Diagnostic Accuracy of Corneal and Epithelial Thickness Map Parameters to Detect Keratoconus and Suspect Keratoconus

Aim To establish the diagnostic accuracy of corneal and epithelial thickness measurements obtained by spectral-domain optical coherence tomography (SD-OCT) in detecting keratoconus (KC) and suspect keratoconus (SKC). Methods This retrospective study reviewed the data of 144 eyes separated into three groups by the Sirius automated corneal classification software: normal (N) (n = 65), SKC (n = 43), and KC (n = 36). Corneal thickness (CT) and epithelial thickness (ET) in the central (0–2 mm) and paracentral (2–5 mm) zones were obtained with the Cirrus high-definition OCT. Areas under the curve (AUC) of receiver operator characteristic (ROC) curves were compared across groups to estimate their discrimination capacity. Results ROC curve analysis revealed excellent predictive ability for ET variables: minimum (Min) ET (0_2), minimum-maximum (Min-Max) ET (0_2), superonasal-inferotemporal (SN-IT) ET (2_5), Min-Max ET (2_5), and Min ET (2_5) to detect keratoconus (AUC > 0.9, all). Min-Max CT (0_2) was the only CT parameter with excellent ability to discriminate between KC and N eyes (AUC = 0.94; cutoff = ≤−32 μm). However, both ET and CT variables were not strong enough (AUC < 0.8, all) to differentiate between SKC and N eyes, with the highest diagnostic power for Min-Max ET (2_5) (AUC = 0.71; cutoff = ≤−9 μm) and central corneal thickness (CCT) (AUC = 0.76; cutoff = ≤533 μm). Conclusion These results demonstrate that OCT-derived CT and ET are able to differentiate between KC and N eyes, with a high level of certainty. However, Min-Max ET (2_5) was the parameter with the highest ability to detect suspect keratoconus.


Introduction
Keratoconus (KC) is a progressive corneal disease characterized by thinning of the central or paracentral portion of the cornea resulting in irregular astigmatism and subsequent visual impairment [1].Several studies from the Middle East have demonstrated a high prevalence rate of KC [2][3][4].KC is typically characterized by thinning of the stroma with the largest thickness changes in the 5 mm zone [5].
Currently, multiple diagnostic modalities are available to measure total corneal thickness and corneal epithelial thickness (ET), including very high-frequency (VHF) digital ultrasound, confocal microscopy, and anterior segment optical coherence tomography (AS-OCT) [6][7][8].AS-OCT is a noncontact, high-resolution imaging technique that allows precise delineation of corneal surfaces and layers.
Among the various parameters proposed by Randelman et al. to predict the risk of postlaser in situ keratomileusis (LASIK) ectasia, unrecognized early-stage keratoconus is considered a vital risk factor of this disease [9].Several corneal topographic-based parameters have demonstrated good accuracy for detecting manifest KC [10][11][12].However, their accuracy in detecting suspect keratoconus (SKC) is limited [13].In early KC, corneal epithelial remodelling can mask the stromal irregularities, resulting in false-negative topographic fndings [14].However, this compensatory epithelial thickness modulation can be an important diagnostic parameter to detect early or subclinical KC.
Our previous studies have demonstrated the diagnostic value of topographic, tomographic, and higher-order aberration parameters utilizing the Sirius (Costruzione Strumenti Oftalmici, Florence, Italy) corneal tomography in detecting SKC and KC [13,15].Evaluating the changes that occur at the level of corneal epithelium may provide additional information for the detection of SKC and prevent postoperative LASIK ectasia.Te aim of this study was to investigate the diagnostic ability of CT and ET parameters obtained with the Zeiss Cirrus 5000 HD-OCT (Carl Zeiss Meditec, Dublin, CA, USA) in diferentiating between SKC, KC, and normal (N) corneas.

Materials and Methods
2.1.Study Participants.Tis was a retrospective, casecontrolled study.Te study adhered to the tenets of the Declaration of Helsinki, and Institutional Review Board approval was obtained (approval number TUH-06023).All participants were at least 18 years old and provided written informed consent from the Department of Ophthalmology of Tishreen University Hospital, Latakia, Syria.
Artifcial intelligence was used to classify the participants into three groups (N, SKC, and KC) utilizing the Sirius Corneal Navigator automated corneal classifcation software.Te Sirius classifcation algorithm (Phoenix) had the highest accuracy in detecting KC (91.24%) and SKC (88.68%) [16].
Te data from a single eye was randomly included if both eyes were positive for KC or SKC.Only data from one eye was included from N patients.Patients were recruited consecutively, and all subjects meeting the inclusion criteria were included in the study.
Patients who had undergone LASIK with 3 years of follow-up and no documented evidence of ectasia or postoperative complications were included in the N group.Eyes in the N group had a normal Sirius software classifcation, and baseline preoperative corneal topography data did not reveal fndings suggestive of KC, such as focal or inferior steepening of the cornea or central keratometry greater than 47.0 D in either eye.All eyes in the N group had a corrected distance Snellen (feet) visual acuity of 20/20 or better.
Diagnosis of SKC and KC was confrmed by an experienced cornea specialist (AS) with more than 10 years of experience.Te SKC group was characterized by a positive Sirius software indicator, absence of clinical (keratometric, retinoscopic, or biomicroscopic) signs of KC in either eye, and corrected distance visual acuity (CDVA) of 20/20 or better.
Te diagnosis of KC was made if there was (a) an irregular cornea determined by distorted keratometry mires and/or distortion of the retinoscopic refex; or one of the following slit-lamp fndings: Vogt striae, 2 mm arc of Fleisher ring, apical thinning, Munson's sign, Rizutti's sign, or corneal scarring consistent with KC, in addition to (b) a positive Sirius software indicator.
Patients were excluded if they had corneal hydrops, previous corneal or ocular surgery (such as cataract, glaucoma, corneal collagen cross-linking, excimer laser surgery, intrastromal corneal rings, and phakic intraocular lens), a history of corneal oedema, any pathology of the cornea and anterior segment that may confound with corneal parameters (dry eye disease, keratitis, glaucoma, uveitis, and Fuch's dystrophy), autoimmune disease, breastfeeding, or pregnancy.Contact lens wearers in the preceding 3 months were also excluded.
In line with previously reported study [17,18], CT and ET acquisition was automated by the anterior segment spectral-domain Cirrus 5000 HD-OCT (Carl Zeiss Meditec, Germany).Te cornea lens attachment was used for the pachymetry map and HD cornea scans (Figure 1).Te subjects were positioned on the headrest with gaze fxed towards the fxation light target, and the image was centred on the pupil centre.Two scans were obtained for each eye by a single operator with 60-second rest periods to optimize the tear flm rest and average values were registered.CT and ET mappings were obtained at a certain time, from 10 : 00 to 14 : 00, and at least 2 hours after awakening.
Te pachymetry map scans include 8 radial scans (1024 axial scans each) repeated 5 times covering a 9 mm diameter area.Te software algorithm measures ET as the distance between the middle of the frst (tear flm) and second (anterior surface of the Bowman layer) hyperrefective lines on the B-scan (Figure 2).CT was measured as the distance between the air-tear and cornea-aqueous interfaces.Te pachymetry analysis tool provided automated cornea thickness measurement in seventeen sectors.Images were captured after the horizontal single scan line was placed on the corneal apex, where the hyperrefective corneal refex was visible.Repeat scans were taken if the initial scan was decentred or had a poor corneal apex refection.

Mean CT and ET Outcome Measures. CT and ET
variables were registered at 0-2 mm central and 2-5 mm paracentral zones.

Corneal Tickness Variables.
Corneal thickness variables were defned as follows: Min CT (0_2): the minimum corneal thickness at the 0-2 mm central zone; Avg CT (0_2): the average corneal thickness at the 0-2 mm central zone; Max CT (0_2): the maximum corneal thickness at the 0-2 mm central zone; Min-Max CT (0_2): the minimum corneal thickness minus the maximum corneal thickness at the 0-2 mm corneal zone; Min CT (2_5): the minimum corneal thickness at the 2-5 mm paracentral zone; Avg CT (2_5): the average corneal thickness at the 2-5 paracentral zone; Max CT (2_5): the maximum corneal thickness at the 2-5 mm paracentral zone; Min-Max CT (2_5): the minimum corneal thickness minus the maximum corneal thickness at the 2-5 mm paracentral zone; S-I CT (2_5): the mean corneal thickness of the superior octant minus that of the inferior octant at the 2-5 mm paracentral zone; SN-IT CT (2_5): the diference between mean superonasal and mean inferotemporal corneal thickness at the 2-5 mm paracentral zone; CCT: the central corneal thickness.Receiver operating characteristic (ROC) curves were used to distinguish KC and SKC from N corneas.Tese curves were obtained by plotting sensitivity against 1-specifcity, calculated for each value observed.Te area under the ROC curve (AUC) measures discrimination, which is the ability of the test to accurately classify eyes with and without disease.

Results
One hundred and forty-four eyes were included and separated into three groups: N, SKC, and KC.Demographic characteristics for the respective groups are presented in Table 1.Te N group included 65 eyes (63.07%female).Te mean age was 27.07 ± 8.43 years.Te mean sphere was −0.47 ± 2.70 dioptres (D), and the mean cylinder was −0.59 ± 1.61 D. Te SKC group included 43 eyes (41.86% female).Te mean patient age was 27.7 ± 6.82 years.Te mean sphere was −0.11 ± 1.15 D, and the mean cylinder was −1.14 ± 1.26 D. Te KC group included 36 eyes (47.22% female).Te mean patient age was 26.49 ± 6.12 years.Te mean sphere was −0.76 ± 1.88 D, and the mean cylinder was −2.72 ± 1.90 D. Tere was no statistically signifcant difference between the mean age of the three groups (ANOVA � 0.79).Te mean sphere values were not signifcantly diferent between the three groups (ANOV-A � 0.43).Te mean cylinder values were highest in the KC group with statistically signifcant diferences between the three groups (ANOVA < 0.0001).
Te mean and standard deviation of OCT-based corneal thickness parameters for all groups are shown in  respectively) in the KC group.Te diference between the minimum and maximum corneal values at the central 2 mm zones was highest (−55.49± 20.94 μm) in the KC group.All corneal thickness parameters obtained from the 2 mm central cornea were statistically signifcant diferent between the three groups (ANOVA < 0.05, for all).When comparing each of the two groups, all of the CT parameters reached statistical signifcance (P < 0.05), except Max CT (0_2) which did not reveal any statistical signifcance between the KC and SKC groups (P � 0.227).
Te minimum corneal thickness value at the 2-5 mm paracentral zone was lowest in the KC group (447.62 ± 49.38 μm).Te associations between opposite octants revealed that the highest diferences between superior and inferior octants and between IT and SN octants were seen in the KC group (28.75 ± 24.05 μm and 46.25 ± 24.72 μm, respectively).While the diference between the minimum and maximum CT was almost similar in the N and SKC, the highest diference (−117.12± 44.84 μm) was registered in the KC group.When comparing the three groups together, all CT parameters at the 2 mm paracentral cornea showed statistical signifcance (ANOVA < 0.05, for all).However, average and maximum CT values were not statistically signifcant between the KC and SKC groups (P > 0.05, for both).S-I CT (2_5) only showed a statistically signifcant diference between the KC and N groups (P � 0.004).SN-IT CT (2_5) was statistically signifcance between the KC and N groups and between the KC and N groups, but not between the KC and SKC groups (P � 0.168).
While the Min-Max CT (2_5) revealed statistical signifcance between the KC and N groups and between KC and SKC groups (P 0.0001, for both), the diference was not statistically signifcant between SKC and N groups (P � 0.0503) ET parameter data are presented in Table 3. Te minimum ETat the 2 mm central zone was thinnest (37.26 ± 4.51) μm in the KC group.Te highest diference (−13.62 ± 6.58 μm) between the minimum and maximum ET at the central zone was registered in the KC group.All ET parameters obtained from the central cornea revealed statistically signifcant diferences between the three groups (ANOVA < 0.0001), except Max ET (0_2), which did not show any statistical signifcance (ANOVA � 0.881).However, Min ET (0_2) and Min-Max ET (0_2) were not statistically diferent between the SKC and N groups (P > 0.05, for both).Te lowest ET value and the highest maximum ET value at the paracentral cornea were reported in the KC group (35.6 ± 4.48 μm and 57.26 ± 8.77 μm, respectively).In the same trend, the diference between minimum ET and maximum ET was highest (−21.59 ± 7.79 μm) in the KC group.While the superior ET was thinner than the inferior ET in the N group, the superior ET was thicker than the inferior ET in the KC group.Te highest diference between SN ET and IT ET (9.35 ± 4.7 μm) was seen in the KC group.All ET parameters at the paracentral cornea were statistically signifcant between the three groups (ANOVA > 0.05), except Avg ET (2_5) which did not show any statistically signifcant diference (ANOVA � 0.183).CET was lowest in the KC (42.97 ± 5.38 μm) with statistical signifcance between the three groups (ANOVA < 0.0001).SN-IT ET (2_5) and Min-Max (2_5) were the only ET parameters with statistically signifcant diferences between SKC and N groups (P � 0.046 and P � 0.0004, respectively).Figure 2 shows the epithelial and topography maps of normal eye, suspect keratoconus, and keratoconus eyes.
Sensitivity, specifcity, and area under the curve values identifed by cutof points of diferent CT parameter sets to diferentiate eyes with SKC from N corneas and KC from N ones are presented in Table 4.To distinguish between SKC and controls, the highest AUC (0.76) was seen for CCT, with a sensitivity of 86.05% and a specifcity of 55.7%.However, none of the CT parameters were strong enough to identify suspect keratoconus (AUC < 0.8, for all).While Min CT (0_2), Avg CT (0_2), Max CT (0_2), Min CT (2_5), Min-Max CT (2_5) SN-IT CT (2_5), and CCT were strong enough to distinguish between KC and N groups (AUC > 0.8, for all), the highest strength (0.94) was seen for Min CT (0_2).Tis parameter with a cutof value of ≤−32 showed the highest sensitivity and specifcity in detecting KC, at 94.12% and 93.4%, respectively.
Table 5 demonstrates the sensitivity, specifcity, and area under the curve values identifed by cutof points of diferent ET parameter sets to diferentiate eyes with SKC from N Corneas and KC from N ones.Min-Max ET (2_5) was the only ET parameter strong enough (AUC > 0.7) to diferentiate between SKC and N eyes.Tis parameter with a cutof value of ≤−9 had a sensitivity of 67.44% and a specifcity of 76.7%.
Figures 3 and 4 show the ROC curves of the corneal thickness parameters to diferentiate suspect KC from N eyes and KC from N eyes, respectively.While there was no CT parameter strong enough to distinguish SKC from N eyes (AUC < 0.8, for all), the same was not true for KC (AUC < 0.8, all).Min-Max CT (0_2) was the parameter with the best AUC to distinguish KC from N eyes, 0.939 (95% CI, 0.871-0.978).Figures 5 and 6 show the ROC curves of the diferent ET parameters to diferentiate suspect KC from N eyes and KC from N eyes.While Min-Max ET (2_5) was the best ET parameter to diferentiate between SKC and N eyes (AUC � 0.707; 95% CI, 0.609-0.793),SN-IT ET (2_5) was

Discussion
Te aim of this study was to determine the diagnostic accuracy of CT and ET parameters measured by the anterior segment spectral-domain OCT (Zeiss Cirrus 5000 HD) in diferentiating SKC, KC, from N corneas.
Te results demonstrate the strong statistical signifcance of CT variables between KC and N eyes.To a lesser extent, SKC and N eyes comparisons showed signifcant diferences for all CT variables except Min CT (2_5), S-I CT (2_5), and Min-Max CT (2_5).In their study, Hashemi et al. used highresolution spectral-domain OCT (Heidelberg Engineering, Heidelberg, Germany) to investigate the diagnostic ability of total corneal thickness in detecting KC, where they found that CCT and Min-Median variables were the most sensitive indices for the diagnosis of KC [5].In the current study, the two CT variables showing the greatest AUC to detect KC were Min-Max CT (0_2) and Min-Max CT (2_5); AUC � 0.94 and AUC � 0.89, respectively.Ambrosio et al. hypothesized that relative corneal thickness as opposed to unique thickness points is a more useful predictor in detecting KC and SKC [19].Although stromal thickness progression variables showed the highest ability to detect KC, this was not the case for SKC eyes, as the highest AUC was seen for the thickness point variable (CCT; AUC � 0.76).Our fndings are consistent with a previously reported study, where the authors found that CCT was the parameter with the highest ability to diferentiate between subclinical KC and N eyes (AUC � 0.782) [18].Reinstein et al. established that stromal thickness progression at a 2 mm radius from the thinnest point in N eyes was 29.9 ± 5.4, whereas it was 60.6 ± 25.6 μm in KC eyes [14].Te data presented here show similar fndings in the diference between the thinnest and thickest cornea at the 2 mm central zone at 27.92 ± 44.61 μm in N eyes and 55.94 ± 20.94 μm in KC eyes.
Coupling ET profles with corneal tomography may further aid in screening for KC and may be useful in the clinical setting.Epithelial thickness data may allow for an earlier diagnosis of KC, as epithelial changes may precede any changes produced on the front surface of the cornea.Such ET changes in KC have been examined by other groups [20][21][22][23][24][25].Fuente et al. evaluated total corneal thickness and corneal layers in 86 healthy young adults using SD-OCT, where they reported a central ET of 54.60 ± 4.25 μm [26].Erie et al. reported a central ET of 46.0 ± 5.0 μm in N corneas measured by confocal microscopy [27].Teir measurements, which excluded the precorneal tear flm thickness, were thinner than the central ET presented in this report of 49.92 ± 3.98 μm in N eyes.Of note, the measurements in this study included the thickness of the tear flm, estimated at 3 μm, which is likely to account for the diference.Li et al. used a Fourier-domain OCT system (RTVue, Optovue Inc., Fremont, CA, USA) to measure corneal epithelium thickness over a 6 mm diameter.Te central epithelial thickness of N eyes from their study, which included the tear flm, was 52.3 ± 3.6 μm.Moreover, the corneal epithelial thickness was demonstrated to be thicker inferiorly by −1.6 μm in the N corneas in comparison with the superior cornea [21].Te data fndings presented here concur with those of Li and colleagues as demonstrated by an S-I mean diference of −1.5 μm.
Previous studies have shown that the apex of early KC focal steepening is compensated by corneal epithelial thinning, and the degree of potential epithelial remodelling is dependent on the severity of the keratoconus [20].In this study, all ET variables showed statistically signifcant differences between KC and N eyes.Te epithelial doughnut pattern described by Reinstein et     zone where the cone is more pronounced.In their study investigating epithelial thickness distribution characteristics in keratoconic eyes, Kanellopoulos and Asimellis found that the epithelium was thinner at the inferotemporal area in comparison with the superonasal area and thinner inferiorly [23].In agreement with these results, we found that the epithelium in the keratoconic eyes was thinner inferiorly than superiorly by 6.62 ± 9.16 μm and thinner inferotemporally than superonasally by 9.35 ± 4.7 μm.Te opposite held true in the N group, the epithelium was thinner superiorly than inferiorly and thinner superonasally than inferotemporally.Tis phenomenon can be justifed by the mechanical chafng and abrasion caused by the accelerated movement of the upper tarsus with a larger force being applied on the superior epithelium [28].

Journal of Ophthalmology
In their study, Li et al. found that root-mean-square pattern deviation (RMSPD) was the only epithelial thickness-based variable with excellent diagnostic power to diferentiate between KC and N eyes, while the other variables varied from poor (central, superior, and inferior zonal epithelial thicknesses), to fair (S-I), to good (Min ET and Min-Max ET) [21].In contrast, several corneal epithelial thickness-based variables evaluated in this study showed excellent (Min ET (0_2), Min-Max ET (0_2), SN-IT ET (2_5), Min-Max ET (2_5), and Min ET (2_5) minimum, AROC > 0.9, for all) diagnostic power in diferentiating KC from N eyes.Te diference in the diagnostic power of ET between the two studies may be due to the diferences in ET variables obtained in each study, although both studies obtained their ET variables from the 5 mm central cornea.For example, the SN-IT ET (2_5) variable was not included in the Li et al. study, despite it proving to be one of the more favourable variables in discriminating between the KC group and the N group (AUC � 96; sensitivity � 88.24%; specifcity � 96.7%).
Ostadian et al. found that the central epithelium was signifcantly thinner in subclinical KC corneas than in N corneas [24].In this study, the central ET was not signifcantly diferent between SKC and N eyes.However, SN-IT ET (2_5) and Min-Max (2_5), were the only ET variables with statistical signifcance between SKC and N eyes, with the highest ability (AUC � 0.71) for Min-Max ET (2_5) to diferentiate between SKC and N eyes.Journal of Ophthalmology Our fndings demonstrated that ET variables were superior to CT variables in diferentiating between KC and N eyes, since only one CT variable (Min-Max (0_2)) revealed excellent diagnostic power (AUC > 0.9) to detect KC eyes, while fve ET variables (Min ET (0_2), Min-Max ET (0_2), SN-IT ET (2_5), Min-Max ET (2_5), and Min ET (2_5)) were excellent in diagnosing KC eyes.
Diagnosing early-stage KC remains a challenge.Our results showed that both CT and ET variables derived from OCT imaging were not sufcient to diferentiate between SKC and N eyes (AUC < 0.8, for both).Estrada and Alio used Placido/Sheimpfug-based corneal tomography (Sirius) to investigate the diagnostic power of posterior corneal surface parameters in diferentiating KC from N eyes, where they found that root-mean-square per unit of area (RMS/A) and keratoconus vertex back (KVb) were the variables with the highest discriminating capabilities between normal and mild KC cases with an area under the curve of 0.96 and 0.97, respectively [29].In our previously reported study, symmetry index back (SIb) derived from the posterior corneal surface was the parameter with the highest diagnostic power to detect SKC eyes [13].Consistent with these results, Heidari et al. found that Sirius SIb was the parameter with the highest power to distinguish subclinical KC from N eyes (AUC � 0.908) [30].However, this was not unexpected, since changes in the posterior corneal surface may be one of the frst clinically detectable signs of KC.In contrast, when we investigated the application of anterior and posterior corneal higher-order aberrations (HOAs) in detecting KC and SKC, coma (3, ±1) derived from the anterior corneal surface was the parameter with the highest ability to discriminate between SKC and N eyes (AUC � 0.922; cutof > 0.2) [15].While Estrada and Alio found that posterior corneal HOAs were useful to diferentiate normal from mild KC cases, our results demonstrated that posterior corneal HOA parameters were unsatisfactory in discriminating between SKC and normal eyes [29].
One limitation of this study is that pachymetric and epithelial thickness maps were confned to the central 5 mm diameter of the cornea.Te cone apex is located in the central 5 mm diameter of the cornea in the vast majority of KC eyes [19].Tus, we opted to evaluate the characteristics of these variables in the 2 mm central and 2-5 mm  paracentral corneas.OCT measurements in our study included the thickness of the precorneal tear flm.Tear flm thickness values range between 3 and 5 μm [27,28].OCT epithelial and corneal measurements including the tear flm thickness may lead to potential inaccuracies of the absolute values and this could be considered a limitation of the study.Te small sample size is another limitation in this study, as using small sample sizes may yield unreliable results.Increasing the sample size and using a power calculation in future studies will be key to limit the risk of sampling bias.
In conclusion, the cutof points proposed in our study had sufcient sensitivity and specifcity to diferentiate between KC and N eyes.In SKC, however, the use of a single parameter to diagnose KC is insufcient.Te combination of OCT-derived data with posterior corneal surface and wavefront data to improve automatic recognition of early KC will be the target of our future research.Nevertheless, epithelial and corneal thickness mapping is a vital tool for the refractive surgeon, with particular strengths in the early detection of keratoconus and monitoring of ectasia due to its predictable transformation in pathological and ectatic conditions as highlighted in this study.

2. 1 . 3 .
Epithelial Tickness Variables.Epithelial thickness variables were defned as follows: Min ET (0_2): the minimum epithelial thickness at the 0-2 mm central zone; Avg ET (0_2): the average epithelial thickness at the 0-2 mm central zone; Max ET (0_2): the maximum epithelial thickness at the 0-2 mm central zone; Min-Max ET (0_2): the minimum epithelial thickness minus the maximum corneal thickness at the 0-2 mm corneal thickness; Min ET (2_5): the minimum epithelial thickness at the 2-5 mm paracentral zone; Avg ET (2_5): the average epithelial corneal thickness at the 2-5 paracentral zone; Max ET (2_5): the maximum epithelial corneal thickness at the 2-5 mm paracentral zone; Min-Max ET (2_5): the minimum epithelial corneal thickness minus the maximum thickness at the 2-5 mm paracentral zone; S-I ET (2_5): the average epithelial thickness of the superior octant minus that of the inferior octant at the 2-5 mm paracentral zone; SN-IT ET (2_5): the diference between mean superonasal and mean inferotemporal corneal epithelial thickness at the 2-5 mm paracentral zone; CET: the thickness of the epithelium at the central point.

Figure 4 :
Figure 4: Comparison of the corneal thickness parameters that showed the best area under the receiver operating characteristic curves to diferentiate between keratoconus and normal eyes.Min-Max � minimum-maximum; CT �corneal thickness; Min � minimum; CCT �central corneal thickness; AUC � area under the receiver operating characteristic curve; SE � standard error; CI � confdence interval.

Figure 6 :
Figure 6: Comparison of the epithelial thickness parameters that showed the best area under the receiver operating characteristic curves to diferentiate between keratoconus and normal eyes.SN-IT �superonasal-inferotemporal; ET �epithelial thickness; Min-� minimum; Min-Max � minimum-maximum; AUC � area under the receiver operating characteristic curve; SE � standard error; CI � confdence interval.

Table 1 :
Demographic characteristics of the study groups.

Table 2 :
OCT-based corneal thickness data in normal, suspect keratoconus, and keratoconus groups.
al. consists of a localized zone of thinning surrounded by an annulus of thickened epithelium.Tis pattern was characteristic of all keratoconic eyes.Reinstein et al. also demonstrated that the mean thinnest epithelium in KC eyes was 38.2 ± 5.8, whereas the mean thickest epithelium was 66.8 ± 7.2 μm with a mean diference between the thinnest and thickest epithelium of 28.6 ± 10.8 μm.± 4.48 μm, 57.26 ± 8.77 μm, and 21.59 ± 7.79 μm, respectively.Tese values were registered at the paracentral

Table 4 :
Sensitivity, specifcity, and area under the curve values identifed by cutof points of diferent OCT-based corneal thickness parameter sets to diferentiate eyes with suspect keratoconus from normal corneas and keratoconus from normal ones.

Table 5 :
Sensitivity, specifcity, and area under the curve values identifed by cutof points of diferent epithelial thickness parameter sets to diferentiate eyes with suspect keratoconus from normal corneas and keratoconus from normal ones.
Figure 5: Comparison of the epithelial thickness parameters that showed the best area under the receiver operating characteristic curves to diferentiate between suspect keratoconus and normal eyes.Min-Max � minimum-maximum; ET �epithelial thickness; SN-IT �superonasal-inferotemporal; Min � minimum; AUC � area under the receiver operating characteristic curve; SE � standard error; CI � confdence interval.