New advances in imaging allow objective measurements for dry eye as well as define new parameters that cannot be measured by clinical assessment alone. A combination of these modalities provides unprecedented information on the static and dynamic properties of the structural and functional parameters in this multifactorial disease. A literature search was conducted to include studies investigating the use of imaging techniques in dry eye disease. This review describes the application of non-invasive tear breakup time, optical coherence tomography, meibomian gland imaging, interferometry, in vivo confocal microscopy, thermography and optical quality assessment for this condition.
- ocular surface
- diagnostic tests/investigation
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
The diagnosis and monitoring of dry eye disease (DED) is challenging in clinical practice.1 This may be attributed to the multifactorial nature of the disease as well as a poor correlation between the signs and symptoms of DED.2 3 The concepts of tear osmolarity and pro-inflammatory cytokines have now been translated into clinical applications with point-of-care devices allowing us to better diagnose and monitor DED.4–6 Paralleling these advances are the different imaging modalities that allow us to evaluate the structural and dynamic properties of the tear film. This review aims to discuss the recent findings and ongoing developments of imaging modalities in DED.
Methods of literature search
The modalities covered in the review will include non-invasive tear breakup time (NITBUT) measurement, meibomian gland imaging, interferometry, optical coherence tomography (OCT), confocal microscopy, thermography and optical quality. A brief technical overview of each modality, its correlation with objective clinical parameters and subjective dry eyes scores, its discriminatory ability in diagnosing DED, as well as the repeatability of measurements will be discussed. We searched for English articles in PubMed and MEDLINE databases up to November 2016. Keywords including ‘dry eye disease’, ‘tear film’, ‘imaging’, ‘diagnosis’, ‘keratography’, ‘meibography’, ‘inteferometry’, ‘OCT’, ‘confocal microscopy’, ‘thermography’ and ‘optical quality’ were used in various and/or logic combinations.
Non-invasive tear breakup time
Fluorescein tear breakup time (FTBUT) is one of the most widely used clinical tests for assessing tear film instability. Despite its widespread use, it cannot simultaneously assess the tear breakup across the entire cornea. Furthermore, FTBUT has suboptimal accuracy and repeatability in DED due to the variability of the concentration and volume of fluorescein installed in the conjunctival fornix, which can also induce reflex tearing.7 To overcome the inherent pitfalls in FTBUT, NITBUT, where fluorescein instillation is not required, was introduced. It measures the reflected grid or videokeratographic mires from an illuminated placido disc and a change in the edges of the mires reflect compromised tear film integrity.8 Different commercially available topography systems with different algorithms used in these tear stability analysis system can determine the changes of these placido mires over time. The first NITBUT (NITBUTf) is the time between blinking until the appearance of first discontinuity in the mires. The average NITBUT (NITBUTavg) represents the average of all the tear film breakup time (TFBUT) over the entire cornea.9 In contrast to calculating the NITBUT based on changes in corneal power, the ring breakup time (RBUT) calculates the NITBUT based on the difference in brightness at each measurement point on the mires.10 Both NITBUTf and NITBUTavg are correlated with Ocular Surface Disease Index (OSDI) score.11 NITBUTf has better within-visit agreement with FTBUT than NITBUTavg 12 A 12.1 s NITBUTf has a 82% sensitivity and 94% specificity while a 5 s RBUT cut-off has a 82% sensitivity and 60% specificity in distinguishing DED from healthy subjects.10 13 These different platforms and algorithms have poor agreement among themselves and are not interchangeable.14 A calibration offset is required in order to make these values comparable.15
Optical coherence tomography
Tear meniscus measurements are important parameters in the diagnosis of DED.16 Anterior segment OCT allows measurement of the tear meniscus by using low coherence interferometry to produce a two-dimensional image of optical scattering.17 Tear meniscus height (TMH) measured by time-domain OCT (TDOCT) was the first tear meniscus parameter identified to be lower in DED and was correlated with the Schirmer’s test score, tear stability and corneal vital staining scores.18 Using lower TMH of <0.30 mm, the sensitivity and specificity in identifying DED were 67% and 81%, respectively.18 Development of spectral-domain (SD) OCT allows higher optical resolution and faster scan rate than time-domain OCT. It has better intraobserver and interobserver repeatability of TMH and tear meniscus area (TMA) measurements; however, the intra-device agreement was poor and is not interchangeable.19 20 In a study that compared the Cirrus SDOCT (Carl Zeiss Meditec, Dublin, California, USA) with the Visante TDOCT (Carl Zeiss Meditec), the 95% limits of agreement (LOA) for TMH and TMA were −22 to 66 µm and −1632 to 5331 µm2 respectively for SDOCT and −125 to 45 µm and −38 050 to 21 874 µm2 respectively for TDOCT.20 SDOCT has the potential to distinguish subgroups of DED and allows better understanding of the tear meniscus and ocular surface relationship.21 TMH was lowest in Sjogren’s syndrome dry eye (SSDE), followed by aqueous deficient non-SSDE, and healthy subjects, where a TMH of <0.21 mm was found to have a 4.65 relative risk ratio of developing severe corneal epithelial disease DED.22 OCT also allows real-time evaluation of the response and changes in tear film to different formulations of artificial tear drops.23 The two-dimensional property of TMA has a higher discriminating ability in diagnosing DED and provides additional details compared with the TMH, which is one-dimensional.24 Swept source OCT (SSOCT) uses a long wavelength of 1310 nm and yields three-dimensional imaging of the anterior segment of eyes. SSOCT enables even faster data acquisition and greater imaging depth than conventional time-domain and spectral-domain methods.25 The three-dimensional imaging makes it possible to measure the tear meniscus volume (TMV) in addition to TMH and TMA where all three parameters have high interobserver intraclass correlations of >95%.26 All three parameters correlated with Schirmer’s test scores, corneal vital staining score and TBUT, with Schirmer’s test score having the strongest correlation, which suggests the OCT tear meniscus parameters likely reflect the tear fluid quantity.27 The increase in TMH, TMA and TMH also varies depending on the instillation of different artificial tears of different retentive properties, which allows quantitative evaluation of the tear fluid dynamics after treatment.28 While Schrimer’s I test is a convenient and low-cost clinical test and acts as a surrogate for tear volume, it has poor reproducibility and is inherently invasive and causes reflex tear secretion. OCT offers a reproducible, repeatable and non-invasive assessment of the tear meniscus parameters with low variability. However, it is important to take into consideration the location of the punctum, palpebral aperture, lid length and presence of conjunctivochalasis in interpreting these tear meniscus parameters.29
Meibomian gland imaging
Meibomian gland dysfunction (MGD) ‘is a chronic, diffuse abnormality of the meibomian glands, commonly characterized by terminal duct obstruction and/or qualitative/quantitative changes in the glandular secretion’ and is the leading cause of DED.30 Non-contact infrared meibography provides two-dimensional details of the silhouette of meibomian glands by retroillumination using an infrared filter, allowing objective evaluation of structural abnormalities or gland dropout. Scoring systems such as meiboscore proposed by Arita et al can be used to quantify the degree of meibomian gland dropout in the upper and lower lids and to correlate the gland loss to clinical parameters.31 Using this proposed four-point integer meiboscore system, the repeatability of meibomian gland dropout shows good interobserver and intraobserver repeatability with a mean deviation of ≤0.18, but different infrared systems cannot be interchanged.32 Expressibility of meibomian glands is one of the most important functional tests in diagnosing MGD and is negatively correlated with the meibomian gland atrophy.33 Correlation between meibomian gland dropout and Schrimer’s test score in patients with MGD suggests that tear fluid secretion may increase as a compensatory response of the loss of tear film stability due to the deficiency of the lipid layer.34 Meibography also provides detailed morphological assessment of the meibomian glands. The ductal length and acini area are also correlated with tear film and ocular surface epithelium in DED.35 However, infrared meibography is limited by its two-dimensional imaging nature, thus it cannot provide any depth information. The infrared wavelength has greater absorption and scattering properties at the epidermis and dermis,36 which will blur and defocus the image. To overcome these challenges, three-dimensional meibomian gland imaging can be obtained via SSOCT. OCT can provide detailed images of the meibomian gland acini and ducts not seen on infrared imaging; however, the correlation between acinar morphological changes (constricted, atrophic, absent) was not well correlated with meibomian gland dropout.37 OCT meibography demonstrated that in obstructive MGD there is a decrease in meibomian gland length and width compared with the control, which are correlated with OSDI score and TFBUT, respectively; however, no differences in the number and diameter of meibomian gland orifices were found between the two groups.38 This suggests that in addition to meibomian gland obstruction, an atrophic process may also play a role in the decrease in meibum secretion in MGD. Furthermore, meibomian gland activity is physiologically asymmetric along the lid margins; therefore, the correlation with clinical symptoms may depend on its location.39 These recent findings using OCT meibography imply that the detection of meibomian gland dropout on infrared meibography should be carefully interpreted and not be used as a single test to diagnose MGD.
Interferometry describes a colour interference pattern produced by specular reflection at the lucent lipid–aqueous interface of the tear film when light is projected over the cornea. Since first described in 1968 by McDonald et al, this phenomenon has been used as a determinant of the integrity of the lipid layer of the tear film.40 LipiView (TearScience, Morrisville, USA) is one of the commercially available interferometers that provides quantitative values of the tear film lipid layer thickness (LLT) in interferometric unit (ICU) by analysing these specular observations in terms of their fringe pattern and colour where 1 ICU corresponds to approximately 1 nm. The LLT intraobserver coefficient of repeatability (COR) was 16 nm and the 95% LOA ranged from −14 to 18 nm, while the interobserver repeatability was comparable with a COR of 13 nm (95% LOA: 9–16 nm).41 Compared with healthy eyes, LLT is significantly thinner in those with obstructive MGD and is negatively correlated with upper and lower meibomian gland loss in both healthy eyes and eyes with obstructive MGD.42 LLT may be a marker of changes in meibum secretion and helps to differentiate obstructive from hypersecrectory MGD. The LLT is reduced in obstructive MGD and increased in hypersecrectory MGD.43 LLT of ≤75 nm has been suggested as a threshold of identifying obstructive MGD (sensitivity of 65.8% and specificity of 63.4%).44 However, demographic factors such as age, sex, ocular surgical history and MGD type can affect the LLT.45 Further studies are needed to establish a normative database to account for other confounding factors in order to define a threshold of LLT value. Apart from the LLT, interferometric colour and fringe pattern may also reflect the balance of aqueous and lipid layers of the tear film and may be useful in identifying the subtypes of DED. These colours and patterns were classified as monotonous grey, multicolour or greyish amorphous interferometric fringe with a κ value of 0.90 in assessing the intraobserver reliability.46
Prior to the development of the LipiView system, the DR-1 tear interference camera (Kowa, Nagoya, Japan) was considered to be the most sophisticated commercially available system for measuring interference. Yokoi et al proposed a grading scheme for the lipid layer interference pattern according to the uniformity of distribution and the colour of the tear interference images, which demonstrated a significant correlation with fluorescein staining and TFBUT.47 Goto et al quantified the interference image by a colorimetric approach, allowing the interference colour from the DR-1 images to be converted to LLT.48 The DR-1 system is also capable of video capture, allowing kinetic analysis of lipid spread time and pattern, stability of the lipid layer after spread and its distribution. In healthy subjects, the lipid spread time was 0.36 s, whereas in patients with MGD-related tear lipid deficiency, it was significantly longer at 3.54 s, with a different spread pattern.49 Kinetic analysis also allows comparison of the lipid layer before and after treatment in order to assess the therapeutic effectiveness of different treatment modalities.50 51 In aqueous tear deficiency dry eyes, kinetic analysis of the tear interference images showed that punctal occlusion can improve lipid spread, uniformity and thickness, suggesting that the lipid layer quality may be associated with the amount of aqueous tear fluid.51
In vivo confocal microscopy
Confocal microscopy produces a focal volume defined by the aperture size, magnification and working distance when the light source and objective lens are focused on a small finite area of interest. This produces a resolution comparable to histological analysis and provides a real-time non-invasive tool to study the ocular surface at a cellular level. In vivo confocal microscopy (IVCM) has been used to analyse the morphological–functional unit of the ocular surface. Corneal epithelial cell density, conjunctival inflammatory cell density and eyelid margin epithelial cell density were different among healthy, SSDE and non-SSDE patients.52 53 Subbasal corneal nerve alternations (lower density, increased tortuosity, number of beadings and width) were found in non-SSDE patients compared with healthy controls. These structural changes are correlated with functional alterations and severity of DED.54 Subbasal nerve alterations may be a characteristic of metabolically active subbasal nerves from tissue damage which are part of the pathological mechanisms in DED.55 This potentially opens new perspective for neurotrophic approach for DED treatment.56 Response to treatment of DED also differs depending on subbasal nerve length. It has been reported that improvements in clinical symptoms were present only in patients with near-normal nerve lengths,57 which may explain the variability in response to treatment and can help to select patients for anti-inflammatory treatment.58 Morphological changes may not be limited to the ocular surface: in moderate to severe DED. There is a significant reduction in central corneal endothelial cell density which is correlated with the severity of DED.59 This observation remains to be investigated but these structural changes may be due to a common inflammatory pathway. It is also possible that the subbasal nerves may have a role in endothelial cell function as patients with reduced subbasal nerve density have an accelerated endothelial loss.60 IVCM findings in patients with DED are also useful in differentiating between DED subtypes, where corneal dendritic cell density is higher in aqueous deficient subtype than evaporative subtypes, while the morphological dendritic cell features also differ between aqueous-deficient subtype with or without underlying systemic immune disease.61 These changes may reflect the different immune and inflammatory activities of the complex DED pathogenesis.
Similar to infrared and OCT meibography, IVCM can be used to provide high-resolution imaging of the meibomian glands. New parameters, such as meibomian gland acinar unit density (MGAUD), longest (MGALD) and shortest diameter (MGASD), and inflammatory cell density can be derived using IVCM.62 63 At cut-off values of 70 glands/mm2, 65 µm, 25 µm, 300 cells/mm2 for MGAUD, MGALD, MGASD and inflammatory cell density, respectively; the sensitivity and specificity values were 81% and 81% for MGAUD, 90% and 81% for MGALD, 86% and 96% for MGASD, and 100% and 100% for inflammatory cell density for the diagnosis of MGD.63 Meibomian gland parameters on IVCM are significantly correlated with tear film metrics, ocular surface damage and meibomian gland expressibility. Confocal features also show discernible meibomian gland characteristics among ocular surface disorders not easily distinguishable with other meibomian imaging techniques, for example, periglandular inflammatory changes without dilatative morphological changes were present in patients with Sjogren’s syndrome, while signs of glandular obstruction and distension and found in patients with MGD.64
Thermography is a non-invasive technique for measuring the surface temperature of an object. Originally designed to detect changes in the skin temperature, there is emerging interest in its ability to detect changes in the ocular surface temperature (OST) caused by tear film evaporation. Infrared ocular thermography measures the amount of infrared radiation emitted from the ocular surface or periorbital skin with an infrared thermal imaging camera. As the corneal surface temperature does not change diurnally in healthy subjects,65 OST can reflect the nature and stability of tear film. An unstable tear film in DED increases tear fluid evaporation and results in heat vaporisation leading to a reduction in OST.66 When the change in corneal apex OST after a 10 s period was used as an index for DED, it had a sensitivity of 0.83 and specificity of 0.80 in identifying patients with DED. The decrease in OST was also significantly correlated with TFBUT.67 Other indices such as temperature differences between differential zones of the cornea, limbal temperature or temperature of the conjunctiva have been used in differentiating between DED and healthy subjects.68 The geometric centre of the cornea, mean OST, extreme or mid-temporal/nasal conjunctiva, temporal/nasal conjunctiva, minimum or maximum temperature of the ocular surface had lower inter-image repeatability measured using coefficient of repeatability (COR): (%COR: 0.2–0.9), interoccasion variability (%COR: 2.1–3.7) and interexaminer variability (%COR: 1.5–3.7) than other indices such as temperature SD of the ocular surface and radial temperature difference, which had a much larger inter-image variability (%COR: 8.9–140.7), interoccasion variability (%COR: 47.5–153.5) and interexaminer variability (%COR: 54.7–142.0).68 The static OST measurements have higher discriminating ability in identifying DED than the dynamic measurements which study the temperature change over time.69 Both upper and lower tarsal conjunctival temperature were lower in patients with obstructive MGD, which can increase the viscosity of the meibum leading to obstructive MGD.70 Future studies need to elucidate whether the reduction of conjunctival temperature is a result of impaired blood flow. Apart from being a non-invasive tool for DED screening, specific baseline thermography image texture features, extracted using Fourier spectrum, fractal dimension and grey-level co-occurrence matrix were predictive of symptomatic improvement after lid warming therapy in MGD.71
Patients with DED often complain of visual disturbances such as blurred vision, fluctuation in vision with blinking and glare. The air–tear film interface is the first optical surface of the eye and has a high refractive power, thus its irregularity can have considerable impact on the optical quality. The deficient or unstable tear film in DED induces local changes of tear film thickness and irregularity which introduce aberration and scattering. Aberrations and scattering are the main factors in optical degradation in human eyes. Aberrations are caused by irregularities of the anterior/posterior corneal surfaces or precorneal tear film. While visual acuity measuring the spheres and cylinders can detect low-order aberration, wavefront sensors such as a Hartmann Shack sensor are needed to evaluate and quantify higher-order aberration (HOAs). HOAs are analysed over the central cornea up to the sixth order by expanding the set of Zernike polynomials. From the Zernike coefficients, the root mean square (RMS) was calculated to represent the wavefront aberrations: S3, S4, S5 and S6 are the RMS of the third-order, fourth-order, fifth-order and sixth-order Zernike coefficients, respectively. Coma-like aberrations (S3 + S5), spherical-like aberrations (S4 + S6) and total HOAs (S3 + S4 + S5 + S6) were calculated. HOA significantly increases over time after a blink in patients with DED and this change in HOA was correlated with the OSDI overall score and TFBUT.72 HOAs after tear film breakup were significantly increased compared with before tear film breakup regardless of pupil size, coma-like, spherical-like or total HOAs.73 Patients with tear film instability show an upward curve in post-blink HOA that continues to increase over time. These patients have a good simulated retinal image immediately after a blink, but the quality of the image continues to degrade over time before the next blink. In contrast to aqueous-deficient patients with DED, the HOAs are consistently higher due to the low tear volume, resulting in an impaired simulated retinal image even immediately post blink.74 75
Light scattering property
Quantifying HOA with wavefront sensors is a useful and objective method to evaluate the serial changes of optical quality in tear film dysfunction; however, it neglects the contribution of scattering and can overestimate the optical quality. Scattering can be divided into forward light scattering (towards the retina) and backward light scattering (scattered backwards off the cornea). The veiling luminance produced by forward light scattering can lead to glare. Whereas corneal backward light scattering, associated with decreased corneal transparency, is less often associated with visual complaints than forward light scattering. A double-pass retinal imaging technique, based on recording from a point source after reflection in the retina and a double-pass through ocular media, was developed to evaluate the objective scatter index (OSI) which measures the amount of light scattering as it passes through the different ocular structure and is not measurable using traditional wavefront aberrometry.76 There are fewer studies evaluating light scattering compared with aberrations in DED. This may be due to a limited availability of instruments and the complex relationship between forward and backward light scattering.77 Higher OSI is correlated with greater light scattering and lower visual quality.78 The change in OSI and its rate of change were found to be highest in more severe forms of DED, followed by mild DED and healthy subjects. These parameters are also correlated with clinical findings of tear film instability and corneal vital staining score.79 In patients with DED with or without superficial punctate keratopathy (SPK), both forward and backward light scattering were greater than healthy eyes, but correlation was found only between the SPK score and backward light scattering and not for forward light scattering.80 In terms of the effect of central SPK on total ocular HOAs, both baseline and sequential post-blink changes were higher in dry eyes with central SPK than those without, which were similar to the pattern of normal eyes.75 Further studies are warranted to understand the relationship between corneal backward light scattering and the degradation of retinal image in DED. The results of optical degradation can affect visual performances such as driving, for example, more targets were missed and the response time in patients with DED was increased in a driving simulator. The response time was found to be correlated with the change in HOA.81 An improvement in HOA and light scattering were reported after instilling artificial tear drops in patients with DED.82 83 Measurement of optical properties allows us to evaluate how different preparations of artificial tear drops can affect HOA and light scattering after therapy.84
A summary of techniques in dry eye imaging is shown in table 1.
Conventional DED diagnostic tests have poor to moderate repeatability.85 Although these imaging modalities are all non-invasive, the imaging sequence may affect the readings of subsequent measurements. Forced eye opening required for NITBUT measurement may cause reflex tearing even in patients with aqueous-deficient DED and results in erroneously high TMH measurement later.86 Careful planning of the sequence of these imaging tests is needed to minimise the potential confounding effects.
Most reported studies evaluated the correlation between the clinical dry eyes testing or patient’s symptoms with the measurement from the imaging modalities. While different measurements may reflect the same tear film properties, few studies compared the different modalities. Temperature difference 3 s after blinking measured on thermography was correlated with TMH measured on OCT and Schrimer’s test value, suggesting that both are suitable parameters to assess tear film structure.87 While IVCM provides excellent details of cornea structure and morphology, it requires contact between the instrument and the patient’s ocular surface and it has limited scanning area. OCT allows non-contact assessment of a larger corneal epithelial surface, extending up to 6×6 mm. Results of epithelial measurements for DED using OCT however have been contrasting, with reports of augmented epithelial thickness in early-stage subclinical DED,88 while another study demonstrated no difference in the central corneal epithelium but a thinner superior corneal epithelium in DED.89 However, the observed difference of corneal epithelium between DED and healthy subjects is small and approaches the 5 µm maximal resolution limit of most anterior segment OCT devices. IVCM may be more suitable than OCT in identifying the cellular changes in corneal epithelial such as epithelial hypertrophy or hyperplasia, or increase in the cellular layers in DED. Detection of DED changes also depends on its severity, for example, while differences of TMV measured on OCT were detected only between patients with moderate to severe DED and controls, different layers of corneal epithelial cell density measured on IVCM were reduced across all severities of DED compared with controls. Thus the ocular surface damage may be present before the changes in the tear film structure.90
DED is a multifactorial disease of the tears and ocular surface that results in not only symptoms of discomfort but also visual disturbance.91 Conventional dry eyes exams are targeted at the static and dynamic properties of the tear film, while imaging modalities offer additional benefits by being non-invasive, repeatable and objective. Yet, the findings from these examinations cannot explain why patients with DED often complain of visual disturbance despite a visual acuity of 20/20. In conventional best-corrected visual acuity (BCVA) testing, patients are allowed to blink as frequently as necessary to compensate for their dysfunctional or unstable tear film, which may lead to a BCVA of 20/20. Functional visual acuity (FVA) is measured after sustained eye opening for at least 10 s. Unlike conventional BCVA testing, the FVA mimics common daily tasks that tend to reduce blinking such as reading or using a computer.92 Goto et al described FVA evaluation by measuring surface regularity index (SRI) with corneal topography after sustained eye blinking. It was found that SRI was significantly reduced in patients with DED with or without Sjogren’s syndrome compared with BCVA, whereas the SRI remained similar to BCVA in the control eyes.92 FVA measurement systems have been developed to assess the change in visual quality over time, such as maximal and minimal visual acuity (maximal and minimal visual acuity obtained during the measurement period) and visual maintenance ratio (FVA divided by baseline visual acuity).93 94 In patients with DED, the mean FVA was significantly lower compared with healthy controls at all time points with significant improvement after punctum plug insertion.93 With the recognition that DED affects visual function, assessment of FVA and sequential HOAs is important in understanding the optical quality in dry eyes with altered tear film dynamics.
Advances in ocular imaging allow repeatable and objective measurements of DED. The information generated furthers our understanding of DED pathogenesis. All these methods are new and need to be proven to be reliable by further investigations. Further research is warranted to improve the correlation of ocular imaging with clinical findings in DED. Cut-off values for the different parameters measured by these imaging modalities still warrant ongoing investigations to establish a sensible level of sensitivity and specificity to diagnose DED. The cost of integrating these instruments into daily clinical practice may be justified if they improve our ability to diagnose DED and help to objectively monitor the therapeutic responses to various treatments.
TCYC and KHW contributed equally.
Contributors All authors contributed to the conception and design, data acquisition, analysis and interpretation of data. All authors contributed to the drafting, revising and final approval of the current study. All authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.