Keratoconus is an ocular disease characterized by thinning and outward protrusion of the cornea.1 The irregular corneal shape causes distorted vision that may continue to deteriorate as the disease progresses. Corneal cross-linking (CXL) is an effective treatment that prevents keratoconus progression by strengthening stromal collagen.2 Early detection of keratoconus is important so that treatment can be instituted prior to vision loss. Early detection is also important in screening candidates for refractive surgery, because undiagnosed keratoconus is the leading risk factor for ectasia after refractive surgery.3,4 Suspicious topographic patterns have been detected in 4.3% of patients evaluated as candidates for laser in situ keratomileusis (LASIK),5 although early forms of the disease may be more common and go unrecognized during routine examinations.6,7
Keratoconus-related decreases in visual acuity and slit-lamp findings typically do not present until the manifest stage of the disease. For early detection to be possible, clinicians must possess more sensitive diagnostic tools. Corneal topography is the current clinical standard for keratoconus detection; however, its sensitivity and specificity are still limited. One reason is that thinning of the epithelium in the region of the cone can mask focal anterior steepening of the underlying stroma.8 Furthermore, it can be difficult to distinguish keratoconus from other sources of focal anterior steepening. Epithelial changes (contact lens–related warpage and epithelial basement membrane dystrophy) and stromal changes (steep central island due to uneven ablation and decentered hyperopic LASIK) can mimic the topographic appearance of ectasia.9,10
Attempts to overcome the limitations of measuring only the anterior corneal surface have been made by measuring corneal thickness (ie, pachymetry). Central pachymetry has low sensitivity and specificity because a keratoconic cornea can have normal central thickness and a relatively thin cornea can be normal.11 Other pachymetry-based parameters such as minimum corneal thickness, superonasal-inferotemporal asymmetry, and the Belin/Ambrósio Enhanced Ectasia Display have been shown to provide better diagnostic value.12–15 It has therefore been suggested that tomography (eg, optical coherence tomography [OCT] or Scheimpflug imaging) is the best method for diagnosing early keratoconus11 because the posterior surface and thickness of the cornea can be measured. In addition to pachymetry, OCT can map epithelial thickness,16 which exhibits characteristic changes in keratoconus.17
The premise of our study was the clinical observation that pachymetric and epithelial thinning occur at similar locations in keratoconus.10 Our goal was to develop a coincident thinning (CTN) index to quantify this co-localized thinning and evaluate its ability to detect keratoconus. Eyes with varying severities of keratoconus and normal controls were imaged, and the CTN index was compared to minimum pachymetry and maximum keratometry.
Patients and Methods
Participants were recruited from the Casey Eye Institute at Oregon Health & Science University (Portland, Oregon) and the Affiliated Eye Hospital of Wenzhou Medical College (Wenzhou, China). The study protocol was approved by the institutional review boards of both institutions, and written informed consent was obtained from all patients. This study adhered to the tenets of the Declaration of Helsinki and the Health Insurance Portability and Accountability Act of 1996. Patients were evaluated for refractive error and underwent slit-lamp examination. Corneal topography maps and maximum keratometry were obtained using either the Orbscan II (Bausch & Lomb) or Pentacam (Oculus Optikgeräte GmbH). The ABCD keratoconus grading indices were recorded for eyes scanned by the Pentacam.18 Eyes with recent contact lens use (soft within 1 week or rigid gas permeable within 3 weeks) or history of ocular surgeries were excluded. Eyes with severe keratoconus with corneal scarring were excluded because they have unpredictable thickness patterns and do not pose a challenge for clinical diagnosis.
Participants were divided into normal and three keratoconus groups using the following criteria: (1) normal: normal slit-lamp examination, corrected distance visual acuity (CDVA) of 20/20 or better, normal topography appearance, and KISA% index less than 100 [The KISA% index combines multiple topography parameters to quantify features associated with keratoconus19]; (2) manifest keratoconus: slit-lamp examination finding consistent with keratoconus (Vogt's striae, Fleischer's ring, Munson's sign, Rizutti's sign, and apparent focal bulging and thinning) or CDVA worse than 20/20, topographic pattern characteristic of keratoconus (asymmetric bowtie with skewed radial axis, central or inferior steep zone), and KISA% index of 100 or greater; (3) subclinical keratoconus: normal slit-lamp examination, CDVA 20/20 or better, topography characteristic of keratoconus, and KISA% index of 100 or greater; and (4) forme fruste keratoconus: better eye of a patient with asymmetric keratoconus who had a normal slit-lamp examination, CDVA 20/20 or better, normal topography, and KISA% index of less than 100; the fellow eye had either manifest or subclinical keratoconus.
OCT Imaging and Pattern Deviation Maps
OCT scans were obtained using commercial Fourier-domain OCT systems (RTVue or Avanti; Optovue, Inc) with a corneal adaptor module for imaging of the anterior eye. A Pachymetry+Cpwr radial scan pattern was used to acquire 6-mm–wide images of the cornea, centered on the pupil (Figure AA, available in the online version of this article). The scan pattern was composed of eight meridians, each with 1,024 axial lines. The pattern was repeated five times within a single scan.
(A) Optical coherence tomography (OCT) radial scan pattern used to obtain 6-mm–wide images of the cornea. (B) Representative OCT image of a cornea with segmented boundaries.
The corneal OCT images were segmented and dewarped (Figure AB)20 using a custom algorithm (MATLAB; MathWorks, Inc) to generate pachymetry and epithelial thickness maps.12,16 Minimum pachymetry was recorded. OCT scans with low signal intensity or motion artifact were excluded to avoid unreliable thickness maps. Eyes with two satisfactory scans were analyzed.
The methodology for calculating pattern deviation (PD) maps was described previously.21 Briefly, PD maps display percent differences in relative thickness compared to the thickness pattern of normal eyes. Pachymetry and epithelial thickness PD maps were used to calculate the CTN index. All calculations were limited to the central 5 mm of the maps to avoid less reliable measurements at the periphery.
Keratoconic Cone Search Area and Gaussian Fitting
Pachymetric and epithelial maximum relative thinning locations were identified in the corresponding PD maps. The correlation and distance between the pachymetric and epithelial maximum relative thinning locations were evaluated for the normal and keratoconus groups. The 99% confidence region for the pachymetric maximum relative thinning locations was calculated from the eyes with manifest keratoconus (Figure B, available in the online version of this article) to characterize the keratoconic cone distribution and serve as the cone search area during Gaussian fitting.
Corneal sector map showing the locations of maximum relative thinning from pachymetry pattern deviation maps of eyes with manifest keratoconus. The search area was defined as the 99% confidence region based on a two-dimensional Gaussian distribution. Locations for left eyes were mirror-image mapped to the right eye coordinate system. CTN = coincident thinning; S = superior; T = temporal; I = inferior; N = nasal
Two-dimensional symmetric Gaussian waveforms were fitted to the pachymetric and epithelial PD maps to capture the cone-shaped focal thinning pattern in keratoconus. The center of the Gaussian fit was the location of maximum relative thinning within the cone search area described above. Full width at half maximum (FWHM) values from the best-fit pachymetric and epithelial Gaussian waveforms for the eyes with manifest keratoconus were used to characterize the keratoconic cone width. Using the median FWHM, a simplified Gaussian fitting approach was used to calculate the CTN index (details in Results section). Both pachymetric and epithelial PD maps were refit using the simplified approach. The correlation between the fitted Gaussian waveform and the thickness PD map, as well as the Gaussian amplitudes, were evaluated.
A CTN map was generated as the product of Gaussian fits of the pachymetric and epithelial PD maps (Figure C, available in the online version of this article). The CTN index was calculated as the maximum value of the CTN map (Equation 1):
are the Gaussian fits from the pachymetry and epithelial thickness PD maps, respectively. CTN index values from repeated scans were averaged to produce one value for each eye.
Gaussian fitting procedure for a representative pachymetry pattern deviation map from an eye with manifest keratoconus. (A) The location of maximum relative thinning within the search area was identified, and (B) a 4-mm diameter surrounding region was used for Gaussian fitting.
The CTN index cutoff value, used to classify eyes as “normal” or “keratoconus,” was calculated using the receiver operating characteristic (ROC) curve. The eyes from all three keratoconus groups were treated as the positive cases, and one eye was randomly selected if both eyes of a patient were measured. The ROC curve was spline-interpolated, and the CTN index value corresponding to 95% specificity was selected as the cutoff. Bootstrapping was used to reduce the influence of sample selection, and the final CTN index cutoff value was the bootstrapped average. The Gaussian fits, CTN index, and CTN index cutoff value were calculated using custom MATLAB algorithms.
Statistical Analysis and Classification Performance
The chi-square test was used to verify data normality. Two-tailed t tests (normality confirmed) or Mann-Whitney U tests (normality rejected) were used to compare differences in average values between each keratoconus group and normal eyes. One eye was chosen randomly if both eyes of a participant were involved. The relationship between the pachymetric and epithelial maximum relative thinning locations was quantified using either Pearson (normality confirmed) or Spearman (normality rejected) correlations. The repeatability of the CTN index, minimum pachymetry, and maximum keratometry were evaluated using the intraclass correlation coefficient (ICC). Only one topography scan per eye was obtained in this study. Therefore, the ICC for maximum keratometry was estimated using the standard deviation of measurement error from the literature.22
Repeated five-fold cross-validation (80% training, 20% test, five repetitions) was performed in MATLAB to estimate the out-of-sample performance of the CTN index, minimum pachymetry, and maximum keratometry. During cross-validation of the CTN index, a new search area and characteristic Gaussian width were calculated within each fold using the training set. Average values across the repeated cross-validations were calculated for the classification cutoff, overall accuracy, sensitivity for each keratoconus group, and area under the ROC curve (AUC).
The McNemar test was used to compare differences in sensitivity between the three parameters. A majority vote method was used to produce one set of predictions (eg, an eye found to be normal in three of five cross-validations was classified as normal). Statistical differences between ROC curves were analyzed using the DeLong test. The CTN index values computed when the eyes were in the test set were averaged across the cross-validations to obtain a single data set. These statistical comparisons were done using MedCalc (MedCalc Software).
A total of 82 normal and 133 keratoconic eyes were included in this prospective cross-sectional study (Table 1). There was no significant difference in age between groups. Compared to normal, all three keratoconus groups had thinner corneas, larger steep keratometry, and larger maximum keratometry. Flat keratometry was higher for the manifest and subclinical keratoconus groups, but not the forme fruste keratoconus group. The average ABCD indices of 64 manifest, 10 subclinical, and 17 forme fruste keratoconus eyes whose topography maps were acquired using Pentacam were reported in Figure 1.
Average ABCD indices for the eyes within each keratoconus group that were scanned by the Pentacam (Oculus Optikgeräte GmbH) (manifest keratoconus = 64 eyes, subclinical keratoconus = 10 eyes, forme fruste keratoconus = 17 eyes). The D index was equal to zero for all eyes with subclinical and forme fruste keratoconus. Error bars = standard deviation.
The locations of pachymetric focal thinning were mostly clustered in the inferotemporal and inferior sectors of the cornea for eyes with manifest keratoconus (Figure B). The 99% confidence region (ie, cone search area) was centered inferotemporally (x = −0.55 mm, y = −1.35 mm) with a radius of 1.91 mm. The pachymetric and epithelial maximum relative thinning locations were significantly closer for the manifest and subclinical keratoconus groups (average distance: 0.49 to 0.63 mm) compared to normal (1.47 mm, Table 2). A strong correlation was found between these locations for the manifest and subclinical keratoconus groups (0.63 to 0.86, Table 2). These findings verified coincident thinning of pachymetry and epithelium thickness in keratoconus. Therefore, the simplified Gaussian fitting approach used the pachymetric maximum relative thinning location inside the cone search area as the Gaussian center for both pachymetric and epithelial map fitting.
Relationship Between Pachymetry and Epithelial Thickness Maximum Relative Thinning Locations
Gaussian fits with flexible width yielded a median FWHM of 3.38 mm for pachymetry and 7.73 mm for epithelial thickness in manifest keratoconus. Due to large variability in the epithelial fits, the simplified Gaussian fitting approach used a fixed symmetric Gaussian width. The Gaussian width was set equal to the median FWHM for the respective map type. The fitting was performed on a 4-mm diameter region around the Gaussian center.
The pachymetric thinning patterns were strongly correlated with the Gaussian functions in keratoconus (average correlation = 0.71 to 0.92, Table 3, Figure 2). The epithelial thinning patterns correlated well with the Gaussian functions in manifest and subclinical keratoconus (0.70 to 0.73), but were less correlated in forme fruste keratoconus (0.49). Normal pachymetric and epithelial thickness PD maps correlated poorly with the Gaussian functions (0.34 to 0.42), indicating the lack of cone-shaped focal thinning patterns in normal eyes.
Characteristics of Gaussian Fits From Pachymetry and Epithelial Thickness PD Maps and the CTN Index
Pattern deviation, coincident thinning (CTN), and Pentacam (Oculus Optikgeräte GmbH) axial power maps for representative eyes from each group. Eyes in the keratoconus groups were generally found to have a higher amplitude and correlation from Gaussian fitting compared to normal eyes (dashed line = Gaussian fit region). Gpachy = pachymetry Gaussian fit; Gepi = epithelial thickness Gaussian fit; D = diopters; K = keratometry
For both pachymetry and epithelial thickness, the amplitude of the Gaussian fit was significantly larger for all three keratoconus groups compared to normal (Table 3). The larger amplitudes were compounded to produce a higher CTN index in keratoconus. The CTN index revealed the highest amount of relative thinning in eyes with manifest keratoconus (average: 11.9%), followed by subclinical keratoconus (9.3%) and forme fruste keratoconus (3.4%), compared to normal.
The ICC across all keratoconus groups was 0.95 for the CTN index and 0.97 for minimum pachymetry. The estimated ICC for maximum keratometry was 0.98. The CTN index was strongly correlated (Figure 3) with both minimum pachymetry (R = −0.80, P < .0001) and maximum keratometry (R = 0.81, P < .0001).
Coincident thinning (CTN) index plotted against (A) minimum pachymetry and (B) maximum keratometry. The 95% specificity cutoff lines are shown for each parameter. The data points in the shaded region represent eyes that were classified as keratoconic eyes by the CTN index but appear normal based on minimum pachymetry or maximum keratometry. D = diopters
The keratoconus cutoff values were 2.22% for the CTN index, 489.3 µm for minimum pachymetry, and 46.90 D for maximum keratometry. Cross-validation showed that the CTN index had statistically higher overall accuracy compared to minimum pachymetry nd maximum keratometry (P < .04, Table 4). The AUC was significantly higher for the CTN index versus both minimum pachymetry and maximum keratometry (P < .05). Small variations were observed for the Gaussian center search area (center x: ±0.03 mm, center y: ± 0.02 mm, radius: ± 0.07 mm) and the characteristic Gaussian FWHM (± 0.1 mm) during cross-validation.
Classification Accuracy of Keratoconus Diagnostic Parameters
In this study, a CTN index was developed to analyze co-localized thinning in pachymetry and epithelial thickness maps. The performance of the CTN index in differentiating normal and keratoconic eyes was evaluated. We verified the premises of the CTN index, showing that focal thinning for pachymetry and epithelial thickness are coincident in keratoconus, the thinning has a characteristic inferotemporal location and Gaussian shape, and the amplitude of thinning correlates well with maximum keratometry.
Corneal epithelial thickness may exhibit thinning at multiple locations, whereas pachymetry usually shows one clear thinning location. Therefore, the search area was defined using pachymetric PD maps. The search area for the location of maximum relative thinning (ie, Gaussian center) was centered inferotemporally to cover the characteristic cone locations in keratoconus.23 The cone location appeared to have a symmetric Gaussian distribution, which allowed us to use a simple circular search area. Most of this search area fell within the 6-mm diameter OCT scan. Current OCT systems offer 9-mm diameter scans that would encompass the entire search area and may further improve the diagnostic accuracy of the CTN index.
The pachymetric width of the keratoconus cone we measured was similar to previous measurements from anterior mean curvature maps (FWHM approximately 3 mm).24 The fixed width of the Gaussian function made it possible to compare the shape of the thinning pattern between groups. The Gaussian function showed excellent cross-correlation with PD maps of keratoconic eyes, validating our choice of using symmetric Gaussian functions to characterize focal thinning in keratoconus.
The calculation of the CTN index used a multiplication scheme that was influenced by both the amplitudes and locations of relative thinning in the pachymetry and epithelial thickness PD maps. The larger CTN index measured in keratoconic eyes can be explained by their larger Gaussian amplitudes (Table 3) and the closer proximity of the maximum relative thinning locations for the two maps (Table 2).
The CTN index outperformed both minimum pachymetry and maximum keratometry in distinguishing keratoconic eyes from normal eyes. The superior performance of the CTN index is likely due to the integration of epithelial thickness information. Because the epithelium is thin relative to overall corneal thickness, decreases in epithelial thickness result in larger percent changes, which could heavily influence the CTN index. Furthermore, whereas thinning of the epithelium over the cone can make it difficult to recognize keratoconus using anterior topography, the CTN index algorithm takes advantage of this characteristic pattern to enhance its diagnostic performance.
Other methods have been proposed for improving keratoconus diagnosis. Detection of early keratoconus using the elevation of the posterior corneal surface has been effective when combined with other measurements such as pachymetry.15,25–27 Other studies have reported high AUC values when distinguishing eyes with forme fruste keratoconus from normal eyes using models that combined up to 13 parameters from tomographic imaging.28,29 Importantly, these models need to be validated to determine out-of-sample performance. Machine learning methods have reported sensitivity and specificity greater than 90% on validation data sets.30–32 Changes in corneal biomechanics are suspected to precede structural changes and have drawn interest.33–36 Direct comparison to the aforementioned methods is difficult due to differences in patient populations and the criteria defining early keratoconus. In this study, we demonstrated that the CTN index was capable of identifying 56% of the forme fruste keratoconus eyes, including 4 eyes that showed no signs of keratoconus based on ABCD indexing (all indices = 0, an example is shown in Figure 2, bottom row). One eye with forme fruste keratoconus with a non-zero ABCD index (ABCD = 0010) was not detected by the CTN index.
Maximum keratometry is used as a metric for keratoconus progression. However, variation between repeated measurements can be as large as 1.00 D,22 so it is not reproducible enough to serve as an endpoint for individual patients.37–39 The CTN index showed excellent repeatability with an ICC of 0.95. The CTN index was also highly correlated with minimum pachymetry and maximum keratometry, suggesting that it could be used to characterize keratoconus severity. Thus, the CTN index may improve the ability of clinicians to monitor progression in early to moderate keratoconus.
There are several limitations of this study. First, the sample size was relatively small for the subclinical keratoconus and forme fruste keratoconus groups. This may explain why a statistically significant difference in sensitivity was not detected between the diagnostic parameters for the early keratoconus groups (except for CTN index versus maximum keratometry for the forme fruste keratoconus group, which had a large difference). Second, longitudinal measurements acquired at multiple time points are needed to verify the ability of the CTN index to detect keratoconus progression. Finally, OCT is capable of measuring both anterior and posterior corneal topography,40 and corneal topographic information can be incorporated into future models to further improve diagnostic accuracy.
We have developed an OCT-based algorithm for the detection of keratoconus that quantifies coincident thinning in pachymetry and epithelial thickness maps. A novel parameter, the CTN index, was designed as a quantitative image-derived biomarker that is specific to keratoconus. The high sensitivity of the CTN index provides evidence that it could be used to complement current clinical standards for detecting and monitoring early to moderate keratoconus.
- Rabinowitz YS. Keratoconus. Surv Ophthalmol. 1998;42(4):297–319. doi:10.1016/S0039-6257(97)00119-7 [CrossRef]
- Chunyu T, Xiujun P, Zhengjun F, Xia Z, Feihu Z. Corneal collagen cross-linking in keratoconus: a systematic review and meta-analysis. Sci Rep. 2014;4(1):5652–5652. doi:10.1038/srep05652 [CrossRef]
- Binder PS, Lindstrom RL, Stulting RD, et al. Keratoconus and corneal ectasia after LASIK. J Cataract Refract Surg. 2005;31(11):2035–2038. doi:10.1016/j.jcrs.2005.12.002 [CrossRef]
- Randleman JB, Woodward M, Lynn MJ, Stulting RD. Risk assessment for ectasia after corneal refractive surgery. Ophthalmology. 2008;115(1):37–50. doi:10.1016/j.ophtha.2007.03.073 [CrossRef]
- Torricelli AAM, Bechara SJ, Wilson SE. Screening of refractive surgery candidates for LASIK and PRK. Cornea. 2014;33(10):1051–1055. doi:10.1097/ICO.0000000000000171 [CrossRef]
- Klein SR, Epstein RJ, Randleman JB, Stulting RD. Corneal ectasia after laser in situ keratomileusis in patients without apparent preoperative risk factors. Cornea. 2006;25(4):388–403. doi:10.1097/01.ico.0000222479.68242.77 [CrossRef]
- Ambrósio R Jr, Dawson DG, Salomão M, Guerra FP, Caiado AL, Belin MW. Corneal ectasia after LASIK despite low preoperative risk: tomographic and biomechanical findings in the unoperated, stable, fellow eye. J Refract Surg. 2010;26(11):906–911. doi:10.3928/1081597X-20100428-02 [CrossRef]
- Kanellopoulos AJ, Asimellis G. Anterior segment optical coherence tomography: assisted topographic corneal epithelial thickness distribution imaging of a keratoconus patient. Case Rep Ophthalmol. 2013;4(1):74–78. doi:10.1159/000350630 [CrossRef]
- Schallhorn JM, Tang M, Li Y, Louie DJ, Chamberlain W, Huang D. Distinguishing between contact lens warpage and ectasia: usefulness of optical coherence tomography epithelial thickness mapping. J Cataract Refract Surg. 2017;43(1):60–66. doi:10.1016/j.jcrs.2016.10.019 [CrossRef]
- Tang M, Li Y, Chamberlain W, Louie DJ, Schallhorn JM, Huang D. Differentiating keratoconus and corneal warpage by analyzing focal change patterns in corneal topography, pachymetry, and epithelial thickness maps. Invest Ophthalmol Vis Sci. 2016;57(9):OCT544–OCT549. doi:10.1167/iovs.15-18938 [CrossRef]
- Gomes JAP, Tan D, Rapuano CJ, et al. Group of Panelists for the Global Delphi Panel of Keratoconus and Ectatic Diseases. Global consensus on keratoconus and ectatic diseases. Cornea. 2015;34(4):359–369. doi:10.1097/ICO.0000000000000408 [CrossRef]
- Li Y, Meisler DM, Tang M, et al. Keratoconus diagnosis with optical coherence tomography pachymetry mapping. Ophthalmology. 2008;115(12):2159–2166. doi:10.1016/j.ophtha.2008.08.004 [CrossRef]
- Qin B, Chen S, Brass R, et al. Keratoconus diagnosis with optical coherence tomography–based pachymetric scoring system. J Cataract Refract Surg. 2013;39(12):1864–1871. doi:10.1016/j.jcrs.2013.05.048 [CrossRef]
- Ambrósio R Jr, Caiado AL, Guerra FP, et al. Novel pachymetric parameters based on corneal tomography for diagnosing keratoconus. J Refract Surg. 2011;27(10):753–758. doi:10.3928/1081597X-20110721-01 [CrossRef]
- Belin MW, Khachikian SS. Keratoconus/ectasia detection with the Oculus Pentacam: Belin/Ambrósio enhanced ectasia display. Highl Ophthalmol. 2007;35(6):5–12.
- Kanellopoulos AJ, Asimellis G. OCT-derived comparison of corneal thickness distribution and asymmetry differences between normal and keratoconic eyes. Cornea. 2014;33(12):1274–1281. doi:10.1097/ICO.0000000000000275 [CrossRef]
- Kanellopoulos AJ, Asimellis G. OCT corneal epithelial topographic asymmetry as a sensitive diagnostic tool for early and advancing keratoconus. Clin Ophthalmol. 2014;8:2277–2287. doi:10.2147/OPTH.S67902 [CrossRef]
- Belin MW, Duncan JK. Keratoconus: the ABCD Grading System. Klin Monatsbl Augenheilkd. 2016;233(6):701–707. doi:10.1055/s-0042-100626 [CrossRef]
- Rabinowitz YS, Rasheed K. KISA% index: a quantitative videokeratography algorithm embodying minimal topographic criteria for diagnosing keratoconus. J Cataract Refract Surg. 1999;25(10):1327–1335. doi:10.1016/S0886-3350(99)00195-9 [CrossRef]
- Huang D, Duker J, Fujimoto J, Lumbrosco B, Schuman J, Weinreb R. Interpretation of corneal images. In: Huang D, Duker J, Fujimoto J, Lumbrosco B, Schuman J, Weinreb R. Imaging the Eye From Front to Back With RTVue Fourier-Domain Optical Coherence Tomography.SLACK Incorporated; 2010:39–46.
- Li Y, Chamberlain W, Tan O, Brass R, Weiss JL, Huang D. Subclinical keratoconus detection by pattern analysis of corneal and epithelial thickness maps with optical coherence tomography. J Cataract Refract Surg. 2016;42(2):284–295. doi:10.1016/j.jcrs.2015.09.021 [CrossRef]
- de Luis Eguileor B, Escudero Argaluza J, Pijoán Zubizarreta JI, Santamaria Carro A, Etxebarria Ecenarro J. Evaluation of the reliability and repeatability of Scheimpflug system measurement in keratoconus. Cornea. 2018;37(2):177–181. doi:10.1097/ICO.0000000000001373 [CrossRef]
- Kenney MC, Gaster RN. Keratoconus. In: Kenney MC, Gaster RN. Ocular Disease. W.B. Saunders; 2010:42–48. doi:10.1016/B978-0-7020-2983-7.00006-1 [CrossRef]
- Tang M, Shekhar R, Miranda D, Huang D. Characteristics of keratoconus and pellucid marginal degeneration in mean curvature maps. Am J Ophthalmol. 2005;140(6):993–1001. doi:10.1016/j.ajo.2005.06.026 [CrossRef]
- Saad A, Gatinel D. Topographic and tomographic properties of forme fruste keratoconus corneas. Invest Ophthalmol Vis Sci. 2010;51(11):5546–5555. doi:10.1167/iovs.10-5369 [CrossRef]
- de Sanctis U, Aragno V, Dalmasso P, Brusasco L, Grignolo F. Diagnosis of subclinical keratoconus using posterior elevation measured with 2 different methods. Cornea. 2013;32(7):911–915. doi:10.1097/ICO.0b013e3182854774 [CrossRef]
- Muftuoglu O, Ayar O, Ozulken K, Ozyol E, Akinci A. Posterior corneal elevation and back difference corneal elevation in diagnosing forme fruste keratoconus in the fellow eyes of unilateral keratoconus patients. J Cataract Refract Surg. 2013;39(9):1348–1357. doi:10.1016/j.jcrs.2013.03.023 [CrossRef]
- Hwang ES, Perez-Straziota CE, Kim SW, Santhiago MR, Randleman JB. Distinguishing highly asymmetric keratoconus eyes using combined Scheimpflug and spectral-domain OCT analysis. Ophthalmology. 2018;125(12):1862–1871. doi:10.1016/j.ophtha.2018.06.020 [CrossRef]
- Golan O, Piccinini AL, Hwang ES, et al. Distinguishing highly asymmetric keratoconus eyes using dual Scheimpflug/Placido analysis. Am J Ophthalmol. 2019;201:46–53. doi:10.1016/j.ajo.2019.01.023 [CrossRef]
- Arbelaez MC, Versaci F, Vestri G, Barboni P, Savini G. Use of a support vector machine for keratoconus and subclinical keratoconus detection by topographic and tomographic data. Ophthalmology. 2012;119(11):2231–2238. doi:10.1016/j.ophtha.2012.06.005 [CrossRef]
- Smadja D, Touboul D, Cohen A, et al. Detection of subclinical keratoconus using an automated decision tree classification. Am J Ophthalmol. 2013;156(2):237–246.e1. doi:10.1016/j.ajo.2013.03.034 [CrossRef]
- 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;55(3):1580–1587. doi:10.1167/iovs.13-12578 [CrossRef]
- Vinciguerra R, Ambrósio R Jr, Elsheikh A, et al. Detection of keratoconus with a new biomechanical index. J Refract Surg. 2016;32(12):803–810. doi:10.3928/1081597X-20160629-01 [CrossRef]
- Ambrósio R Jr, Lopes BT, Faria-Correia F, et al. Integration of Scheimpflug-based corneal tomography and biomechanical assessments for enhancing ectasia detection. J Refract Surg. 2017;33(7):434–443. doi:10.3928/1081597X-20170426-02 [CrossRef]
- Steinberg J, Siebert M, Katz T, et al. Tomographic and biomechanical Scheimpflug imaging for keratoconus characterization: a validation of current indices. J Refract Surg. 2018;34(12):840–847. doi:10.3928/1081597X-20181012-01 [CrossRef]
- Shao P, Eltony AM, Seiler TG, et al. Spatially-resolved Brillouin spectroscopy reveals biomechanical abnormalities in mild to advanced keratoconus in vivo. Sci Rep. 2019;9(1):7467. doi:10.1038/s41598-019-43811-5 [CrossRef]
- Szalai E, Berta A, Hassan Z, Módis L Jr, . Reliability and repeatability of swept-source Fourier-domain optical coherence tomography and Scheimpflug imaging in keratoconus. J Cataract Refract Surg. 2012;38(3):485–494. doi:10.1016/j.jcrs.2011.10.027 [CrossRef]
- Hashemi H, Yekta A, Khabazkhoob M. Effect of keratoconus grades on repeatability of keratometry readings: comparison of 5 devices. J Cataract Refract Surg. 2015;41(5):1065–1072. doi:10.1016/j.jcrs.2014.08.043 [CrossRef]
- Hashemi K, Guber I, Bergin C, Majo F. Reduced precision of the Pentacam HR in eyes with mild to moderate keratoconus. Ophthalmology. 2015;122(1):211–212. doi:10.1016/j.ophtha.2014.08.026 [CrossRef]
- Nakagawa T, Maeda N, Higashiura R, Hori Y, Inoue T, Nishida K. Corneal topographic analysis in patients with keratoconus using 3-dimensional anterior segment optical coherence tomography. J Cataract Refract Surg. 2011;37(10):1871–1878. doi:10.1016/j.jcrs.2011.05.027 [CrossRef]
|Patient Group||No. of Eyes (Patients)||Age (y)||Minimum Pachymetry (µm)||Flat K (D)||Steep K (D)||Max K (D)|
|Normal||82 (43)||32.5 ± 14.1||530 ± 30||42.6 ± 1.4||43.8 ± 1.4||44.3 ± 1.4|
|Manifest keratoconus||91 (65)||35.6 ± 15.2||439 ± 38a||46.1 ± 3.8a||50.1 ± 4.5a||57.6 ± 6.7a|
|Subclinical keratoconus||16 (14)||37.1 ± 14.3||455 ± 52a||45.0 ± 4.1b||48.1 ± 4.5b||53.1 ± 5.0a|
|Forme fruste keratoconus||26 (26)||31.1 ± 13.5||491 ± 35a||43.2 ± 1.3||44.5 ± 1.5a||46.1 ± 2.9b|
Relationship Between Pachymetry and Epithelial Thickness Maximum Relative Thinning Locations
|Patient Group||Xepi − Xpachy (mm)||Yepi − Ypachy (mm)||X Location Correlation||Y Location Correlation||Distance Between Pachymetric & Epithelial Maximum Relative Thinning Locations (mm)|
|Normal||0.19 ± 1.13||−0.55 ± 1.07||0.10||0.13||1.47 ± 0.76|
|Manifest keratoconus||−0.14 ± 0.48||0.37 ± 0.47a||0.63b||0.72b||0.63 ± 0.45a|
|Subclinical keratoconus||−0.03 ± 0.42||0.29 ± 0.25c||0.7b||0.86b||0.49 ± 0.27c|
|Forme fruste keratoconus||0.24 ± 1.04||0.16 ± 0.61c||0.35||0.45b||0.98 ± 0.72c|
Characteristics of Gaussian Fits From Pachymetry and Epithelial Thickness PD Maps and the CTN Index
|Patient Group||Gaussian Amplitude (%)||Gaussian Correlation||CTN Index (%)|
|Pachymetry||Epithelial Thickness||Pachymetry||Epithelial Thickness|
|Normal||−1 ± 1||−3 ± 1||0.42 ± 0.25||0.34 ± 0.17||1.0 ± 0.8|
|Manifest keratoconus||−10 ± 4a||−21 ± 8a||0.91 ± 0.07a||0.70 ± 0.22a||11.9 ± 3.8a|
|Subclinical keratoconus||−7 ± 3a||−18 ± 8a||0.92 ± 0.06b||0.73 ± 0.13a||9.3 ± 2.4a|
|Forme fruste keratoconus||−3 ± 2a||−7 ± 6a||0.71 ± 0.18b||0.49 ± 0.22a||3.4 ± 3.1a|
Classification Accuracy of Keratoconus Diagnostic Parametersa
|Diagnostic Parameter||Cutoff Value||Overall Accuracy (%)||Sensitivity by Group (%)||AUC|
|Manifest Keratoconus||Subclinical Keratoconus||Forme Fruste Keratoconus|
|CTN index (%)||2.22 ± 0.02||93 ± 0||100 ± 0||100 ± 0||56 ± 3||0.970 ± 0.002|
|Minimum pachymetry (µm)||489.3 ± 1.0||86 ± 0b||91 ± 0b||69 ± 2||49 ± 3||0.937 ± 0.003c|
|Maximum keratometry (D)||46.9 ± 0.4||8 ± 2b||98 ± 1||95 ± 3||28 ± 7b||0.945 ± 0.002c|