images. As would be expected, grading scales of corneal oedema imaged over the pupil with sclerotic scatter illumination (Annunziato pictorial and Vistakon–Synoptik photographic scales), were best correlated to average intensity. However, oedema induced by patching an eye wearing a low Dk/t hydrogel contact lens for 3 h did not induce consistent marked changes in average intensity or any of the other parameters measured in this way. It can, therefore, be concluded, that observation of light scatter over the pupil is not the most sensitive way to monitor moderate changes (<15% increase in corneal thickness) in soft contact lensinduced corneal oedema. The Efron grading scale of an oedematous corneal section was correlated most strongly to the variation in intensity
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
167
Fig. 5. Corneal thickness and distance visual acuity with time for the right eye with induced oedema and the left control eye. Error Bars = 1 S.D.
across the section width, as artistic rendition of oedematous features resulted in a decrease in average intensity with increasing scale grade. Such artistic rendition can represent a range of magnifications, illuminations and slit-lamp biomicroscope angles, used to grade the anterior eye in
clinical practice. This range is not always possible with photographic images. The 14.3% increase in corneal thickness induced by patching an eye wearing a low Dk/t soft lens for 3 h caused a significant increase, followed by a progressive decrease
Fig. 6. Bulbar hyperaemia with time for the right eye with induced oedema and the left control eye. Error bars = 1 S.D.
168
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
Fig. 7. Corneal section images: Average intensity, edge detection, intensity variation and maximum intensity changes with time for the eye (filled symbols) with induced oedema. Error bars = 1 S.D. of three repeat readings. For clarity, the control eye and 95% confidence interval over the trial period is represented by the non-filled symbols.
(with corneal recovery) in corneal section average intensity, detected edges, intensity variation and maximum intensity. There were also transient changes in visual acuity and bulbar redness, although these had dissipated when corneal oedema had reduced to 8%. Striae and folds were present until the corneal oedema had reduced to 4 and 7% respectively, although changes in the analysed corneal section images were still evident even when corneal thickness had returned to baseline (Fig. 7). In agreement with the grading scale images, linear analysis identified that intensity variation
across the slit width, in isolation, was the strongest correlate, accounting for between 85–90% of the variance with recovery time and corneal thickness. Therefore, despite previous studies solely examining corneal intensity (average greyscale value), intensity variation across a corneal section would appear to be the most sensitive measure of oedema and related corneal physiology. Compared to clinician variability, even taking the level of induced oedema as the maximal range of grading, repeatability of the objective grading of oedema (2.7% of the scale range) was found to be
Fig. 8. Pupil images: Average intensity, edge detection, intensity variation and maximum intensity changes with time for the eye with induced oedema (filled symbols). Error bars = 1 S.D. of three repeat readings. For clarity, the control eye and 95% confidence interval over the trial period is represented by the nonfilled symbols and maximum intensity values divided by 10.
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
1.9–7.0 times better than subjective grading from printed scales [40,44–46]. In conclusion, transparency is an important indicator of corneal physiology and simple image analysis of the variation across a corneal section imaged by a slit-lamp biomicroscope will provide more sensitive monitoring capability than traditional observation of sclerotic scatter, the presence of striae and folds, and measurement of visual acuity.
References [1] Maurice DM. The structure and transparency of the cornea. J Physiol 1957;136:263–86. [2] Meek KM, Leonard DW. Ultrastructure of the corneal stroma: a comparative study. Biophys J 1993;64:273–80. [3] Daxer A, Misof K, Grabner B, Ettl A, Fratzl P. Collagen fibrils in the human corneal stroma: structure and aging. Invest Ophthalmol Vis Sci 1998;39:644–8. [4] Freegard TJ. The physical basis of transparency of the normal cornea. Eye 1997;11:465–71. [5] O’Donnell C, Maldonado-Codina C. A new paradigm: in vivo confocal microscopy part two. Optician 1999;217(5706):14–6. [6] Moller-Pedersen T, Vogel M, Li HF, Petroll WM, Cavanagh HD, Jester JV. Quantification of stromal thinning, epithelial thickness, and corneal haze after photorefractive keratectomy using in vivo confocal microscopy. Ophthalmology 1997;104:360–8. [7] Moller-Pedersen T, Cavanagh HD, Petroll WM, Jester JV. Corneal haze developed after PRK is regulated by volume of stromal tissue removal. Cornea 1998;17:627–39. [8] Quantock AJ, Meek KM, Chakravarti S. An X-ray diffraction investigation of corneal structure in lumican-deficient mice. Invest Ophthalmol Vis Sci 2001;42:1750–6. [9] Song J, Lee YG, Houston J, Petroll WM, Chakravarti S, Cavanagh HD, et al. Neonatal corneal stromal development in the normal and lumican-deficient mouse. Invest Ophthalmol Vis Sci 2003;44:548–57. [10] Pisella PJ, Auzerie O, Bokobza Y, Debbasch C, Baudouin C. Evaluation of corneal stromal changes in vivo after laser in situ keratomileusis with confocal microscopy. Ophthalmology 2001;108:1744–50. [11] Connon CJ, Marshall J, Patmore AL, Brahma A, Meek KM. Persistent haze and disorganization of anterior stromal collagen appear unrelated following phototherapeutic keratectomy. J Refr Surg 2003;19:323–32. [12] Quantock AJ, Kratz-Owens KL, Leonard DW, Meek KM, Schanzlin DJ. Remodelling of the corneal stroma after lamellar keratoplasty. A synchrotron X-ray diffraction study. Cornea 1994;13:20–7. [13] Quantock AJ, Meek KM, Thonar EJ, Assil KK. Synchrotron X-ray diffraction in atypical macular dystrophy. Eye 1993;7:779–84. [14] Hockwin O, Lerman S, Ohrloff C. Investigations on lens transparency and its disturbances by microdensitometric analyses of Scheimpflug photographs. Curr Eye Res 1984;3:15–22. [15] Wegener A, Laser H. Image-analysis and Scheimpflug-photography of the anterior segment of the eye - a review article. Klinische Monatsblatter Fur Augenheilkunde 2001;218:67–77. [16] Dubbelman M, Van der Heijde GL, Weeber HA, Vrensen GFJM. Changes in the internal structure of the human crystalline lens with age and accommodation. Vision Res 2003;43:2363–75. [17] Morgan AJ, Harper J, Hosking SL, Gilmartin B. The effect of corneal thickness and corneal curvature on pneumatonometer measurements. Curr Eye Res 2002;25:107–12. [18] Smith GT, Brown NA, Shun-Shin GA. Light scatter from the central human cornea. Eye 1990;4:584–8. [19] Binder PS, Bosem M, Weinreb RN. Scheimpflug anterior segment photography assessment of wound healing after myopic excimer laser photorefractive keratectomy. J Cataract Refract Surg 1996;22:205–12.
169
[20] Kaji Y, Obata H, Usui T, Soya K, Machinami R, Tsuru T, et al. Threedimensional organization of collagen fibrils during corneal stromal wound healing after excimer laser keratectomy. J Cataract Refract Surg 1998;24:1441–6. [21] Vinciguerra P, Azzolini M, Radice P, Sborgia M, De Molfetta V. A method for examining surface and interface irregularities after photorefractive keratectomy and laser in situ keratomileusis: predictor of optical and functional outcomes. J Refr Surg 1998;14:Error: FPage (s204) is higher than LPage (s206)!. [22] Soya K, Amano S, Oshika T. Quantification of simulated corneal haze by measuring back-scattered light. Ophthal Res 2002;34:380–8. [23] van de Pol C, Soya K, Hwang DG. Objective assessment of transient corneal haze and its relation to visual performance after photorefractive keratectomy. Am J Ophthalmol 2001;132:204–10. [24] Huang Y, Meek KM, Ho MW, Paterson CA. Analysis of birefringence during wound healing and remodeling following alkali burns in rabbit cornea. Exp Eye Res 2001;73:521–32. [25] Goggin M, Kenna P, Lavery F. Haze following photorefractive and photoastigmatic refractive keratectomy with the Nidek EC5000 and the Summit ExciMed UV200. J Cataract Refract Surg 1997;23:50–3. [26] Vigo L, Scandola E, Carones F, Scraping. mitomycin. C to treat haze and regression after photorefractive keratectomy for myopia. J Refr Surg 2003;19:449–54. [27] O’Donnell C, Efron N, Boulton AJM. A prospective study of contact lens wear in diabetes mellitus. Ophthal Physiol Opt 2001;21:127–38. [28] Wollensak G, Spoerl E, Seiler T. Riboflavin/ultraviolet-A-induced collagen crosslinking for the treatment of keratoconus. Am J Ophthalmol 2003;135:620–7. [29] Fonda S, Baraldi P, Vecchi D, Bocca E. Modifications of the slit-lamp for digital image processing of the anterior segment of the eye. Graefes Arch Clin Exp Ophthalmol 1986;224:284–7. [30] Merin LM, Brown MF, Howdeshell LL. Darkfield biomicrography of eye bank donor corneas. Cornea 2001;20:210–3. [31] Lohmann CP, Fitzke F, O‘Brart D, Muir MK, Timberlake G, Marshall J. Corneal light scattering and visual performance in myopic individuals with spectacles, contact lenses, or excimer laser photorefractive keratectomy. Am J Ophthalmol 1993;115:444–53. [32] Chan JW, Edwards MH, Woo GC, Woo VC. Objective method to measure corneal clarity before and after laser in situ keratomileusis. J Cataract Refract Surg 2003;29:118–24. [33] Lohmann CP, Timberlake GT, Fitzke FW, Gartry DS, Muir MK, Marshall J. Corneal light scattering after eximer laser photorefractive keratectomy: the objective measurement of haze. Refract Corn Surg 1992;8:114–21. [34] Braunstein RE, Jain S, McCally RL, Stark WJ, Connolly PJ, Azar DT. Objective measurement of corneal light scattering after excimer laser keratectomy. Ophthalmology 1996;103:439–43. [35] Maldonado MJ, Arnau V, Navea A, Martinez-Costa R, Mico FM, Cisneros AL, et al. Direct objective quantification of corneal haze after excimer laser photorefractive keratectomy for high myopia. Ophthalmology 1996;103:1970–8. [36] Polunin GS, Kourenkov VV, Polunina EG. Corneal transparency and measurement of corneal permeability in excimer laser photorefractive keratectomy. J Refr Surg 1998;14:S230–4. [37] Maldonado MJ, Arnau V, Martinez-Costa R, Navea A, Mico FM, Cisneros AL, et al. Reproducibility of digital image analysis for measuring corneal haze after myopic photorefractive keratectomy. Am J Ophthalmol 1997;123:31–41. [38] Corbett MC, Prydal JI, Verma S, Oliver KM, Pande M, Marshall J. An in vivo investigation of the structures responsible for corneal haze after photorefractive keratectomy and their effect on visual function. Ophthalmology 1996;103:1366–80. [39] Annunziato T, Davidson RG, Christensen MT, Deloach J, Pazandak BB, Thurburn L, et al. Atlas of slit-lamp findings and contact-lens related anomalies. circa 1992. Fort Worth, TX: Southwest Independent Institutional Review Board.
170
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
[40] Efron N, Morgan PB, Katsara SS. Validation of grading scales for contact lens complications. Ophthal Physiol Opt 2001;21:17–29. [41] Andersen JS, Davies IP, Kruse A, Lofstrom T, Ringmann LA. Handbook of Contact Lens Management. Vistakon Publication; 1996:6–7. [42] Bailey IL, Bullimore MA, Raasch TW, Taylor HR. Clinical grading and the effects of scaling. Invest Ophthalmol Vis Sci 1991;32:422–32. [43] Lloyd M. Lies, statistics, and clinical significance. J Br Contact Lens Assoc 1992;15:67–70.
[44] Twelker JD, Bailey IL. Grading conjunctival hyperaemia using a photography-based method. Invest Ophthalmol Vis Sci 2000;41: s927. [45] Chong T, Simpson T, Pritchard N, Dumbleton K, Richter D, Fonn D. Repeatability of discrete and continuous clinical grading scales. Optom Vis Sci 1996;73:S232. [46] Papas EB. Key factors in the subjective and objective assessment of conjunctival erythema. Invest Ophthalmol Vis Sci 2000;41:687–91.
Grading of corneal transparency Clare O’Donnella,*, James S Wolffsohnb a
Department of Optometry and Neuroscience, UMIST, PO Box 88, Manchester M60 1QD, UK b Neurosciences Research Institute, Aston, Birmingham, UK
Abstract Aim: To examine the academic literature on the grading of corneal transparency and to assess the potential use of objective image analysis. Method: Reference databases of academic literature were searched and relevant manuscripts reviewed. Annunziato, Efron (Millennium Edition) and Vistakon–Synoptik corneal oedema grading scale images were analysed objectively for relative intensity, edges detected, variation in intensity and maximum intensity. In addition, corneal oedema was induced in one subject using a low oxygen transmissibility (Dk/ t) hydrogel contact lens worn for 3 h under a light eye patch. Recovery from oedema was monitored over time using ultrasound pachymetry, high and low contrast visual acuity measures, bulbar hyperaemia grading and transparency image analysis of the test and control eyes. Results: Several methods for assessing corneal transparency are described in the academic literature, but none have gained widespread use in clinical practice. The change in objective image analysis with printed scale grade was best described by quadratic parametric or sigmoid 3parameter functions. ‘Pupil image scales’ (Annunziato and Vistakon–Synoptik) were best correlated to average intensity; however, the corneal section scale (Efron) was strongly correlated to variations in intensity. As expected, patching an eye wearing a low Dk/t hydrogel contact lens caused a significant (F = 119.2, p < 0.001) 14.3% increase in corneal thickness, which gradually recovered under open eye conditions. Corneal section image analysis was the most affected parameter and intensity variation across the slit width, in isolation, was the strongest correlate, accounting for 85.8% of the variance with time following patching, and 88.7% of the variance with corneal thickness. Conclusion: Corneal oedema is best determined objectively by the intensity variation across the width of a corneal section. This can be easily measured using a slit-lamp camera connected to a computer. Oedema due to soft contact lens wear is not easily determined over the pupil area by sclerotic scatter illumination techniques. # 2004 British Contact Lens Association. Published by Elsevier Ltd. All rights reserved. Keywords: Transparency; Oedema; Cornea; Grading; Image analysis
1. Introduction Corneal oedema and transparency is usually monitored clinically with the use of a slit-lamp biomicroscope. Although this instrument offers the advantages of variable magnification and illumination techniques, the subjective monitoring of oedema is limited to the observation of signs such as microcysts, striae and folds, and perhaps the comparison of corneal haze against grading scale images. The ability to monitor changes in corneal transparency non-invasively and objectively could assist in improving patient management. Measurement of corneal transparency could provide a useful benchmark when evaluating the ocular response to different types and modalities of contact * Corresponding author. Tel.: +44 161 200 3872; fax: +44 161 200 4442. E-mail address: [email protected] (C. O’Donnell).
lens wear. There would also be considerable benefits of being able to quantify the degree of sub-epithelial haze and stromal scarring in refractive surgery patients in the clinical setting. The transparency of the cornea results from the fact that the normal cornea does not absorb visible light and light scatter is minimal. Maurice [1] explained the transparency of the cornea on the basis of the uniform diameter and regular separation of stromal collagen. Maurice suggested that the collagen fibrils of the stroma were arranged regularly in a lattice and that scattered light is eliminated by destructive interference so that only forward travelling light is permitted [1]. Factors involved in maintaining collagen fibril size and spatial order are not fully understood. It has been proposed that collagen fibril diameters may be controlled by the incorporation of minor collagens [2] and that fibril spacing is
1367-0484/$ – see front matter # 2004 British Contact Lens Association. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.clae.2004.08.001
162
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
a function of proteoglycan–collagen interaction. Reduced corneal transparency occurs with age, due to the threedimensional growth of collagen fibrils in the stroma [3]. The state of corneal hydration is an important factor in corneal transparency. Physiologically, corneal hydration is maintained at approximately 78%. When the cornea swells, light scatter increases with a loss in transparency due to the disruption of the collagen matrix. The collagen fibrils themselves swell very little and most of the additional water enters the inter-fibrillar spaces. 1.1. Techniques for assessing corneal transparency The measurement of light scatter has been used to probe characteristics of the corneal fibrillar matrix, based on the assumption that fibrils are the dominant source of scattering within the living cornea. Other possible sources of scattering in the cornea are the epithelial and endothelial cells, Bowman’s layer, keratocytes and folds or striae (where present). 1.1.1. Confocal microscopy Discussions of corneal transparency centre on the ultrastructure and optical properties of the cornea, particu-
larly the stroma. Most of the information on corneal ultrastructure has come from electon microscopy [4]. However, in recent times in vivo confocal microscopy has been used to make direct observations on corneal tissue, avoiding the shrinkage and distortion associated with conventional processing and sectioning. Confocal microscopy enables visualisation of all corneal layers even weakly reflecting layers like the basal cells. Although stromal collagens and ground substance are not directly visible in the healthy cornea with confocal microscopy, the technique does enable the early detection of deposits and stromal matrix disruption (Fig. 1b). Confocal microscopy through focusing (CMTF) [5] has been shown to be a powerful tool that enables quantitative evaluation of refractive surgery outcomes over time by providing corneal thickness analysis, photoablation depth assessment, and unbiased haze measurement [6]. MollerPederson et al. [7] used CMTF to assess back scattered light from the anterior cornea as an objective estimate of corneal haze in rabbits after transepithelial photoablation. The magnitude of corneal wound repair and the development and duration of corneal haze increased proportionally with the volume of stromal tissue removed [7]. CMTF has also been
Fig. 1. Confocal images, prior to (left) and after (right) maximally induced oedema. (a) basal layer (b) anterior stroma (c) posterior stroma (d) endothelium. (e) Confocal microscopy through focussing (CMTF) scan showing reflected light intensity (y-axis) and distance into the cornea (x-axis) (courtesy of C. O’Donnell and C. Maldonado-Codina, 1999).
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
used on lumican-deficient mice to quantify epithelial and stromal thickness and corneal light-scattering [8,9]. Pisella et al. [10] carried out a confocal microscopy evaluation of corneal changes after laser assisted in situ keratomileusis (LASIK). The z-scan feature was used to evaluate haze after LASIK for myopia and the authors concluded that haze was negatively correlated with flap thickness [10]. 1.1.2. X-ray X-ray scattering patterns have been used to determine the biological structure of corneal collagen. When X-rays interact with biological tissues they are scattered through differing angles. Depending on the degree of order in the structure, different levels of interference between the scattered rays will occur. X-ray diffraction techniques have been applied to samples of diseased human corneas, as well as to corneas post-refractive surgery [8,11–13]. 1.1.3. Scheimpflug photography The Scheimpflug principle enables cross-sectional images of the anterior eye to be captured (Fig. 2). It can be used for the assessment of cataract [14–16] and for measuring corneal curvature and thickness [17]. Some researchers have examined the transparency of the cornea using this technique [18–21], but as the image covers a relatively large area, resolution can be an issue. The
163
objective analysis of tissue transparency involves processing the image, usually in terms of the average greyscale of a selected area. Smith et al. [18] found no correlation in healthy subjects between the amount of light scatter and age, but found that more light was scattered at the anterior and posterior surfaces of the cornea, rather than from the stroma. Recently the Nidek EAS-1000 Scheimpflug camera has been adapted and evaluated for use as an objective corneal ‘hazemeter’ [22,23]. 1.1.4. Polarised microscopy Evaluation of corneal birefringence provides an alternative way to monitor wound healing and tissue remodelling, both qualitatively and quantitatively. Huang and colleagues [24] attempted a quantitative measurement of birefringence, in terms of the optical path difference using a polarized microscopy technique based on the analysis of interference colours. The results showed that certain agents effectively prevented deep corneal ulceration in rabbit corneas injured with alkali. Some of these agents were found to delay the onset of ulceration and the treated corneas appeared to have better transparency than the control corneas [24]. 1.1.5. Slit-lamp biomicroscopy Slit-lamp biomicroscopy is the most popular clinical technique for assessing corneal transparency and integrity.
Fig. 2. Scheimpflug image of the anterior eye taken with a Pentacam (Oculus Inc).
164
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
Subjective viewing has been used in many applications, such as evaluations of the effects of excimer laser surgery on corneal transparency [25], the efficacy of agents to treat haze after PRK [26], corneal changes with contact lens wear in diabetics [27] and to assess the usefulness of collagen crosslinking for altering the progression of keratoconus [28]. Initial attempts to objectively grade anterior eye transparency using slit-lamp biomicroscopy involved the use of video camera attachments and image digitisers [29]. Darkfield illumination has been used to evaluate the transparency of donor corneal tissue for transplantation, but this technique cannot be applied to in vivo human corneas [30]. Virtually all other attempts to objectively grade corneal transparency have been related to refractive surgery studies although Lohmann and colleagues [31] compared corneal light scatter in myopic individuals with spectacles, RGP or soft contact lenses and excimer laser PRK, and Chan et al. [32] assessed the clarity of the cornea after soft contact lens wear as well as in LASIK patients. Typically an optic section of the cornea is imaged with a charged couple device (CCD) and the image analysed by a frame grabber [33] or directly with computer software [34,35]. Descriptions of the technique vary (with terms such as opacification index determination, corneal clarity index measurement, videography and microdensitometry used), but essentially the average greyscale of the area selected in an 8 bit (256 level) monochrome image is calculated [32,36], with edge detection algorithms applied initially in some studies [35,37]. Polarisers can also be inserted in front of the light-source to remove reflected light, allowing backscattered light to be assessed [32,33,38]. Corbett and colleagues [38] examined the cornea retro-illuminated in addition to using an optic section. The repeatability of objective image analysis techniques has been shown to be good [32,37]. Although several methods for assessing corneal transparency are described in the academic literature, none have gained widespread usage in clinical practice. The introduction of desktop computers into eye care practices and the decreasing costs of digital imaging technology offers the potential for corneal transparency (and other ocular parameters) to be assessed objectively. This paper examines the incremental nature of corneal oedema printed grading scale images and the most appropriate illumination technique and image analysis variable for quantifying oedema induced by a hydrogel contact lens.
the pupil, the Efron grading scale images are pictorial sections of the cornea including features such as striae and folds and the Vistakon–Synoptik images are photographic images of the pupil illuminated by sclerotic scatter. The other commonly used grading scale, from the CCLRU, only has representative images of oedematous signs, such as microcysts and striae, rather than incremental scale images of corneal oedema. The CCLRU scale was therefore not included in this work. Each of the grading scale images was converted into 8 bit greyscale (256 shades). The average intensity, edges detected by a 3 3 kernal, variation in intensity across the width of the corneal section (standard deviation) and maximum intensity of the corneal area was determined three times for the manually selected area of interest. The results were curve fitted using linear (y = mx + c), quadratic (y = ax2 + bx + c) and sigmoid 3-parameter (y = a/1 + e (x x0)/b) functions. The second aim of this work was to determine the most appropriate corneal illumination technique and image analysis variable for assessing corneal oedema in vivo. This work was approved by the Institutional Research Ethics Committee at UMIST, Manchester, UK and written informed consent was obtained from the subject (JSW) after the nature and possible consequences of the study were discussed. Corneal thickness (Allergan-Humphrey 850 ultrasonic pachometer, CA, USA) and logMAR distance high (90%) and low (10%) contrast visual acuity (normal room illumination) was measured for both eyes of the subject. Digital images of the bulbar conjunctiva (5 magnification, under diffuse illumination), the cornea over the pupil (pupils dilated with 50 ml 0.5% Tropicamide Hydrochloride, 8 magnification, sclerotic scatter illumination) and a corneal section (8 magnification, direct illumination 1 mm beam at 458 to the vertical) were taken (Fig. 3). To induce corneal oedema, a soft contact lens (nominal parameters +10.00DS, 8.60 base curve, 13.80 diameter
2. Methods Images from the Annunziato [39], Efron (Millenium Edition) [40] and Vistakon–Synoptik [41] grading scales were scanned at a resolution of 600 dpi, stored in lossless tagged image format (TIFF) and analysed using a purpose written program (Labview, National Instruments, USA). The Annunziato images are pictorial images of transparency over
Fig. 3. Corneal section viewed through a slit-lamp biomicroscope. (a) prior to lens wear (b) loss of corneal transparency with maximum swelling in the present study (c) stria and (d) endothelial fold.
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
Durasoft 2; Phemfilcon A 38% water content material, Dk 9, centre thickness 0.07 mm; CIBA Vision, USA) was placed onto the right cornea and the eye was lightly patched for 3 h. After 3 h, the lens and patch were removed and the clinical measures described above were repeated at 15–30 min intervals on both eyes, until corneal thickness in the test eye had returned to baseline values. Corneal confocal microscopy images (ConfoScan 3, NIDEK, Japan) were taken of the right eye at baseline and immediately after removing the contact lens and patch. 2.1. Statistical analysis Analysis of variance was used to examine overall effects with time and Tukey’s pairwise multiple comparison test was used to assess individual time differences. The variance accounted for by image analysis measures was assessed using best subset Pearson’s product moment correlation.
3. Results 3.1. Grading scale images Best-fit linear analysis using average intensity (A), edge detection (E), variation in intensity (V) and maximum intensity (M) to describe grading scale variation showed that greatest variance was described by: Annunziato: 5.8 + 0.4E + 0.1A (adj r2 = 0.94) Efron: 26.5 + 6.4E 0.4A + 0.1M (adj r2 = 0.96) Vistakon–Synoptik: 4.4 + 1.0E + 0.1A (adj r2 = 0.97). However, the change in scale grade was best described by quadratic parametric or sigmoid 3-parameter functions (Table 1; Fig. 4). The pictorial (Annunziato) and photographic (Vistakon–Synoptik) pupil image scales were best correlated to average intensity. However, the Efron grading scale showed a decrease in average intensity with increasing scale grade, due to the artistic rendition of oedematous features pictorial scale, with variation in intensity being the strongest correlate. 3.2. Oedema trial Patching an eye wearing a low oxygen transmissibility hydrogel contact lens caused a significant (F = 119.2, p <0.001) 14.3% increase in corneal thickness, which gradually thinned under open eye conditions (Fig. 5). Although both high and low contrast distance visual acuity were affected by the induced corneal oedema, the effect returned to baseline within 55 min (Fig. 5). Bulbar hyperaemia increased significantly from baseline (F = 12.0, p <0.001) for 20 min following removal of the contact lens and patch and transient contra-lateral, increased redness
165
Table 1 Regression variance (r2) with image analysis parameter and regression function for each of the transparency grading scales Annunziato Efron Vistakon–Synoptik Average intensity
Linear 0.95a Quadratic 0.95a Sigmoid 0.93a
0.89 0.90 0.87
0.96a 0.98a 0.94a
Edge detection
Linear 0.63a Quadratic 0.67a Sigmoid 0.80a
0.61 0.98 0.98
0.04 0.76 0.53
Intensity variation
Linear 0.07 Quadratic 0.11 Sigmoid 0.13
0.86a 0.88a 0.88a
0.67a 0.76a 0.88a
0.44a 0.53a 0.00a
0.89a 0.91a 0.90a
Maximum intensity Linear 0.94a Quadratic 0.95a Sigmoid 0.93a
a Indicates variables showing an incremental increase between at least three out of four increments.
was also evident (F = 3.4, p <0.01; Fig. 6). The average standard deviation of three repeated measures of each corneal section image relative to the change with maximal oedema (compared to baseline) was similar for maximum intensity (2.8%), intensity variation (3.1%) and maximum intensity (2.8%) and was lower in the control eye (1.0; 1.4; 1.2%, respectively). However, the variation between images was larger for edge detection (7.7%) and was similar in the control eye (7.8%). Image analysis of a corneal section was able to determine changes in oedema by increases in average intensity (F = 167.0, p <0.001); detected edges (F = 91.0, p <0.001; only at 0, 85 and 100 min following lens removal); intensity variation (F = 168.9, p <0.001); and maximum intensity (F = 190.0, p <0.001; Fig. 7). Best-fit linear analysis identified that intensity variation across the slit width, in isolation, was the strongest correlate, accounting for 85.8% of the variance with recovery time and 88.7% of the variance with corneal thickness. Image analysis over the pupil zone showed significant changes with time (average intensity: F = 25.1, p <0.001; detected edges: F = 22.1, p <0.001; intensity variation: F = 35.7, p <0.001; maximum intensity: F = 20.5, p <0.001), but only average intensity showed an increase from baseline values immediately following removal of the lens and patch and this had dissipated within 20 min of open eye recovery (Fig. 8). Slit-lamp observations of striae and folds were present until the corneal oedema had reduced to less than 4 and 7%, respectively (Fig. 3). Confocal microscopy images of the oedematous eye, compared to baseline, showed increased clarity of the basal cell borders, decreased definition of the anterior stromal keratocyte nuclei and the presence of horizontal and vertical lines in the posterior corneal stroma. The corneal endothelium appeared to be relatively unaffected (Fig. 1).
166
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
Fig. 4. Average intensity, edge detection, intensity variation and maximum intensity for Annunziato, Efron and Vistakon–Synoptik grading scale images. Error bars = 1 S.D. of three repeat readings.
4. Discussion Corneal clarity is an important indicator of corneal physiology. Although contact lenses and refractive surgery techniques are constantly being improved, more sensitive measures of corneal transparency have the potential to identify future complications with contact lens wear and to allow more sensitive monitoring of corneal integrity following surgery, trauma and disease. The grading scale images were best described by quadratic parametric or sigmoid functions. This could be considered appropriate for clinical human grading as the majority of eyes have minimal oedema, so a smaller increment between low grades will make this part of the scale more sensitive. However, this could lead to errors if clinicians interpolate between scale grade images to improve sensitivity [42–44] or if a 0.1 improvement at one end of the scale is considered equivalent to a similar change at the other end. The strong correlation with a
3-parameter sigmoid function suggests that the quadratic increase in increments between grades is not maintained for the top grades, perhaps in photographic scales because of the limited availability of severe oedematous images. As would be expected, grading scales of corneal oedema imaged over the pupil with sclerotic scatter illumination (Annunziato pictorial and Vistakon–Synoptik photographic scales), were best correlated to average intensity. However, oedema induced by patching an eye wearing a low Dk/t hydrogel contact lens for 3 h did not induce consistent marked changes in average intensity or any of the other parameters measured in this way. It can, therefore, be concluded, that observation of light scatter over the pupil is not the most sensitive way to monitor moderate changes (<15% increase in corneal thickness) in soft contact lensinduced corneal oedema. The Efron grading scale of an oedematous corneal section was correlated most strongly to the variation in intensity
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
167
Fig. 5. Corneal thickness and distance visual acuity with time for the right eye with induced oedema and the left control eye. Error Bars = 1 S.D.
across the section width, as artistic rendition of oedematous features resulted in a decrease in average intensity with increasing scale grade. Such artistic rendition can represent a range of magnifications, illuminations and slit-lamp biomicroscope angles, used to grade the anterior eye in
clinical practice. This range is not always possible with photographic images. The 14.3% increase in corneal thickness induced by patching an eye wearing a low Dk/t soft lens for 3 h caused a significant increase, followed by a progressive decrease
Fig. 6. Bulbar hyperaemia with time for the right eye with induced oedema and the left control eye. Error bars = 1 S.D.
168
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
Fig. 7. Corneal section images: Average intensity, edge detection, intensity variation and maximum intensity changes with time for the eye (filled symbols) with induced oedema. Error bars = 1 S.D. of three repeat readings. For clarity, the control eye and 95% confidence interval over the trial period is represented by the non-filled symbols.
(with corneal recovery) in corneal section average intensity, detected edges, intensity variation and maximum intensity. There were also transient changes in visual acuity and bulbar redness, although these had dissipated when corneal oedema had reduced to 8%. Striae and folds were present until the corneal oedema had reduced to 4 and 7% respectively, although changes in the analysed corneal section images were still evident even when corneal thickness had returned to baseline (Fig. 7). In agreement with the grading scale images, linear analysis identified that intensity variation
across the slit width, in isolation, was the strongest correlate, accounting for between 85–90% of the variance with recovery time and corneal thickness. Therefore, despite previous studies solely examining corneal intensity (average greyscale value), intensity variation across a corneal section would appear to be the most sensitive measure of oedema and related corneal physiology. Compared to clinician variability, even taking the level of induced oedema as the maximal range of grading, repeatability of the objective grading of oedema (2.7% of the scale range) was found to be
Fig. 8. Pupil images: Average intensity, edge detection, intensity variation and maximum intensity changes with time for the eye with induced oedema (filled symbols). Error bars = 1 S.D. of three repeat readings. For clarity, the control eye and 95% confidence interval over the trial period is represented by the nonfilled symbols and maximum intensity values divided by 10.
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
1.9–7.0 times better than subjective grading from printed scales [40,44–46]. In conclusion, transparency is an important indicator of corneal physiology and simple image analysis of the variation across a corneal section imaged by a slit-lamp biomicroscope will provide more sensitive monitoring capability than traditional observation of sclerotic scatter, the presence of striae and folds, and measurement of visual acuity.
References [1] Maurice DM. The structure and transparency of the cornea. J Physiol 1957;136:263–86. [2] Meek KM, Leonard DW. Ultrastructure of the corneal stroma: a comparative study. Biophys J 1993;64:273–80. [3] Daxer A, Misof K, Grabner B, Ettl A, Fratzl P. Collagen fibrils in the human corneal stroma: structure and aging. Invest Ophthalmol Vis Sci 1998;39:644–8. [4] Freegard TJ. The physical basis of transparency of the normal cornea. Eye 1997;11:465–71. [5] O’Donnell C, Maldonado-Codina C. A new paradigm: in vivo confocal microscopy part two. Optician 1999;217(5706):14–6. [6] Moller-Pedersen T, Vogel M, Li HF, Petroll WM, Cavanagh HD, Jester JV. Quantification of stromal thinning, epithelial thickness, and corneal haze after photorefractive keratectomy using in vivo confocal microscopy. Ophthalmology 1997;104:360–8. [7] Moller-Pedersen T, Cavanagh HD, Petroll WM, Jester JV. Corneal haze developed after PRK is regulated by volume of stromal tissue removal. Cornea 1998;17:627–39. [8] Quantock AJ, Meek KM, Chakravarti S. An X-ray diffraction investigation of corneal structure in lumican-deficient mice. Invest Ophthalmol Vis Sci 2001;42:1750–6. [9] Song J, Lee YG, Houston J, Petroll WM, Chakravarti S, Cavanagh HD, et al. Neonatal corneal stromal development in the normal and lumican-deficient mouse. Invest Ophthalmol Vis Sci 2003;44:548–57. [10] Pisella PJ, Auzerie O, Bokobza Y, Debbasch C, Baudouin C. Evaluation of corneal stromal changes in vivo after laser in situ keratomileusis with confocal microscopy. Ophthalmology 2001;108:1744–50. [11] Connon CJ, Marshall J, Patmore AL, Brahma A, Meek KM. Persistent haze and disorganization of anterior stromal collagen appear unrelated following phototherapeutic keratectomy. J Refr Surg 2003;19:323–32. [12] Quantock AJ, Kratz-Owens KL, Leonard DW, Meek KM, Schanzlin DJ. Remodelling of the corneal stroma after lamellar keratoplasty. A synchrotron X-ray diffraction study. Cornea 1994;13:20–7. [13] Quantock AJ, Meek KM, Thonar EJ, Assil KK. Synchrotron X-ray diffraction in atypical macular dystrophy. Eye 1993;7:779–84. [14] Hockwin O, Lerman S, Ohrloff C. Investigations on lens transparency and its disturbances by microdensitometric analyses of Scheimpflug photographs. Curr Eye Res 1984;3:15–22. [15] Wegener A, Laser H. Image-analysis and Scheimpflug-photography of the anterior segment of the eye - a review article. Klinische Monatsblatter Fur Augenheilkunde 2001;218:67–77. [16] Dubbelman M, Van der Heijde GL, Weeber HA, Vrensen GFJM. Changes in the internal structure of the human crystalline lens with age and accommodation. Vision Res 2003;43:2363–75. [17] Morgan AJ, Harper J, Hosking SL, Gilmartin B. The effect of corneal thickness and corneal curvature on pneumatonometer measurements. Curr Eye Res 2002;25:107–12. [18] Smith GT, Brown NA, Shun-Shin GA. Light scatter from the central human cornea. Eye 1990;4:584–8. [19] Binder PS, Bosem M, Weinreb RN. Scheimpflug anterior segment photography assessment of wound healing after myopic excimer laser photorefractive keratectomy. J Cataract Refract Surg 1996;22:205–12.
169
[20] Kaji Y, Obata H, Usui T, Soya K, Machinami R, Tsuru T, et al. Threedimensional organization of collagen fibrils during corneal stromal wound healing after excimer laser keratectomy. J Cataract Refract Surg 1998;24:1441–6. [21] Vinciguerra P, Azzolini M, Radice P, Sborgia M, De Molfetta V. A method for examining surface and interface irregularities after photorefractive keratectomy and laser in situ keratomileusis: predictor of optical and functional outcomes. J Refr Surg 1998;14:Error: FPage (s204) is higher than LPage (s206)!. [22] Soya K, Amano S, Oshika T. Quantification of simulated corneal haze by measuring back-scattered light. Ophthal Res 2002;34:380–8. [23] van de Pol C, Soya K, Hwang DG. Objective assessment of transient corneal haze and its relation to visual performance after photorefractive keratectomy. Am J Ophthalmol 2001;132:204–10. [24] Huang Y, Meek KM, Ho MW, Paterson CA. Analysis of birefringence during wound healing and remodeling following alkali burns in rabbit cornea. Exp Eye Res 2001;73:521–32. [25] Goggin M, Kenna P, Lavery F. Haze following photorefractive and photoastigmatic refractive keratectomy with the Nidek EC5000 and the Summit ExciMed UV200. J Cataract Refract Surg 1997;23:50–3. [26] Vigo L, Scandola E, Carones F, Scraping. mitomycin. C to treat haze and regression after photorefractive keratectomy for myopia. J Refr Surg 2003;19:449–54. [27] O’Donnell C, Efron N, Boulton AJM. A prospective study of contact lens wear in diabetes mellitus. Ophthal Physiol Opt 2001;21:127–38. [28] Wollensak G, Spoerl E, Seiler T. Riboflavin/ultraviolet-A-induced collagen crosslinking for the treatment of keratoconus. Am J Ophthalmol 2003;135:620–7. [29] Fonda S, Baraldi P, Vecchi D, Bocca E. Modifications of the slit-lamp for digital image processing of the anterior segment of the eye. Graefes Arch Clin Exp Ophthalmol 1986;224:284–7. [30] Merin LM, Brown MF, Howdeshell LL. Darkfield biomicrography of eye bank donor corneas. Cornea 2001;20:210–3. [31] Lohmann CP, Fitzke F, O‘Brart D, Muir MK, Timberlake G, Marshall J. Corneal light scattering and visual performance in myopic individuals with spectacles, contact lenses, or excimer laser photorefractive keratectomy. Am J Ophthalmol 1993;115:444–53. [32] Chan JW, Edwards MH, Woo GC, Woo VC. Objective method to measure corneal clarity before and after laser in situ keratomileusis. J Cataract Refract Surg 2003;29:118–24. [33] Lohmann CP, Timberlake GT, Fitzke FW, Gartry DS, Muir MK, Marshall J. Corneal light scattering after eximer laser photorefractive keratectomy: the objective measurement of haze. Refract Corn Surg 1992;8:114–21. [34] Braunstein RE, Jain S, McCally RL, Stark WJ, Connolly PJ, Azar DT. Objective measurement of corneal light scattering after excimer laser keratectomy. Ophthalmology 1996;103:439–43. [35] Maldonado MJ, Arnau V, Navea A, Martinez-Costa R, Mico FM, Cisneros AL, et al. Direct objective quantification of corneal haze after excimer laser photorefractive keratectomy for high myopia. Ophthalmology 1996;103:1970–8. [36] Polunin GS, Kourenkov VV, Polunina EG. Corneal transparency and measurement of corneal permeability in excimer laser photorefractive keratectomy. J Refr Surg 1998;14:S230–4. [37] Maldonado MJ, Arnau V, Martinez-Costa R, Navea A, Mico FM, Cisneros AL, et al. Reproducibility of digital image analysis for measuring corneal haze after myopic photorefractive keratectomy. Am J Ophthalmol 1997;123:31–41. [38] Corbett MC, Prydal JI, Verma S, Oliver KM, Pande M, Marshall J. An in vivo investigation of the structures responsible for corneal haze after photorefractive keratectomy and their effect on visual function. Ophthalmology 1996;103:1366–80. [39] Annunziato T, Davidson RG, Christensen MT, Deloach J, Pazandak BB, Thurburn L, et al. Atlas of slit-lamp findings and contact-lens related anomalies. circa 1992. Fort Worth, TX: Southwest Independent Institutional Review Board.
170
C. O’Donnell, J.S. Wolffsohn / Contact Lens & Anterior Eye 27 (2004) 161–170
[40] Efron N, Morgan PB, Katsara SS. Validation of grading scales for contact lens complications. Ophthal Physiol Opt 2001;21:17–29. [41] Andersen JS, Davies IP, Kruse A, Lofstrom T, Ringmann LA. Handbook of Contact Lens Management. Vistakon Publication; 1996:6–7. [42] Bailey IL, Bullimore MA, Raasch TW, Taylor HR. Clinical grading and the effects of scaling. Invest Ophthalmol Vis Sci 1991;32:422–32. [43] Lloyd M. Lies, statistics, and clinical significance. J Br Contact Lens Assoc 1992;15:67–70.
[44] Twelker JD, Bailey IL. Grading conjunctival hyperaemia using a photography-based method. Invest Ophthalmol Vis Sci 2000;41: s927. [45] Chong T, Simpson T, Pritchard N, Dumbleton K, Richter D, Fonn D. Repeatability of discrete and continuous clinical grading scales. Optom Vis Sci 1996;73:S232. [46] Papas EB. Key factors in the subjective and objective assessment of conjunctival erythema. Invest Ophthalmol Vis Sci 2000;41:687–91.