FP1575 : “Bowman’s Imaging- A novel imaging marker for detecting keratoconus”

AIOS – Optics / Refraction / Contact Lens Award

Dr.Kalyani Deshpande, D15134, Dr. Rohit Shetty, Dr.Rushad Shroff, Dr. Abhijit Sinha Roy

Introduction

Keratoconus (KC) is a degenerative corneal disease, accompanied by chronic steepening of the cornea, thinning of the epithelium and stroma.¹ Corneal topography and tomography play an important role in diagnosis of the disease.¹ However, assessment of the fellow eyes in unilateral KC subjects is a challenging task as the progression of the disease can last over a decade and serial follow-up of the patient is required to assess the true onset of KC.^2-6 These eyes may be classified as forme fruste keratoconus or FFKC.⁴ Several detection scores, based on corneal tomography, have been developed for diagnosis of FFKC eyes but these are almost entirely dependent on the measurement device, which may differ in terms of resolution and accuracy.^2-5

While tomography forms the crux of diagnosis of KC, morphological changes may also serve as sensitive indicators of presence of KC.⁷ The Bowman’s layer is thinner in KC eyes and this has been shown both in vivo and in vitro.^7,8 A recent study has speculated that this irregular thinning may be a marker for FFKC eyes supported by in vivo evidence from KC eyes.⁷ New indices, using high resolution optical coherence tomography (OCT), quantified the thinning of the Bowman’s layer in KC eyes, and showed high sensitivity and specificity.⁷ A major advantage of the proposed indices was that the results were reproducible with any OCT of resolution adequate to image the Bowman’s layer.⁷ In this study, a new index, called the Bowman’s roughness index (BRI), was developed to quantify the irregularity in the Bowman’s layer in KC and FFKC eyes using high resolution OCT. Sensitivity and specificity of BRI was also compared with the other tomography based indices, namely, corneal wavefront aberrations^2,4 and Belin-Ambrosio enhanced ectasia display overall deviation index (BAD-D)⁹.

Methods

The study was a prospective, observational, cross-sectional study approved by the ethics committee of Narayana Nethralaya Multi-specialty eye hospital, Bangalore, India. Informedconsent was obtained from all the subjects. The study was conducted in accordance with the tenets of the Declaration of Helsinki. 82 eyes of 61 subjects between 20 to 50 years of age were included in the study. Eyes were classified as normal, FFKC or KC based on corneal tomography (Pentacam, OCULUS Optikgerate Gmbh, Germany) and slit-lamp examination. Eyes (n=30) with normal tomography and no clinical signs of KC such as Fleischer ring,Vogt striae, scissoring of the red reflex, an abnormal retinoscopy, and curvature asymmetry leading to abnormalcorneal astigmatism were classified as normal. Eyes (n=31) were classified as KC based on evidence of corneal steepening, stromal thinning, asymmetric astigmatism, corneal scarring and other clinical indicators as described above. FFKC (n=21) eyes were the clinically normal fellow eyes of unilateral KC subjects. These eyes appeared similar to normal eyes upon clinical examination and may also be referred to as “subclinical” KC.⁴

The Bowman’s layer images were acquired with high resolution (870 nm) hand held spectral domain OCT (Envisu, Bioptigen Inc., Morrisville, USA). A scan of the central cornea, 3 mm in size, was acquired along the nasal-temporal direction such that the scan divided the cornea into two halves approximately upon visual examination. The acquired image had a lateral resolution of 3 μm and an axial resolution of 1.93 μm. The Bowman’s layer was segmented from the image using Graph theory and Dijkstra’s algorithm (Math Works Inc., USA).¹⁰ The irregularity was quantified using the method described. The segmented anterior edge of the Bowman’s layer is marked. The 3^rd order polynomial fit to the segmented edge is marked. The algebraic sum of the enclosed areas was named the Bowman’s roughness index (BRI). In a perfectly smooth layer, BRI would be 0.0 as there would be no shaded areas enclosed between the actual edge and its’ polynomial fit. A 3^rd order polynomial was chosen since a paraxial approximation (2^nd order polynomial) cannot capture mathematically the spatial variation of the 2^nd order gradient of corneal elevation. An order of the polynomial greater than 3 would capture the irregularity (pink line) better and may reduce the magnitude of BRI. However, the same order of the polynomial was used in normal, FFKC and KC eyes and therefore, the theoretical spatial resolution of the polynomial was the same in all the eyes.

Also the anterior edge of the epithelium and it’s polynomial fit of order 3 was obtained. The difference in elevation between the polynomial fit to the anterior edge of the epithelium and the anterior edge of the Bowman’s layer provided the epithelium thickness. From Pentacam, flat axis keratometry (K1), steep axis keratometry (K2), maximum keratometry (Kmax), mean keratometry (Kmean), central corneal thickness (CCT) and the BAD-D index were analyzed. Also from Pentacam, the lower (LOA) and higher order (HOA) anterior surface and total (anterior plus posterior surface) corneal wavefront aberrationswere analyzed using Zernike polynomials up to 8^th order and an analysis zone size of 6 mm. All variables were presented as mean ± standard error of the mean (SEM) after assessing normality of distribution. Group means of variables were assessed with analyses of variance. The sensitivity, specificity and accuracy of each variable were assessed with receiver operator characteristics (ROC). A p-value less than 0.05 was considered statistically significant and was Bonferroni corrected for multiple group comparisons. Statistical analyses were performed with MedCalc v15.8 (MedCalc Software, Belgium).

Results

Mean age of subjects in normal, FFKC and KC group was 26±1, 24±2 and 25±2 years (p=0.79), respectively . K1 (p=0.02), K2 (p<0.001), Kmean (p=0.001), Kmax (p<0.001) and CCT (p<0.001) of KC eyes differed significantly from normal and FFKC eyes. However, IOP was similar among all the groups (p=0.11). Mean BRI of normal was significantly higher than mean BRI of KC and FFKC eyes (p=0.001). Mean epithelium thickness of normal was significantly greater than mean epithelium thickness of KC eyes (p=0.001). BAD-D differed significantly between all the groups (p=0.001). Root mean square of anterior and total corneal lower and higher order corneal aberrations of normal and FFKC were significantly lower than KC eyes (p=0.001).

BAD-D index had overall the best area under the ROC curve, sensitivity and specificity between normal and KC eyes as well as between normal and FFKC eyes. BRI had overall the next best area under the ROC curve, sensitivity and specificity between normal and KC eyes as well as between normal and FFKC eyes. The RMS of anterior and total LOA and HOA had similar area under the ROC curve, sensitivity and specificity between normal and KC eyes as well as between normal and FFKC eyes. Interestingly, BRI had the highest sensitivity among all the indices as compared to the RMS of corneal wavefront aberrations and BAD-D index.

Discussion

Screening of FFKC is vital so that such corneas can be eliminated from potential risks of ectasia as a result of intervention.⁵ Morphological changes in KC eyes such as thinning and steepening have been characterized extensively in several studies but only a few of them have  assessed both FFKC and KC eyes in the same study.^2-5 A classification system using anterior corneal surface aberrations reported nearly 100% sensitivity but a widely varying specificity ranging from 68% to 97% in detection of FFKC eyes using OrbscanIIz topographer (Bausch & Lomb, New York, USA).²Another study using GALILEI Scheimpflug imaging system (ZiemerOphthalmic Systems AG, Port, Switzerland) reported a sensitivity and specificity of 92% and 97%, respectively in detection of FFKC with a classification system based on corneal tomographic and aberrometric parameters.³A third study using Pentacam reported a sensitivity and specificity of 89% and 82%, respectively, in the detection of FFKC eyes using the BAD-D index.⁹ Another study using OPD-scan (NidekCo. Ltd., Gamagori, Japan) reported a sensitivity and specificity ranging from 71% to 90% and from 73% to 95%, respectively, in the detection of FFKC with corneal aberrations.¹¹ Thus, there is considerable variability among studies, which may be attributed to differences between the patient populations in different studies and also differences between the measurement techniques (Placido disk, Scheimpflug).

Another confounder in assessment of corneal anterior topographic parameters is the masking effect of the true severity of the degenerate cornea by the epithelium.^12,13 Since epithelium in normal corneas is of uniform thickness, curvatures measured at the anterior corneal surface are expected to be similar to curvatures at the Bowman’s layer-epithelium interfaceoperatively in KC eyes of grade 1 severity on Amsler-Krumeich severity scale indicating that a FFKC eye could have undergone adverse changes at the Bowman’s layer well before any manifestation at the anterior corneal surface was observed.^12,13 This also indicates that degradation (thinning and irregularity) of the Bowman’s layer can be completely missed in conventional tomography (Palcido-disk and Scheimpflug) unless the epithelium is removed. The epithelium is thinner at the location of the cone and thicker in regions away from the cone, which explains the compensatory role of the epithelium in masking the true severity of the disease.¹³In this study as well, epithelium thickness of KC eyes was significantly lower than the thickness in normal eyes.

In this study, high resolution OCT was used to quantify the irregularity in the central region of normal, FFKC and KC eyes without any contact with the patient eye unlike ultrasound based techniques.¹³ It was able to quantify in vivo the sub-epithelium morphological changes independent of the masking effect of the epithelium, which would not have been possible with either Placido disk or Scheimpflug imaging systems in normal clinical setting. The study clearly showed that BRI was altered in KC eyes but was similar in FFKC and KC eyes. The decrease in BRI in KC indicated that with thinning, the irregularity  also became less. This may have a biomechanical attribute since a thinner layer will be stretched more due to the in-plane mechanical stresses induced by IOP or it may simply be an effect of tissue volume loss in FFKC and KC. These effects need to be evaluated further in future studies. Thus, morphological changes in the Bowman’s layer probably occur very early during the progression of the disease as stated earlier. While BAD-D enjoyed overall the best diagnostic performance, it had considerably lower sensitivity than BRI (96.6% vs. 80.7%) in the detection of FFKC from normal eyes. In contrast, BRI had lower specificity than BAD-D (53.3% vs. 100%). In an earlier study, thinning of the Bowman’s layer was studied in two ways: local thickness vs ratio of thickness at the location of the cone to thickness away from the cone in an unaffected region of the cornea.⁷ It was found that the latter method had significantly higher sensitivity and specificity in the detection of KC from normal eyes than the former method of using just local thickness of Bowman’s layer as a diagnostic index. Thus, it may be possible to further improve the overall sensitivity and specificity of BRI by normalizing the BRI at the location of the cone with it’scorresponding magnitude in the peripheral unaffected cornea of the same eye.

A limitation in the imaging of the peripheral corneal is a significant drop in signal to noise ratio away from the central cornea, which can affect the resolution of the Bowman’s layer.¹⁰ However, advancements in the field of OCT are rapidly occurring, which may allow quantification of the BRI of the peripheral cornea as well in the near future. In this study, corneal aberrations had lower area under the ROC curve compared to BAD-D and BRI but its’ sensitivity and specificity were similar to some recent studies on FFKC eyes, particular in case of anterior  surface aberrations.^2,11 In summary, BRI is a new technique for the quantification of the irregularity of the Bowman’s layer and was able to show disease specific changes in the Bowman’s layer of the FFKC eyes, which were similar to the changes in the KC eyes.

Financial disclosures: Dr. Sinha Roy has received research funding in the area of biomechanical modeling of the eye from Carl Zeiss Inc., Germany, Avedro Inc., USA, Topcon Medical Systems Inc., USA and Bioptigen Inc., USA. Dr. Sinha Roy has intellectual property related to computational modeling through Cleveland Clinic Innovations, Cleveland, USA. No other author has any financial or proprietary interests to declare. Dr. Sinha Roy and Dr Shetty have pending patent application on image based quantification of structural changes in the Bowman’s layer after surgery and in disease.

References

References

1. Ruberti JW, Sinha Roy A, Roberts CJ. Corneal biomechanics and biomaterials. Annu Rev Biomed Eng. 2011; 13: 269-95.
2. Bühren J, Kook D, Yoon G, Kohnen T. Detection of subclinical keratoconus by using corneal anterior and posterior surface aberrations and thickness spatial profiles. Invest Ophthalmol Vis Sci. 2010;51:3424-32.
3. 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: 2231-8.
4. 4.Smadja D, Touboul D, Cohen A, et al. Detection of subclinical keratoconus using an automated decision tree classification. Am J Ophthalmol. 2013;156:237-246.
5. 5.Chan C, Ang M, Saad A, etc. Validation of an objective scoring system for forme fruste keratoconus detection and post-LASIK ectasia risk assessment in asian eyes. Cornea. 2015;34:996-1004.
6. 6.Li X, Rabinowitz YS, Rasheed K, Yang H. Longitudinal study of the normal eyes in unilateral keratoconus patients. Ophthalmology. 2004;111:440-6.
7. 7.Abou Shousha M, Perez VL, Fraga Santini Canto AP, et al. The use of Bowman’s layer vertical topographic thickness map in the diagnosis of keratoconus. Ophthalmology. 2014;121:988-93.
8. .Sawaguchi S, Fukuchi T, Abe H, Kaiya T, Sugar J, Yue BY. Arch Ophthalmol. 1998;116:62-8.
9. 9.Ruiseñor Vázquez PR, Galletti JD, Minguez N, et al. Pentacam Scheimpflug tomography findings in topographically normal patients and subclinical keratoconus cases. Am J Ophthalmol. 2014;158:32-40.
10. 10.Larocca F, Chiu SJ, McNabb RP, Kuo AN, Izatt JA, Farsiu S. Robust automatic segmentation of corneal layer boundaries in SDOCT images using graph theory and dynamic programming. Biomed Opt Express. 2011;2:1524-38.
11. 11.Saad A, Gatinel D. Evaluation of total and corneal wavefront high order aberrations for the detection of forme fruste keratoconus. Invest Ophthalmol Vis Sci. 2012;53:2978-92.
12. 12.Touboul D, Trichet E, Binder PS, Praud D, Seguy C, Colin J. Comparison of front-surface corneal topography and Bowman membrane specular topography in keratoconus. J Cataract Refract Surg. 2012;38:1043-9.
13. 13.Reinstein DZ, Archer TJ, Urs R, Gobbe M, RoyChoudhury A, Silverman RH. Detection of Keratoconus in Clinically and Algorithmically Topographically Normal Fellow Eyes Using Epithelial Thickness Analysis. J Refract Surg. 2015; 31: 736-44.