The Evolution of Diagnostics for Keratoconus: From Ophthalmometry to Biomechanics

ABSTRACT

Purpose

To enumerate the various diagnostic modalities used for keratoconus and their evolution over the past century.

Methods

A comprehensive literature search including articles on diagnosis on keratoconus were searched on PUBMED and summarized in this review.

Results

Initially diagnosed in later stages of the disease process through clinical signs and retinoscopy, the initial introduction of corneal topography devices like Placido disc, photokeratoscopy, keratometry and computer-assisted videokeratography helped in the earlier detection of keratoconus. The evolution of corneal tomography, initially with slit scanning devices and later with Scheimpflug imaging, has vastly improved the accuracy and detection of clinical and sub-clinical disease. Analyzing the alteration in corneal biomechanics further contributed to the earlier detection of keratoconus even before the tomographic changes became evident. Anterior segment optical coherence tomography has proven to be a helpful adjuvant in diagnosing keratoconus, especially with epithelial thickness mapping. Confocal microscopy has helped us understand the alterations at a cellular level in keratoconic corneas.

Conclusion

Thus, the collective contribution of the various investigative modalities have greatly enhanced earlier and accurate detection of keratoconus, thus reducing the disease morbidity

KEYWORDS:

• Corneal biomechanics
• keratoconus
• machine learning
• optical coherence tomography
• Scheimpflug tomography

INTRODUCTION

The first comprehensive description of keratoconus was provided by John Nottingham in 1854 when he described a condition called ‘conical cornea’.¹ Since then, our understanding of the disease has evolved significantly. Keratoconus has been classically described as a bilateral, progressive, non-inflammatory disorder of the cornea leading to corneal thinning, protrusion, and irregular astigmatism, eventually resulting in impaired quality of vision.^2,3 The usual onset of the disease is in adolescence, and it progresses until the third or fourth decade of life.⁴

Very few articles are available in the literature that focuses on the evolution of the diagnosis of keratoconus. This review aims to provide a comprehensive analysis of the various diagnostic modalities used for keratoconus over the past century and describe their importance in ultimately diagnosing and treating the disease early.

METHODS

A thorough literature search for the components of this review was done using PubMed. The keywords in relation to keratoconus used for the search were “retinoscopy”, “corneal topography”, “pentacam”, “corneal biomechanics”, “ocular coherence tomography” and “confocal microscopy”. The search yielded a total of 3510 articles. The articles were screened, and only those focused on the diagnosis of keratoconus were included in this review.

Moderate to advanced keratoconus can be detected by clinical examination but these clinical signs (Table 1) usually appear later in the course of the disease. Amongst these, a good retinoscopy can pick up keratoconus much earlier but is subjective.

Table 1. Clinical signs of Keratoconus.

CLINICAL SIGNS OF KERATOCONUS
External signs	Munson’s sign Rizzuti phenomenon
Slit lamp signs	Corneal thinning Vogt striae Fleischer ring Apical scarring
Retroillumination signs	Scissoring on retinoscopy Charleaux oil droplet sign

RETINOSCOPY

Retinoscopy, invented by Hugo Wolff,⁵ is an indispensable component of objective refraction. In keratoconus, a scissor reflex is usually seen on retinoscopy, comprising an “against” motion in the centre and a “with” motion in the periphery.³ This split in the light reflex is called the scissor reflex. Goebels et al.⁶ found that retinoscopy was comparable to the Pentacam and Ocular Response Analyzer in confirming the diagnosis of keratoconus. Al-Mahrouqi et al. found that the scissoring reflex on retinoscopy is very sensitive for detecting keratoconus and is an economical and easy to use method to screen for keratoconus especially in population-based screening programs.⁷

CORNEAL TOPOGRAPHY

Helmholtz first introduced the keratometer (ophthalmometer) in 1854 as an instrument that provided information on 2–3 locations on the anterior corneal surface approximately 3 mm apart. The keratometer can detect keratoconus by identifying distortion of the reflected mires, and central or inferior steepening and is a sensitive indicator of keratoconus.³ The limitations of the keratometer include lack of information regarding the topography central or peripheral to the points of measurement and mire distortion even with mild surface irregularity preventing meaningful assessment.⁸

Marc Amsler, in 1938, used a photographic Placido disk to describe the topographic changes in keratoconus.⁹ He classified keratoconus based on the degree of deviation of the horizontal axis of the Placido disc and documented the natural history of progression of the disease. Amsler concluded that progression was likely to occur in patients between 10 and 20 years of age, whereas progression was less likely after 30 years.^9,10 However, reproducibility is a significant problem with this device due to instrument tilt or poor alignment resulting in incorrect interpretation.¹¹

Placido¹² and Gullstrand were the first to develop the photokeratoscope that provides topography of 55–80% of the anterior corneal surface, excluding the central 3 mm. The earliest sign detected on photokeratoscopy in keratoconus was compression of mires in the inferotemporal cornea (egg-shaped mires).¹³

In 1880, Antonio Placido¹⁴ introduced the principle of Placido ring-based videokeratoscopes, which use concentric rings of light and a digital camera to capture the reflection of the rings on the corneal surface, and the data is then processed using algorithms to illustrate the topography of the anterior corneal surface. These devices incorporate features of both the keratometer and a photokeratoscope. Rabinowitz and co-workers have described various patterns and indices for early detection of keratoconus using videokeratography.^15–18 Inferior steepening with asymmetric bowtie with a skewed radial axis (AB/SRAX) pattern was found in nearly 100% of patients with early keratoconus but only 0.5% of the normal population.¹⁹ Additionally, it was shown that 50% of eyes with an AB/SRAX pattern in the normal fellow eye of unilateral keratoconus ultimately progress to keratoconus, proving that this pattern could be a marker for forme fruste keratoconus (FFKC).^18,20 A few indices, including I-S value, the SRAX index and KISA% index, were also introduced to improve the detection of keratoconus.^15,16,18 However, none of these indices has proven to be 100% accurate in differentiating normal corneas from those that had early keratoconus.

CORNEAL TOMOGRAPHY

Corneal tomography provides a three-dimensional representation of the cornea with a detailed analysis of both the anterior and posterior corneal surfaces as well as pachymetric analysis. The evolution of corneal tomography has revolutionized the diagnosis of keratoconus and screening for refractive surgery.

The Orbscan (Bausch & Lomb; Rochester, USA) was the first corneal tomography method introduced in 1995. It uses the innovative principle of slit scanning technology to acquire information from the cornea’s anterior and posterior surfaces. In 1999, the Orbscan IIz (Bausch and Lomb, Rochester, USA) was introduced, which used a combination of Placido rings and slit scanning to assess the tomography and pachymetry. The Placido rings consist of 40 monochromatic concentric rings, and analysis of the reflection of these rings on the corneal surface provides the anterior corneal topography. The slit-scanning system utilizes 40 white light slits at angles of 45°, each measuring 12.5 × 0.3 mm, and gives information regarding the topography of the posterior corneal surface and corneal thickness. Recently, an artificial intelligence system named the Screening Corneal Objective Risk of Ectasia (SCORE) has been linked to the Orbscan IIz. This system uses a combination of 12 tomographic indices, including maximum posterior elevation, vertical decentration of thinnest point, thinnest corneal thickness and the difference between mean central and thinnest corneal thickness. These indices are used to derive a score and categorize the cornea as positive or negative for the risk of ectasia. The SCORE software has been validated in various subsets like FFKC and post LASIK ectasia.^21–23

Scheimpflug imaging of the cornea has more recently become the mainstay of corneal tomography. Jules Carpentier first described the Scheimpflug principle in 1901, which was cited and credited in the original patent by Theodor Scheimpflug in 1904. In Scheimpflug imaging, the planes are aligned to provide an extended depth of focus and more sharpness to points of the image with minimal distortion.²⁴

Digital rotating Scheimpflug tomography is an evolution compared to parallelopiped horizontal cross-sectioning for tomographic evaluation of the cornea and anterior segment used in the Orbscan.²⁵ The rotating system in Scheimpflug imaging has a common centre that makes image registration more accurate than horizontal slit scanning.^26,27 The Oculus Pentacam (Wetzlar, Germany) was the first system to use a digital rotating Scheimpflug camera for corneal and anterior segment tomography.²⁸ In addition to the rotating Scheimpflug camera, the pentacam has a central frontal camera that controls fixation and compensates for alignment. This high-resolution frontal camera also records pupil measurements of white-to-white diameter and provides orientation for reconstructing a three-dimensional tomography model. In around 2 seconds, a three-dimensional cornea and anterior segment model is generated from as many as 25,000 elevation points in the Pentacam and 138,000 elevations points in the Pentacam HR. The resultant Scheimpflug images provide data from the anterior and posterior surfaces of the cornea, angle, anterior iris and lens. Although the Pentacam provides very accurate tomography data, it has a few limitations. The wavelength of light used, 475 nm, is sensitive to corneal opacities leading to inaccuracies in contour analysis. Another limitation is the inability to visualize the anterior chamber angle directly due to total internal reflection in the peripheral cornea. However, the Pentacam has extrapolation software integrated into the system, accurately estimating the anterior chamber angle.^29,30 The other devices that use Scheimpflug photography include Galilei (Port, Switzerland), Sirius (Firenze, Italy), Tomey TMS-5 (Nagoya, Japan), Costruzione Strumenti Oftalmici (CSO; Florence, Italy) and the Preciso (iVIS Technologies, Taranto, Italy). The features of these devices in elaborated in Table 2.

Table 2. Comparison of various Scheimpflug-based tomography devices.

Imaging system	Pentacam	Galilei	Sirius	TMS −5	Preciso
Manufacturer	Oculus, Germany	Ziemer, Switzerland	CSO, Italy	Tomey, Japan	Ivis, Italy
Photography	rotating Scheimpflug + frontal static camera	dual rotating Scheimpflug + Placido disc	rotating Scheimpflug + Placido disc	rotating Scheimpflug camera incorporated into Placido disc	rotating Scheimpflug + Placido disc
Elevation maps	Yes	Yes	Yes	Yes	Yes
Refractive power of cornea map	Yes	Yes	Yes	Yes	Yes
Pachymetric maps	Yes	Yes	Yes	Yes	Yes
Thickness spatial profile graphs	Yes	No	No	No	No
Anterior chamber analysis	Yes	Yes	Yes	Yes	Yes

The elevation^31–33 and pachymetric indices (Figure 1A) derived from tomographic maps have been reported to be very effective in detecting early keratoconus.^{31,32,34–37} The corneal thickness spatial profile and percentage thickness increase graphs, available on Pentacam, have proven important in identifying keratoconus.^34,38–45 Pachymetric progression indices (PPI) are assessed for all hemi-meridians across the cornea so that the average (PPI Ave) and the meridian with maximal (PPI Max) pachymetric increase are noted.²⁵ The ratio between the thinnest point and PPI gives the Ambrosio’s relational thickness (ART). The Pentacam HR also provides the Belin-Ambrósio Enhanced Ectasia Display (BAD) (Figure 1B), with data demonstrating nine input parameters; anterior elevation and posterior elevation at the thinnest point, change in anterior and posterior elevation, location of the thinnest point, corneal thickness at the thinnest point, pachymetric progression, ART and Kmax.⁴⁶ Anterior and posterior elevation data is calculated utilizing a method of enhanced best-fit spheres (BFS), which are generated by excluding an area of 3.5 mm in diameter centred on the thinnest point. The resultant enhanced BFS is flatter than a standard BFS, which has minimal effect in normal corneas but highlights the abnormal areas of an ectatic cornea.^47,48 The BAD summarises the analyses in the form of normality indices as provided in Table 3. The software highlights values of these indices between 1.6 and 2.6 in yellow (suspect keratoconus) and values >2.6 in red (clinical keratoconus).⁴⁸ The final D value (BAD-D) is deduced from a regression analysis of all the parameters, and a value of more than 2.1 indicates ectasia. BAD-D has been shown to have the highest accuracy in detecting clinical and subclinical keratoconus.^25,49–55

Figure 1. Pentacam maps of a keratoconus patient A) 4 Maps refractive output showing specific tomographic features of keratoconus. B) Belin Ambrosio enhanced ectasia display of the same patient showing all BAD indices including BAD-D in red indicating keratoconus.

Table 3. Belin/Ambrósio enhanced ectasia display values.

Display Value	Description of Parameter
Df	Standard deviation from the mean anterior elevation
Db	Standard deviation from the mean of posterior elevation
Dp	Standard deviation from the mean of average pachymetric progression
Dt	Standard deviation from the mean of thinnest corneal thickness
Dy	Standard deviation from the mean displacement of the thinnest point along the vertical meridian
Da	Standard deviation from the mean of Ambrósio relational thickness
D	Final overall map reading

CORNEAL BIOMECHANICS

A decrease in corneal biomechanical strength has been recognized as a critical factor in the pathophysiology of keratoconus.^56–58 A combination of factors, including a decrease in the number of collagen lamellae and their altered orientation, reduced cross-links, and alterations in keratocyte density and extracellular matrix, contribute to the decrease in biomechanical resistance of the cornea to permanent deformation.^58–60 Investigating these changes in biomechanical properties could help in the earlier diagnosis of keratoconus, even before topographic and tomographic changes become evident.

The Ocular Response Analyzer (ORA, Reichert Ophthalmic Instruments, NY) was the first commercially available instrument to measure the biomechanical properties of the cornea in vivo.⁶¹ This instrument is a non-contact tonometer with a metered air puff to applanate the cornea. The infrared reflectance produces two distinct peaks: when the cornea moves inwards (P1) and when the cornea moves outwards (P2). The values for P1 and P2 equate to the emitted air-pulse pressure at the respective applanation events within the 3 mm sampling zone.⁶¹

The two main biomechanical outputs from the ORA are corneal hysteresis (CH) and corneal resistance factor (CRF). CH is defined as the difference between pressures at P1 and P2, and CRF is calculated as (P1 – kP2), with a proprietary factor k (0.7) derived by the manufacturer.⁶¹ Though these parameters have been noted to be lower in keratoconus than normal corneas,^61–67 they lack sensitivity and specificity for diagnosing keratoconus as a significant overlap exists between the two groups. Hence, these parameters cannot be used for the diagnosis of early keratoconus.^65–67 However, newer parameters were derived from the ORA waveform to improve the diagnostic accuracy.^68–70 Avetisov et al.⁷¹ found a statistical difference in a mathematically derived coefficient of elasticity between healthy and keratoconic corneas. Luz et al.⁷² used logistic regression analysis to produce a combined tomographic and biomechanical linear model, which improved the accuracy of diagnosis of FFKC.

The CorVis ST (CST, Oculus; Wetzlar, Germany) was introduced as an alternative to measuring the biomechanical properties of the cornea in vivo. CST is also a non-contact tonometer like the ORA and emits an air pulse. However, in contrast to ORA, it uses an ultra-high-speed Scheimpflug camera (4330 frames/sec) to record the ocular response to the air pulse across an 8 mm wide horizontal cross-section of the cornea. The data is analyzed in real-time to derive various parameters (Table 4) that help provide information regarding biomechanical properties⁷³ (Figure 2). Similar to the ORA, though several parameters of CST were significantly different in keratoconic corneas when compared with healthy corneas,^73–75 a substantial overlap existed between the two groups.^74,75 However, analysis of deformation amplitude and deflection amplitude was found to differentiate suspect and keratoconic eyes from normal eyes.⁷⁶

Figure 2. Tomographic biomechanical display (TBI) from Pentacam and Corvis ST showing abnormal corneal biomechanics and abnormal CBI, BAD D and TBI values indicative of keratoconus.

Table 4. Corneal deformation parameters provided by the corvis ST.

1st Applanation Moment of first applanation of the cornea during the air puff (in milliseconds). In parenthesis is the length of the applanation at this moment (in millimeters). Highest Concavity The instant that the cornea assumes its maximum concavity during the air puff (in milliseconds). In parenthesis is the length of the distance between the two peaks of the cornea at this moment (in millimeters). 2nd Applanation The second applanation of the cornea during the air puff (in milliseconds). In parenthesis is the length of the applanation at this moment (in millimeters). Maximum Deformation Measurement (in millimeters) of the maximum cornea deformation during the air puff. Wing Distance The length of the distance between the two peaks of the cornea at this moment (in millimeters). Maximum Velocity (in) Maximum velocity during the ingoing phase (in meters per seconds (m/s)). Maximum Velocity (out) The maximum velocity during the outgoing phase (in meters per seconds (m/s)). Curvature Radius Normal Radius curvature of the cornea in its natural state (in millimeters). Curvature Radius HC Radius of curvature of the cornea at the time of maximum concavity during the air puff (in millimeters). Cornea Thickness Measurement of the corneal thickness (in millimeters). Integrated Inverse Radius Inverse of the radius of curvature during the concave phase of the deformation. Deformation Amplitude Ratio 1 or 2 mm The central cornea deformation divided by an average of the deformation 1 or 2 mm at either side of the center with maximum value just before 1st applanation IOP Measurement of the intraocular pressure (in millimeters of mercury (mmHg)). bIOP Biomechanically-corrected IOP

Vinciguerra et al.⁷⁷ combined corneal deformation parameters and horizontal thickness profile to develop the Corvis corneal biomechanical index (CBI), which was highly sensitive and specific to differentiate keratoconic and normal eyes. Ambrósio et al.⁷⁸ integrated data obtained from Scheimpflug-based corneal tomography and analysis of corneal biomechanical properties using artificial intelligence techniques to develop the tomographic biomechanical index (TBI). TBI has proven better than other investigative parameters, including CBI and BAD-D, in detecting corneal ectasia.^78,79 Validation studies have also demonstrated that TBI increases the ability to detect early ectasia even in subclinical cases^80–82

Brillouin Spectroscopy is a newer modality for analyzing corneal biomechanics. The principle is based on the interaction of light (laser) and the intrinsic acoustic waves within the tissue. The laser’s interaction with the tissue causes a Doppler frequency shift associated with the elastic modulus,⁸³ thus analyzing the biomechanics of the cornea, sclera and lens.^84–87 Scarcelli et al.⁸⁸ reported ex vivo ectatic corneas having significantly smaller Brillouin frequency shift than normal corneas. It was also found that there was a significant difference between Brillouin frequency shift in the cone region compared to other areas in vivo.⁸³ Further studies have distinguished mild KC from normal and clinical KC by comparing their Brillouin frequency, proving its efficacy and accuracy.^89,90

OPTICAL COHERENCE TOMOGRAPHY

Anterior segment optical coherence tomography (AS-OCT) helps assess the microstructure of the cornea in vivo by producing high resolution, cross-sectional images using the principle of optical light scattering in combination with low coherence interferometry.⁹¹

AS-OCT has gained importance in epithelial thickness mapping in keratoconus, providing a potential avenue for early diagnosis. Early histological changes in keratoconus include reduced corneal basal epithelial cell density, Bowman’s layer instability and loss of anterior stromal collagen fibrils.^92,93 The features that have been demonstrated in epithelial thickness maps on AS-OCT in eyes with keratoconus are apical epithelial thinning, decreased overall epithelial thickness, thicker epithelial layer supero-nasally (in a typical inferotemporal cone), greater differences between the minimum and maximum epithelial thickness and greater variability of the pattern standard deviation and map standard deviation compared with normal eyes.^39,94–96 Apical epithelial thinning occurs due to a compensatory remodelling in response to the underlying stromal thinning to reduce anterior surface irregularity^94,97 (Figure 3). This may lead to masking the stromal thinning when the total corneal thickness is measured or cause topographic images to appear relatively normal, especially in early keratoconus.^39,98 Therefore, analysis of the epithelium and stroma separately with the AS-OCT could help in improving the accuracy of diagnosis. However, it has been shown that epithelial thickness parameters are insufficient as independent diagnostic tools and should be used in conjunction with corneal tomography to diagnose keratoconus.^56,99

Figure 3. AS-OCT of a keratoconic cornea showing compensatory changes in the epithelial thickness (white arrow) as a response to underlying stromal thinning thus resulting in a regular anterior surface.

Li et al.^94,100 used the OCT to develop an extended epithelial thickness map and various indices to detect early KC. Pahuja et al.¹⁰¹ studied Bowman’s layer irregularity in normal and ectatic corneas using OCT and introduced the Bowman’s roughness index. This index successfully detected keratoconus and, in combination with the BAD-D and epithelial thickness data, improved the sensitivity for the detection of early keratoconus. Hwang et al.¹⁰² reported that the combination of parameters from Scheimpflug imaging and OCT improved the accuracy in detecting very asymmetric keratoconus.

Polarization-sensitive optical coherence tomography (PS-OCT) enhances the contrast of fibrous tissues by measuring birefringence.¹⁰³ It is well known that the arrangement of collagen fibril lamellae is altered in keratoconus,^59,104,105 which results in modifications in their birefringence. Based on this principle, it was shown that phase retardation of the posterior corneal surface was increased in ectatic corneas due to an increase in birefringence.¹⁰⁶ It was also shown that phase retardation was sensitive in discriminating keratoconus suspects and can help diagnose early or subclinical keratoconus.¹⁰⁶

AS-OCT can also be used for detecting risk factors for developing acute hydrops in keratoconus. Fuentes et al.¹⁰⁷ reported features on AS-OCT such as increased epithelial thickness, a higher epithelial to stromal thickness ratio, stromal thinning at the corneal apex and hyper-reflectivity of Bowman’s layer are associated with an increased risk of developing hydrops. The authors also hypothesized that corneal scarring might increase corneal rigidity, thus protecting against developing hydrops. In cases of corneal hydrops, AS-OCT can help assess and classify the extent and edges of Descemet’s membrane (DM) detachment. Basu et al.¹⁰⁸ reported that eyes with deeper DM detachments and larger DM breaks required more time to resolve corneal oedema.

CONFOCAL MICROSCOPY

In vivo confocal microscopy (IVCM) allows imaging of the cornea at a cellular level. The features of keratoconus on IVCM include decreased keratocyte density in the anterior and posterior stroma,^60,109–112 decreased basal epithelial cell and endothelium densities along with increased pleomorphism and polymegethism,¹⁰⁹ significantly lower nerve density¹¹³ and alteration in the arrangement of the subbasal nerve plexus.¹¹⁴ Interestingly, keratoconus patients who use contact lenses have significantly lower anterior keratocyte density than those who do not.^60,111,112

WAVEFRONT ANALYSIS

In keratoconus, irregular astigmatism increases lower and higher-order aberrations.¹¹⁵ Many authors have shown that corneal and total wavefront ocular aberrations are higher in ectatic corneas when compared to healthy ones.^115,116 Higher-order aberrations, especially vertical coma and trefoil, have been reported to be significantly higher in mild keratoconus than in normal eyes.^{115,117–120} and can potentially be used for the early detection of the disease.

MACHINE LEARNING

Machine learning refers to an artificial intelligence technique where a dataset of known inputs is entered into a machine programmed to maximize output accuracy.¹²¹ Over the past two decades, multiple machine learning methods have been employed to diagnose keratoconus with varying parameters to improve detection accuracy.^{43,97,120,122–133} Recently, Lopes et al.¹³³ reported the most extensive machine learning study to detect clinical ectasia and ectasia susceptibility. The authors introduced a new Pentacam random forest index (PRFI), proven to be statistically superior to BAD-D when assessing all ectasia cases. The advantage of machine learning methods is that many refractive indices can be evaluated at once, improving the diagnostic accuracy and decreasing the dependence of subjective operators. However, lack of standardization across the data subsets provides the most significant challenge in drawing comparative conclusions currently but has excellent potential for the future.

SUMMARY

The increasing popularity of laser refractive surgery over the past two decades has accelerated the need for accurate detection of subclinical keratoconus. The constant evolution and distinctive contribution of the diagnostic modalities in keratoconus have vastly improved the detection and aided in the timely treatment of the disease. Corneal tomography is still the most widely used tool for detecting keratoconus. However, the increasing importance of the other modalities like corneal biomechanics assessment and anterior segment optical coherence tomography has proven that a multi-pronged approach can help in increasing the diagnostic accuracy. Analysis of multiple parameters with machine learning will significantly enhance the speed and accuracy of detection and decrease subjective dependence.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Additional information

Funding

The authors reported there is no funding associated with the work featured in this article.