Clinical Application of Ocular Imaging

Abstract

The broadening frontier of technology used in ocular imaging is continuously affecting the landscape of clinical eye care. With each wave of enhanced imaging modalities, the field faces the difficulties of optimally incorporating these devices into the clinic. Ocular imaging devices have been widely incorporated into clinical management after their diagnostic capabilities have been documented in a wide range of ocular disease. In this review we are presenting the main commercially available devices for imaging of the posterior segment of the eye.

In the past two decades eye care has seen the emergence of three major technologies for diagnostic imaging of the posterior segment of the eye: Scanning Laser Polarimetry (SLP), Confocal Scanning Laser Ophthalmoscopy (CSLO), and Optical Coherence Tomography (OCT). All three devices provide real-time, non-invasive and high-resolution images of the eye. These devices went through the challenging process of translating technological advances into meaningful, robust and validated applications that are clinically useful and have become an integral part of clinical eye care management in detecting and monitoring various ocular pathologies. While each deserves its own separate consideration, this manuscript will serve as an overview of the primary state-of-the-art technologies found in the clinic for imaging the posterior segment of the eye and providing quantitative measurements of the ocular structures. The manuscript will first describe each technology, followed by their applications in monitoring and diagnosing ocular diseases.

Scanning Laser Polarimetry

Scanning Laser Polarimetry (SLP) determines the thickness of nerve tissue by examining the birefringence of polarized light as it is reflected from the eye. The size, orientation, and more specifically the microtubular structure of the nerve fibers cause birefringence of incident light.¹ The projected polarized light is subject to reflected phase delay (retardation) that provides a means to estimate tissue thickness.² Because other ocular structures in addition to the retina exhibit birefringence, current commercial devices examine and compensate for corneal and other confounding birefringence by using the radial birefringence of Henle’s fiber layer in the macula as a reference. This approach substantially reduced the occurrence of atypical retardation pattern that hampered the images at earlier version of this technology. The most recent commercially available SLP device, GDx PRO (Carl Zeiss Meditec, Dublin, CA), uses a diode laser with a wavelength of 785 nm, and acquires data from 40° × 20° area of retina. Current scanning pattern starts with scanning of the macular region to determine the proper eye-specific corneal compensation properties, followed by scanning centered on the optic nerve head. An annulus centered on the optic nerve head is sampled from which the retinal nerve fiber thickness is automatically quantified and reported, both as a thickness map and by the TSNIT thickness measurement and profile (Figure 1). The measured thickness is compared to normative thickness values and reported as a deviation map where areas with either statistically significant thinning or thickening of the RNFL are highlighted. The machine also provides the Nerve fiber index (NFI), which is generated by an artificial intelligence algorithm that takes into account multiple variables and reports a value that has been shown to correspond to disease status.^3–6

Figure 1.

Scanning laser polarimetry printout with the GDx PRO; the left eye has a near normal Retinal Fiber Layer Map appearance as evidenced by the thickening (red color) superior and inferior to the optic nerve with a small and early localized thinning at the inferior temporal aspect of the disc noticeable in the Deviation Map. The right eye exhibits extensive thinning in the superior and inferior regions, marked by red and yellow indicators in the deviation map. The center table contains the quantitative measurements with the color coded comparison to normative database. Figures 1–3 were all acquired from the same subject using different imaging devices.

Confocal Scanning Laser Ophthalmoscopy

Confocal Scanning laser ophthalmoscopy (CSLO) is a confocal microscopy technique with the advantage of having high transverse resolution. CSLO focuses a beam onto tissue and filters reflecting light from outside the focal point by use of a confocal pinhole situated in front of a photodetector at a focal conjugate. The commercially available CSLO device, Heidelberg Retina Tomography III (HRT III; Heidelberg Engineering, Heidelberg, Germany) uses a diode laser beam with a wavelength of 670 nm and captures a series of evenly spaced 16–64 sequential 2-dimensional frames each covering an area of 15° × 15°. The optical transverse resolution of the HRT is 10μm and the axial resolution is 300μm. The initial scans are focused anterior to the retina, and the last scans are posterior to the bottom of the optic cup and the depth of optic nerve head dictates the number of frames. For each focal plane the HRT acquires 384 × 384 samplings which records a reflection intensity value at each of the particular (x, y) coordinates of the focal plane. For each (x, y) point, the machine determines the surface location in the z-axis derived from the center of gravity of the reflected light along the frames. Three sets of scans are acquired in a rapid succession and aligned and the average of the three scans is used for the analysis. The operator is required to trace the optic nerve head margin and accordingly the device automatically defines a reference plane to differentiate between the cup and the neuroretinal rim structures. The acquired scans are summarized as a reflectance image resembling the clinical view of the optic nerve head region along with topographic map overlaid with the cup and disc boundaries (Figure 2). Additionally, the machine provides numerous parameters which quantify the optic nerve head region. HRT III also provides another method for analyzing the optic nerve head structure without the need of manual delineation of the optic disc margin. This method, the Glaucoma Probability Score, is based on fitting a predefined model of the optic nerve head region and provides global and sectoral quantification of the region.⁷

Figure 2.

Confocal scanning laser ophthalmoscopy printout of the HRTIII with quantitative parameters shown in the center. The right eye shows signs of glaucomatous damage as highlighted by the red and yellow warning markings on the reflectance image shown in the center. The topography image (uppermost) demonstrates enlarged optic nerve head cupping marked by the red color. The left eye falls within the normal limits.

Optical Coherence Tomography

Optical Coherence Tomography (OCT) uses low-coherence interferometry, where backscattered light is interfered with a reference beam to create an axial scan (A-scan) of depth resolved tissue reflectivity. Consecutive beams are scanned transversely to recreate a cross-sectional slice of scattering media (B-scan).⁸ Initially, OCT was conducted in the time domain (TD-OCT), where a reference mirror was moved to acquire a depth reflectivity profile for a particular locale in the tissue. The physical limitation of a moving reference mirror restricts the speed of TD-OCT systems to ~400 A-scans/second. In order to limit the deleterious effect of eye movements, corneal dryness, and blinking, the available scanning patterns were limited to a circumpapillary scan or six radial slices in a spoke configuration centered on either the macula, or the optic nerve head. More recently, spectral domain OCT (SD-OCT) has been introduced eliminating the need of physically moving mirror, the device samples broad spectral information for a particular location in the tissue. By Fourier transforming the power spectra it is able to recover tissue reflectivity information. SD-OCT uses a spectrometer and charge-coupled-device (CCD) camera to separate and detect the spectrally resolved signal.⁹ With the dramatic increase in speed and axial resolution afforded by SD-OCT over TD-OCT a full three dimensional volumetric imaging reconstructed from rapid raster scan patterns can be obtained.¹⁰

Several SD-OCT commercial devices are currently available, each offering their own distinct advantages. Most operate at a speed around 27,000 A-scans/second, with a current maximum of 52,000 A-scans/second (SOCT Copernicus; Optopol, Miami, FL). Axial resolutions range from 3μm (SOCT Copernicus), 3.9μm (Spectralis; Heidelberg Engineering, Heidelberg, Germany; Figure 3) and 4μm (Bioptigen SDOIS; Bioptigen, Durham, NC) to 6μm (Spectral OCT SLO, OPKO Health, Miami, FL). Cirrus HD-OCT (Carl Zeiss Meditec, Dublin, CA; Figure 4), RTVue-100 (Optovue, Fremont, CA; Figure 5), and SOCT Copernicus all include progression analysis software for monitoring pathologies. The Bioptigen device is more geared towards research and permits user control over the scanning protocol. Some devices incorporate features of other imaging modalities: the 3D-OCT 2000 device (Topcon, Tokyo, Japan; Figure 6) has a built in high-resolution fundus camera for improved registration. The OPKO device includes an SLO image and microperimetry. The Spectralis device includes eye-tracking, and SLO coupled with fluorescein angiography, and autofluorescence in addition to the OCT system. The autofluorescence uses a 488 nm wavelength light source to illuminate the retina, inducing fluorescence without injecting dye. This highlight structures such as drusens and improve the ability of detecting these structures. It should be noted that both fluorescein angiography and autofluorescence lack the quantitative analysis characteristic of the other modes but are clinically very useful for visualizing retinal pathologies.

Figure 3.

Optic nerve head (ONH) and retinal nerve fiber layer (RNFL) printout from Cirrus SD-OCT. The left eye shows normal RNFL Thickness Map with thickening adjacent to the ONH poles. The right eye exhibits extensive thinning of the RNFL in the superior and inferior regions in the RNFL Thickness Map. The thinning is noticeable by the large area of significant thinning at the RNFL Deviation Map. Quantitative parameters of the ONH and RNFL appear at the top center with color coded comparison to normative dataset. Quadrant and 1 clock-hour RNFL thickness and RNFL thickness profiles are provided at the center bottom.

Figure 4.

Cross sectional (B-scan) image taken with the Spectralis. The green line in the fundus image on the left represents the placement of the OCT scan extending through the entire macula and most of the optic nerve head. The layered pattern of the retina is clearly evident in the image.

Figure 5.

Ganglion cell complex (GCC) thickness and deviation map for a progressing glaucomatous eye taken with RTVue. The top row contains the overall thickness map, progressing longitudinally from left to right with thickness values indicated by the adjacent color bar. The middle row shows a deviation of thickness from normal. The significance map at the bottom highlights regions of statistically significant thinning of the GCC. The defect starts as a localized region temporally, and over time progresses nasally occupying most of the superior region, and eventually extending inferiorly.

Figure 6.

Three-dimensional reconstruction of the optic nerve head morphology by the Topcon 3D-OCT 2000. Cross-section maps occur at the margin of the cube.

The parameters provided vary among the devices though most report the retinal nerve fiber layer thickness globally and in sectors. Macula scans are typically quantified by the total retinal thickness or volume and some devices provide further segmentation of the macular layers (Cirrus and RTVue).

Imaging Pathologies

The imaging technologies outlined above provide a detailed view and structural quantitative measures at the posterior segment of the eye. Considering their attributes, the most prominent use of the ophthalmic imaging devices is for retinal diseases and glaucoma. The enhanced visualization has been shown to be of important diagnostic value in diabetic macular edema, age related macular degeneration (AMD; Figure 7), macular hole (Figure 8), vitreomacular traction, epiretinal membrane, and others.^11–15 Hard drusen, a hallmark of AMD, appears as highly reflective deposits underneath the retinal pigment epithelium layer, in OCT scans. The number and size of the drusen can be quantified by OCT and has been shown to be associated with the risk of AMD.^16–19 Ocular imaging devices have been used for grading macular holes and accordingly dictate the preferred surgical approach.^20–23 Similarly, the ability to visualize the epiretinal membrane and vitreomacular traction strand assist in determining the surgical approach and treatment outcome.^24–27 The devices have also been valuable in assessing central serous chorioretinopathy without the need of invasive testing such as fluorescein angiography.¹¹ In some instances the break of the retinal pigment epithelium through which the leakage occur could be detected by OCT.²⁸ Consecutive imaging allows tracing the longitudinal dynamic of the serous collection and treatment response. The ability to quantify the retinal thickness is of use in assessing pathologies such as macular edema and the response to treatment.^{29, 30} These examples illustrate some of the clinical uses of the imaging devices from a wide array of indications. Recently, an automated method of classifying macular hole, macular edema, and AMD, has been shown to successfully identify these macular pathologies.³¹

Figure 7.

High definition raster scan printout of age-related macular degeneration from the Cirrus HD-OCT. The scan includes 5 linear scans (top left) with the center scan magnified at the bottom. Marked elevation of the foveal region is evident with irregularity at the retinal pigment epithelium photoreceptor layer.

Figure 8.

High definition raster scan printout of macular hole from the Cirrus HD-OCT. Full thickness hole with intra-retinal edema at the hole margin occurring as the black spaces are clearly visible.

Ocular imaging devices have become pivotal in clinical glaucoma assessment due to their ability of obtaining micron scale measurements. Glaucoma is a slow progressing and irreversible neurodegenerative disease that remains mostly asymptomatic until late stages. The disease manifests as the thinning of retinal layers, as well as cupping of the ONH. The devices have been shown to provide highly reproducible measurements of the retinal nerve fiber layer and optic nerve head structures.^32–42 These measurements can quantify the thinning of the retinal nerve fiber layer, the typical enlargement of the optic disc cup and the thinning of the neuroretinal rim. The improved segmentation ability of recent iteration of the OCT allows the quantification of the macular ganglion cell layer that has been shown to be an alternative sampling location for glaucoma diagnosis.^43–46 Numerous studies have demonstrated good discrimination ability of the devices between healthy and glaucomatous eyes.^47–53 Using cross-sectional data it has been shown that in early stages of glaucoma, thinning of the retinal nerve fiber layer thickness, as measured by OCT, occurs without changes in visual field as measured by current technology.⁵⁴ After reaching a threshold level there is a decline by both structure and function that might indicate that structural changes precede functional changes in some glaucoma patients further emphasizing the benefit of ocular imaging devices. Furthermore, baseline HRT and GDx measurements have been demonstrated to predict future development of glaucomatous visual field damage.^{4, 55–57} Taken together, the important role of ophthalmic imaging devices in glaucoma assessment has been widely accepted.

Longitudinal Studies

A focal point of clinical research is understanding the trajectory a pathology takes in an individual subject. One of the most powerful means of examining this is a longitudinal study, where subjects are followed over an extended period of time. However, longitudinal studies of clinical subjects are laced with difficulty. Diseases can be slow progressing, and the technology involved in data acquisition is imperfect and constantly changing. Comparing scans longitudinally requires registration of spatially coincident information collected at different time points. The great challenge of rapidly evolving technology is maintaining congruence by finding ways to translate between generations so that past data can be still used. Furthermore, numerous manufacturers are currently available for OCT, yet no standard of scan patterns or analysis were established and therefore measurements obtained with one OCT instrument are not interchangeable with measurements obtained from an instrument manufactured by another company. At the time of writing this review there is no universal reader for OCT that can be used with all available devices.

In many retinal diseases, the ability to obtain consecutive highly detailed visualization over time is sufficient for qualitative assessment of the disease process. For example, the disappearance of retinal puckers after peeling of epiretinal membrane will be evaluated by visualizing the images. However, in many instances, precise quantification of the changes occurring along the disease process is desirable. Due to the high reproducibility of the measurements obtained by optical imaging devices, any change exceeding the inherent measurement variability might be attributed to the disease process. HRT defines the measurements variability for each eye by assessing the difference between several baseline images and accordingly sets the limit for any follow-up image.⁵⁸ Other technologies use the variability as recorded in population based studies. This method is often described as an event analysis and it is mainly useful for situations where changes are expected to occur in a stepwise manner or in acute events. Another approach, called trend analysis, is based on computing the slope of change over time either by a regression analysis or Bayesian analysis.⁵⁹ A slope that statistically significantly exceeds the rate of zero change is marked or when the rate of progression exceeds the normal age related rate of change. This method is mostly suitable for situations where the disease is progressing in a homogenous or continuous pattern. A few studies have shown that the ocular imaging devices are capable of detecting changes over time in healthy and glaucomatous subjects.^60–62 However, the correspondence between progression as detected by visual field and imaging devices in various glaucoma studies was limited, as was the agreement on progression between the various devices.^61–63 These results might be explained either by the different time scale for structure and function progression or by the different physical properties of the various devices that might be suitable for disease detection in different stages of the disease. Further studies with larger cohorts and longer period of follow-up are needed to fully elucidate the role of imaging devices in longitudinal assessment in glaucoma.

CONCLUSIONS

The current forefront of technological advancement in ocular imaging plays a major role in clinical assessment and monitoring of the posterior segment of the eye. Clinical studies evaluating current commercial technology have demonstrated the ability to proficiently detect disease. However, characterizing the trajectory of a disease felt by individuals remains incomplete. The rapid evolution of the imaging devices hold promise for further strengthening of their roles in clinical practice.