Fig. 13.1
Mean epithelial thickness profile for a population of 110 normal eyes and a population of 54 keratoconic eyes. The epithelial thickness profiles for all eyes in each population were averaged using mirrored left eye symmetry. The colour scale represents epithelial thickness in microns. A Cartesian 1-mm grid is superimposed with the origin at the corneal vertex. Reprinted with permission from SLACK Incorporated: Reinstein, DZ., Archer, T., Gobbe M. (2009). “Corneal Epithelial Thickness Profile in the Diagnosis of Keratoconus.” Journal or Refractive Surgery, 25, 604–610
Figure 13.2a shows a B-scan of a normal cornea. The epithelium appears regular in thickness.
Fig. 13.2
Plot showing the mean location of the thinnest epithelium in a population of 110 normal eyes and 54 keratoconic eyes. The blue dot represents the mean location of the thinnest point for the normal population and the dotted blue line represents one standard deviation. The red dot represents the mean location of the thinnest point for the keratoconic population and the dotted red line represents one standard deviation. Reprinted with permission from SLACK Incorporated: Reinstein, DZ., Gobbe, M., Archer, T., Silverman, R., Coleman, J. (2010). “Epithelial, Stromal and Total Corneal Thickness in Keratoconus.” Journal or Refractive Surgery, 26, 259–271
Figure 13.4, Column 1 shows the keratometry , Atlas 995 (Carl Zeiss Meditec, Jena, Germany) corneal topography map and PathFinder™ corneal analysis, Orbscan II (software version 3.00) anterior elevation BFS, Orbscan II posterior elevation BFS and Artemis epithelial thickness profile of a normal eye.
Epithelial thickness can now also be measured using some optical coherence tomography systems, notably the RTVue (Optovue, Fremont, CA) [51–53]. These studies have confirmed this superior–inferior and nasal-temporal asymmetric profile for epithelial thickness in normal eyes [53].
This non-uniformity seems to provide evidence that the epithelial thickness is regulated by eyelid mechanics and blinking , as we suggested in 1994 [50]. We postulated that the eyelid might effectively be chafing the surface epithelium during blinking and that the posterior surface of the semi-rigid tarsus provides a template for the outer shape of the epithelial surface. During blinking, which occurs on average between 300 and 1500 times per hour [54], the vertical traverse of the upper lid is much greater than that of the lower lid. Doane [55] studied the dynamics of eyelid anatomy during blinking and found that during a blink the descent of the upper eyelid reaches its maximum speed at about the time it crosses the visual axis. As a consequence, it is likely that the eyelid applies more force on the superior than inferior cornea. Similarly, the friction on the cornea during lid closure is likely to be greater temporally than nasally as the outer can thus is higher than the inner can thus (mean intercanthal angle = 3°), and the temporal portion of the lid is higher than the nasal lid (mean upper lid angle = 2.7°) [56]. Therefore, it seems that the nature of the eyelid completely explains the non-uniform epithelial thickness profile of a normal eye.
Further evidence for this theory is provided by the epithelial thickness changes observed in orthokeratology [57]. In orthokeratology , a shaped contact lens is placed on the cornea overnight that sits tightly on the cornea centrally but leaves a gap in the mid-periphery. Therefore, the natural template provided by the posterior surface of the semi-rigid tarsus of the eyelid is replaced by an artificial contact lens template designed to fit tightly to the centre of the cornea and loosely paracentrally. We found significant epithelial thickness changes with central thinning and mid-peripheral thickening showing that the epithelium had remodelled according to the template provided by the contact lens, i.e. the epithelium is chafed and squashed by the lens centrally while the epithelium is free to thicken paracentrally where the lens is not so tightly fitted.
13.4 Epithelial Thickness Profile in Keratoconic Eyes
It is well known that the epithelial thickness changes in keratoconus since extreme steepening leads to epithelial breakdown, as often seen clinically. Epithelial thinning over the cone has been demonstrated using histopathologic analysis of keratoconic corneas by Scroggs et al. [58] and later using custom software and a Humphrey-Zeiss OCT system (Humphrey Systems, Dublin, CA) by Haque et al. [59].
We have characterized the in vivo epithelial thickness profile in a population of keratoconic eyes. The subjects included for the study had previously been diagnosed with keratoconus, and the diagnosis was confirmed by clinical signs of keratoconus such as microscopic signs at the slit-lamp, corneal topographic changes, high refractive astigmatism, reduced best-corrected visual acuity and contrast sensitivity, and significant level of higher order aberrations, in particular vertical coma. We measured the epithelial thickness profile across the central 10 mm diameter of the cornea for 54 keratoconic eyes of 30 patients and averaged the data in the population [60]. Epithelial thickness values for left eyes were reflected in the vertical axis and superimposed onto the right eye values so that nasal/temporal characteristics could be combined.
The average epithelial thickness profile in keratoconus revealed that the epithelium was significantly more irregular in thickness compared to normals. The epithelium was thinnest at the apex of the cone and this thin epithelial zone was surrounded by an annulus of thickened epithelium (Fig. 13.1b). While all eyes exhibited the same epithelial doughnut pattern, characterized by a localized central zone of thinning surrounded by an annulus of thick epithelium, the thickness values of the thinnest point and the thickest point as well as the difference in thickness between the thinnest and thickest epithelium varied greatly between eyes. There was a statistically significant correlation between the thinnest epithelium and the steepest keratometry (D), indicating that as the cornea became steeper, the epithelial thickness minimum became thinner. In addition, there was a statistically significant correlation between the thickness of the thinnest epithelium and the difference in thickness between the thinnest and thickest epithelium. This indicated that as the epithelium thinned, there was an increase in the irregularity of the epithelial thickness profile, i.e. that there was an increase in the severity of the keratoconus. The location of the thinnest epithelium within the central 5 mm of the cornea was displaced on average 0.48 mm (±0.66 mm) temporally and 0.32 mm (±0.67 mm) inferiorly with reference to the corneal vertex (Fig. 13.2). The mean epithelial thickness for all eyes was 45.7 ± 5.9 μm (range: 33.1–56.3 μm) at the corneal vertex, 38.2 ± 5.8 μm (range: 29.6–52.4 μm) at the thinnest point and 66.8 ± 7.2 μm (range: 54.1–94.4 μm) at the thickest point [60].
Figure 13.3b shows a B-scan for a keratoconic cornea which demonstrates the lack of homogeneity in epithelial thickness as well as central corneal thinning. There is epithelial thinning over the cone and relative epithelial thickening adjacent to the stromal surface cone.
Fig. 13.3
(a) (left) Horizontal non-geometrically corrected B-scan of a normal cornea obtained using the Artemis very high-frequency digital ultrasound arc scanner. The epithelium appears uniform in thickness across the 10 mm diameter of the scan. (b) (right) Vertical non-geometrically corrected B-scan of a keratoconic cornea obtained using the Artemis very high-frequency digital ultrasound arc-scanner. The epithelium appears very thin centrally coincident with a visible cone on the back surface. The epithelium is clearly thicker either side of the cone. The central epithelium is much thinner and the peripheral epithelium is much thicker compared to that seen in the normal eye
Figure 13.4, Column 2 shows the keratometry , Atlas 995 corneal topography map and PathFinder™ corneal analysis, Orbscan II anterior elevation BFS, Orbscan II posterior elevation BFS and Artemis epithelial thickness profile of a keratoconic eye. As expected, the front surface topography shows infero-temporal steepening with steep average keratometry and high astigmatism; the anterior and posterior elevation BFS maps demonstrate that the apex of the cone is located infero-temporally; the epithelial thickness profile shows epithelial thinning at the apex of the cone surrounded by an annulus of thicker epithelium. The steepest cornea coincides with the apex of the anterior and posterior elevation BFS as well as with the location of the thinnest epithelium.
Fig. 13.4
Central keratometry , Atlas corneal topography and PathFinder™ corneal analysis, Orbscan anterior and posterior elevation BFS and Artemis epithelial thickness profile for one normal eye, one keratoconic eye, and three example eyes where the diagnosis of keratoconus might be misleading from topography. The final diagnosis based on the epithelial thickness profile is shown at the bottom of each example. Reprinted with permission from SLACK Incorporated: Reinstein, DZ., Gobbe, M., Archer, T., Silverman, R., Coleman, J. (2010). “Epithelial, Stromal and Total Corneal Thickness in Keratoconus.” Journal or Refractive Surgery, 26, 259–271
As for normal eyes, the epithelial thickness profile for keratoconus as described here has been confirmed by studies using OCT [53, 61–63]. The study by Laroche’s group [63] elegantly described the different stages of advanced keratoconus demonstrating that as keratoconus moves into its latter stages, a very different epithelial thickness profile becomes apparent. In advanced keratoconus , there is stromal loss often in the location of the cone, for example due to hydrops. This means that rather than the cone being elevated relative to the rest of the stroma, this region is now a depression. Therefore, the epithelium changes from being thinnest over the cone to being thickest in this region, as it is compensating for a depression instead of an elevation (see next section). There can be significant stromal loss in such advanced keratoconus , so the epithelium can be as thick as 200 μm in some cases. Examples of this epithelial thickening were also reported by Rocha et al. [61] who concluded that focal central epithelial thinning was suggestive but not pathognomonic for keratoconus (i.e. the presence of an epithelial doughnut pattern did not prove beyond any doubt that an eye has keratoconus). However, as described by Laroche, these cases only appear in very advanced keratoconus, which means that they are of no interest with respect to keratoconus screening. Eyes with early keratoconus will never present with epithelial thickening in the location of the cone as by definition if there has been stromal loss, then the keratoconus must be more advanced and the cornea will be obviously abnormal.
13.5 Understanding the Predictable Behaviour of the Corneal Epithelium
Epithelial thickness changes in keratoconus provide another example of the very predictable mechanism of the corneal epithelium to compensate for irregularities on the stromal surface. Epithelial thickness changes have also been described after myopic excimer laser ablation [64–67], hyperopic excimer laser ablation [68], radial keratotomy [69], intra-corneal ring segments [70], irregularly irregular astigmatism after corneal refractive surgery [45, 71–75] and in ectasia [76].
In all of these cases, the epithelial thickness changes are clearly a compensatory response to the change to the stromal surface and can all be explained by the theory of eyelid template regulation of epithelial thickness [46]. Compensatory epithelial thickness changes can be summarized by the following rules:
- 1.
- 2.
- 3.
- 4.
The amount of epithelial remodelling is defined by the rate of change of curvature of an irregularity [46, 77]; there will be more epithelial remodelling for a more localized irregularity [45, 72, 73, 75]. The epithelium effectively acts as a low pass filter, smoothing local changes (high curvature gradient) almost completely, but only partially smoothing global changes (low curvature gradient). For example, there is almost twice as much epithelial thickening after a hyperopic ablation [68] compared with a myopic ablation [64, 65, 67], and there is almost total epithelial compensation for small, very localized stromal loss such as after a corneal ulcer [68].
13.6 Diagnosing Early Keratoconus Using Epithelial Thickness Profiles
We have shown that mapping of the epithelial thickness profile reveals a very distinct thickness profile in keratoconus compared to that of normal corneas, due to the compensatory mechanism of the epithelium for stromal irregularities. We have also shown that the epithelial thickness profile changes with the progression of the disease; as the keratoconus becomes more severe, the epithelium at the apex of the cone becomes thinner and the surrounding annulus of epithelium in the epithelial doughnut pattern becomes thicker. Therefore, the degree of epithelial abnormality in both directions (thinner and thicker than normal) can be used to confirm or exclude a diagnosis of keratoconus in eyes suggestive but not conclusive of a diagnosis of keratoconus on topography at a very early stage in the expression of the disease [78].
13.6.1 Pattern of Epithelial Thickness Profile
The epithelial thickness profile in normal eyes demonstrates that the epithelium is on average thicker inferiorly than superiorly and slightly thicker nasally than temporally. There is very little variation in epithelial thickness within both the inferior hemi-cornea and the superior hemi-cornea. In contrast, in keratoconic eyes, the average epithelial thickness map showed an epithelial doughnut pattern characterized by a localized central zone of thinning overlying the stromal cone, surrounded by an annulus of thick epithelium. In early keratoconus, we would expect to see the pattern of localized epithelial thinning surrounded by an annulus of thick epithelium coincident with a suspected cone on posterior elevation BFS. The coincidence of epithelial thinning together with an eccentric posterior elevation BFS apex may reveal whether or not to ascribe significance to an eccentric posterior elevation BFS apex occurring concurrently with a normal front surface topography. In other words, in the presence of normal front surface topography, thinning of the epithelium coincident with the location of the posterior elevation BFS apex would represent total masking or compensation for a sub-surface stromal cone and herald posterior elevation BFS changes which do represent keratoconus. Conversely, finding thicker epithelium over an area of topographic steepening or an eccentric posterior elevation BFS apex would imply that the steepening is not due to a keratoconic sub-surface stromal cone, but more likely due to localized epithelial thickening. Localized compensatory changes in epithelial thickness profiles can be detected by Artemis VHF digital ultrasound once they exceed 1–2 μm. In a way, examination of epithelial thickness profile irregularities provides a very sensitive method of examining stromal surface topography—by proxy. Therefore, this technique provides increased sensitivity and specificity to a diagnosis of keratoconus well in advance of any detectable corneal front surface topographic change .
Case Examples
Figure 13.4 shows three selected examples where epithelial thickness profiles helped to interpret and diagnose anterior and posterior elevation BFS abnormalities. In each case, the epithelial thickness profile appears to be able to differentiate cases where the diagnosis of keratoconus is uncertain, from normal [78].
Case 1 (OS) represents a 25-year-old male, with a manifest refraction of −1.00 −0.50 × 150 and a best spectacle-corrected visual acuity of 20/16. Atlas corneal topography demonstrated inferior steepening which would traditionally indicate keratoconus. The keratometry was 45.25/43.25 D × 76, and PathFinder™ corneal analysis classified the topography as normal. Orbscan II posterior elevation BFS showed that the posterior elevation BFS apex was decentred infero-temporally. Corneal pachymetry minimum by handheld ultrasound was 479 μm. Contrast sensitivity was slightly below the normal range measured using the CSV-1000 (Vector Vision Inc., Greenville, Ohio). There was −0.30 μm (OSA notation) of vertical coma on WASCA aberrometry. Corneal hysteresis was 7.5 mmHg and corneal resistance factor was 7.1 mmHg, which are low, but these could be affected by the low corneal thickness. The combination of inferior steepening, an eccentric posterior elevation BFS apex and thin cornea raised the suspicion of keratoconus although there was no suggestion of keratoconus by refraction, keratometry or PathFinder™ corneal analysis. Artemis epithelial thickness profile showed a pattern typical of keratoconus with an epithelial doughnut shape characterized by a localized zone of epithelial thinning displaced infero-temporally over the eccentric posterior elevation BFS apex, surrounded by an annulus of thick epithelium. The coincidence of an area of epithelial thinning with the apex of the posterior elevation BFS, as well as the increased irregularity of the epithelium confirmed the diagnosis of early keratoconus.
Case 2 (OD) represents a 31-year-old female, with a manifest refraction of −2.25 −0.50 × 88 and a best spectacle-corrected visual acuity of 20/16. Atlas corneal topography demonstrated a very similar pattern to case 1 of inferior steepening, therefore suggesting that the eye could also be keratoconic. The keratometry was 44.12/44.75 D × 148, and PathFinder™ corneal analysis classified the topography as suspect subclinical keratoconus. Orbscan II posterior elevation BFS showed that the apex was slightly decentred nasally. Corneal pachymetry minimum by handheld ultrasound was 538 μm. Contrast sensitivity was in the normal range. There was 0.32 μm (OSA notation) of vertical coma on WASCA aberrometry. Corneal hysteresis was 10.1 mmHg and corneal resistance factor was 9.8 mmHg, which are well within normal range. The combination of inferior steepening, against-the-rule astigmatism and high degree of vertical coma raised the suspicion of keratoconus, which was also noted by PathFinder™ corneal analysis. Artemis epithelial thickness profile showed a typical normal pattern with thicker epithelium inferiorly and thinner epithelium superiorly. Thicker epithelium inferiorly over the suspected cone (inferior steepening on topography) was inconsistent with an underlying stromal surface cone, and therefore the diagnosis of keratoconus was excluded. This patient would have been rejected for surgery given a documented PathFinder™ corneal analysis warning of suspect subclinical keratoconus, but given the epithelial thickness profile, this patient was deemed a suitable candidate for LASIK.
The anterior corneal topography in case 3 (OD) bears no features related to keratoconus . The patient is a 35-year-old female with a manifest refraction of −25 −0.50 × 4 and a best spectacle-corrected visual acuity of 20/16. The refraction had been stable for at least 10 years and the contrast sensitivity was within normal limits. The keratometry was 43.62/42.62 D × 74 and PathFinder™ analysis classified the topography as normal. Orbscan II posterior elevation BFS showed that the apex was slightly decentred infero-temporally, but the anterior elevation BFS apex was well centred. Corneal pachymetry minimum by handheld ultrasound was 484 μm. Pentacam (Oculus, Wetzlar, Germany) keratoconus screening indices were normal. WASCA ocular higher order aberrations were low (RMS = 0.19 μm) as well as the level of vertical coma (coma = 0.066 μm). Corneal hysteresis was 8.9 mmHg and corneal resistance factor was 8.8 mmHg, both within normal limits. In this case, only the slightly eccentric posterior elevation BFS apex and the low–normal corneal thickness were suspicious for keratoconus, while all other screening methods gave no indication of keratoconus. However, the epithelial thickness profile showed an epithelial doughnut pattern characterized by localized epithelial thinning surrounded by an annulus of thick epithelium, coincident with the eccentric posterior elevation BFS apex. Epithelial thinning with surrounding annular thickening over the eccentric posterior elevation BFS apex indicated the presence of probable sub-surface keratoconus. In this case, it seems that the epithelium had fully compensated for the stromal surface irregularity so that the anterior surface topography of the cornea appeared perfectly regular. Given the regularity of the front surface topography and the normality of nearly all other screening parameters, it is feasible that this patient could have been deemed suitable for corneal refractive surgery and subsequently developed ectasia. As we were able to also consider the epithelial thickness profile, this patient was rejected for corneal refractive surgery. This kind of case may explain some reported cases of ectasia “without a cause” [79].
13.7 Automated Algorithm for Classification by Epithelium
Based on this qualitative diagnostic method, we then set out to derive an automated classifier to detect keratoconus using epithelial thickness data, together with Ron Silverman and his group at Columbia University [80]. We used stepwise linear discriminant analysis (LDA) and neural network (NN) analysis to develop multivariate models based on combinations of 161 features comparing a population of 130 normal and 74 keratoconic eyes. This process resulted in a six-variable model that provided an area under the receiver operating curve of 100 %, indicative of complete separation of keratoconic from normal corneas. Test-set performance averaged over ten trials, gave a specificity of 99.5 ± 1.5 % and sensitivity of 98.9 ± 1.9 %. Maps of the average epithelium and LDA function values were also found to be well correlated with keratoconus severity grade (see Figs. 13.5 and 13.6). Other groups have also been working on automated classification algorithms based on epithelial thickness data obtained by OCT [53, 81].
Fig. 13.5
Epithelial thickness maps averaged over all normal corneas and for each keratoconus grade. The departure from the normal epithelial distribution is evident even in grade 1 keratoconus but becomes more obvious with severity. Reprinted with permission from IOVS: Silverman RH, Urs R, Roychoudhury A, Archer TJ, Gobbe M, Reinstein DZ. Epithelial remodeling as basis for machine-based identification of keratoconus. Invest Ophthalmol Vis Sci. 2014 Mar 13;55(3):1580–7
Fig. 13.6
Box and whisker plot of discriminant function value versus keratoconus severity grade. Grade 0 represents normal subjects. Grades 1–4 are based on Krumeich classification. Boxes represent ±1 quartile about median value (horizontal line), and whiskers represent full range of values for each group. Circles indicate outliers. Reprinted with permission from IOVS: Silverman RH, Urs R, Roychoudhury A, Archer TJ, Gobbe M, Reinstein DZ. Epithelial remodeling as basis for machine-based identification of keratoconus. Invest Ophthalmol Vis Sci. 2014 Mar 13;55(3):1580–7
Following this study, we then applied the algorithm to a population of 10 patients with unilateral keratoconus (clinically and algorithmically topographically normal in the fellow eyes), on the basis that the fellow eye in such patients represents a latent form of keratoconus, and as such, has been considered a gold standard for studies aimed at early keratoconus detection. These eyes were also analysed using the Belin-Ambrosio enhanced ectasia display (BAD-D parameter and ART-Max) [20, 24, 82] and the Orbscan SCORE value as described by Saad and Gatinel [28–30].
Table 1 summarizes the diagnosis derived for the fellow eyes using the classification function based on epithelial thickness parameters, the classification function combining VHF digital ultrasound (epithelial and stromal thickness) and Pentacam HD parameters, the BAD-D and ART-Max values, and the Orbscan SCORE value. The last column of the table indicates whether the topographic map displayed suspicious features of keratoconus such as inferior steepening and asymmetric bow-tie. The table also shows the percentage of eyes that were classified as keratoconus by each method.
The most interesting finding of this study was that more than 50 % of the fellow eyes were classified as normal by all methods. This was similar to the result reported by Bae et al. [26], who found no difference in the BAD-D or ART-Max values between normal and topographically normal fellow eyes of keratoconus patients. This is in contrast to other studies using unilateral keratoconus populations where a much higher sensitivity was reported; however, these studies often included patients with a suspicious topography in the fellow eye (i.e. some studies use a more rigorous definition of unilateral keratoconus than others) [27]. Therefore, the main conclusion from the study was to put into question the validity of using unilateral keratoconus patients for keratoconus screening studies. The fact that a number of these fellow eyes showed absolutely no indication of keratoconus by any method implies that it is likely that these were truly normal eyes. However, it is generally agreed that keratoconus as a disease must be bilateral [83], therefore it appears that these cases are patients who do not have keratoconus, but have induced an ectasia in one eye, for example by eye rubbing or trauma. This means that using “unilateral keratoconus” populations to study keratoconus screening may be flawed.
The alternative is somewhat more alarming, as this would mean that there are eyes with keratoconus that are literally undetectable by any existing method. This would, however, explain any case of “ectasia without a cause” [79, 84]. Detection of keratoconus in such cases may require development of new in vivo measurements of corneal biomechanics, although this appears to be outside the scope of current methods such as the Ocular Response Analyzer [85–87] and Corvis (Oculus, Wetzlar, Germany) [86, 87] due to the wide scatter in the data acquired. Another factor, as has been described using Brillouin microscopy [88], may be that the biomechanical tensile strength of the cornea may not be different from normal in early keratoconus when measuring the whole cornea globally, but there may only be a difference in the region localized of the cone (or in the location of a future cone). Another potential and final solution would be whether a genotype or other molecular marker for keratoconus could be found [89–91].
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
