Phase-Decorrelation Optical Coherence Tomography Measurement of Cold-Induced Nuclear Cataract

Abstract

Purpose: The purpose of this work is to determine the sensitivity of phase-decorrelation optical coherence tomography (OCT) to protein aggregation associated with cataracts in the ocular lens, as compared to OCT signal intensity.

Methods: Six fresh porcine globes were held at 4°C until cold cataracts developed. As the globes were re-warmed to ambient temperature, reversing the cold cataract, each lens was imaged repeatedly using a conventional OCT system. Throughout each experiment, the internal temperature of the globe was recorded using a needle-mounted thermocouple. OCT scans were acquired, their temporal fluctuations were analyzed, and the rates of decorrelation were spatially mapped. Both decorrelation and intensity were evaluated as a function of recorded temperature.

Results: Both signal decorrelation and intensity were found to change with lens temperature, a surrogate of protein aggregation. However, the relationship between signal intensity and temperature was not consistent across different samples. In contrast, the relationship between decorrelation and temperature was found to be consistent across samples.

Conclusions: In this study, signal decorrelation was shown to be a more repeatable metric for quantification of crystallin protein aggregation in the ocular lens than OCT intensity-based metrics. Thus, OCT signal decorrelation measurements could enable more detailed and sensitive study of methods to prevent cataract formation.

Translational Relevance: This dynamic light scattering-based approach to early cataract assessment can be implemented on existing clinical OCT systems without hardware additions, so it could quickly become part of a clinical study workflow or an indication for use for a pharmaceutical cataract intervention.

Introduction

Prevalence and Epidemiology of Age-Related Cataracts

Cataracts are exceptionally common, and their prevalence increases with age. Large studies in the United States have found that 17% of the population over 40 years of age have cataracts, and that proportion increases to 68% for those over 80 years of age.¹ Age-related nuclear cataracts are by far the most common type of cataract, accounting for over 80% of all age-related cataracts.² 

Protein Aggregation as Mechanism of Cataract Formation

Age-related nuclear cataracts are the result of microstructural abnormalities in the ocular lens. The lens is composed of α-, β-, and γ-crystallins. The α-crystallins act as chaperone molecules, binding to misfolded γ-crystallins to prevent improper aggregation.³ Unfortunately, proteins in the lens are not regenerated during adulthood. Over time, crystallins are damaged through mechanisms such as oxidative stress, ultraviolet irradiation, and truncation.⁴ As the misfolded γ-crystallins bind to an increasing number of α-crystallins, the α-crystallins are depleted from the lens. Unchecked, protein aggregation increases to the point where the protein aggregates are large enough to scatter light (approximately 1 µm).⁵ 

During cold cataract formation, γ-crystallin precipitates out of the solution with decreasing temperature, undergoing a liquid-liquid phase separation and forming dense liquid droplets.⁶ The nucleus, at the center of the lens, has a higher concentration of γ-crystallin than the surrounding cortex.⁷ Accordingly, the change in scattering is most pronounced in the nucleus.⁸ This increasing scattering (opacity) is a nuclear cataract. 

Clinical Motivation for Early Cataract Detection

The only clinically approved and effective treatment for restoring vision in the case of cataracts is removal of the cataractous lens contents and replacement with an artificial intraocular lens (IOL). This is indicated when visual function is compromised. In early-stage cataracts, visual acuity may be improved through refractive correction, however, this does not stop the progressive development of cataracts.⁹ However, a growing number of pharmacological solutions are being explored that could treat early cataracts and prevent or reduce further progression.¹⁰ Unfortunately, there is a lack of methodologies to accurately diagnose and measure cataracts in early stages. 

A critical need in the development of pharmaceutical cataract interventions is the ability to quantify changes in cataracts over time. There have been a number of sustained efforts to quantify cataract progression. These methods, such as slit-lamp cataract classification,¹¹ optical coherence tomography (OCT)–based backscatter signal intensity quantification,^12,13 or Scheimpflug densitometry,^14,15 primarily rely on static light scattering. These methods generally have reported relatively high variability and lower-than-desired repeatability. This ambiguity creates challenges in determining how accurate a particular method is at assessing protein aggregation within the lens. Previously, benchtop and clinical studies have shown strong correspondence between dynamic light scattering (DLS)-based measurements of protein aggregation (see Methods for further explanation of DLS) and both (1) gold standard testing for protein aggregate size,¹⁶ and (2) known, gradual modulators of protein aggregation, such as age.^17,18 These studies also suggest that DLS measurements are superior in early detection and sensitivity, as compared to static light scattering methods, such as Scheimpflug and OCT imaging, because light scattering intensity only significantly increases once protein aggregates reach at least 1 µm in diameter, whereas DLS is sensitive to changes in particle size down to several nanometers.^18,19 However, there is no clinically available DLS measuring device and there are significant barriers to creating a new ophthalmic device, including financial, regulatory, and garnering clinical acceptance. We propose the use of clinical OCT systems to make DLS measurements of the lens as an indicator of protein aggregation. This approach may share the benefits of specificity and sensitivity offered by DLS while making clinical adoption more likely through capitalizing on available equipment. 

In this study, the cold cataract is intended to serve as an approximation of age-related cataracts. In cold cataract, liquid-liquid phase separation (LLPS) is induced through lowering the temperature of the lens. In one mechanism of age-related cataracts, damage, dimerization, or misfolding of γ-crystallin can raise the threshold temperature LLPS, resulting in LLPS and therefore increased scattering at body temperature.^20,21 

Porcine eyes were chosen for the study because of their accessibility and similarity to human eyes, specifically in the geometry of the ocular lens. The goal of this study was to demonstrate that DLS measurements can be made in the lens using an OCT system, and that DLS measurements may be a more repeatable indicator of cataract-associated changes in the lens than backscatter light intensity-based OCT measurements. 

Methods

Porcine Globe

Porcine globes were obtained from an abattoir (E. R. Boltiantz Co., Ashland, OH, USA) within 6 hours post mortem. All globes were examined and determined to be free of cataracts and other abnormalities. Whole globes were kept moist and chilled for transport, and upon arrival at the research laboratory, globes were immersed in OptiSol-GS (Bausch & Lomb Inc., Bridgewater, NJ, USA), and stored in a refrigerator with an average temperature of approximately 4°C. OptiSol-GS was chosen as storage media because it optimally maintains corneal hydration and transparency, as desired for imaging of the lens. 

Cold Cataract

Cold cataract is a reversible type of cataract (opacity of the lens, in this case, primarily the nucleus of the lens) induced by prolonged exposure to temperatures below the phase separation temperature of lens proteins.^22,23 During cold cataract, γ-crystallin undergoes a phase transition and the nuclear cytoplasm separates into protein-rich and protein-poor phases (LLPS).²³ Cold cataract has been found to also stiffen the lens,²⁴ similar to cataracts which present clinically.²⁵ Upon rewarming, this phase separation reverses. During storage, globes were periodically examined to monitor the development of cold cataract opacity. After 24 to 48 hours of 4°C storage, globes were examined, confirmed to have visually apparent cold cataracts, and used for study immediately after removal from the refrigerator. The imaging experiment is conducted during rewarming and concurrent reversal of the cold cataract. As the cold cataract reverses, the protein-rich particles decrease in size, reducing scattering and opacity.²⁶ Thus, temperature may be considered a surrogate measurement for relative particle size during this rewarming phase. 

Experimental Setup, Recording, and Controls

During imaging (Fig. 1), globes were positioned in a custom-designed, 3D-printed dish. Intraocular pressure (IOP) of each globe was maintained at 15 mm Hg. To maintain IOP, a 21-gauge needle was inserted into the posterior segment through the sclera, and connected to a gravity-fed pressure system with an adjustable-height, 300 mL phosphate-buffered saline reservoir. A digital in-line pressure transducer attached to a continuous pressure monitor (Siemens SC 7000; Siemens, Berlin, Germany) provided real-time pressure readings and the reservoir height was manually adjusted to regulate pressure. Before each experiment, all tubes were checked for air bubbles and the pressure monitor was zeroed to atmospheric pressure. Globes were given at least 2 minutes to equilibrate with the induced 15 mm Hg pressure prior to imaging. A second needle with a type T thermocouple mounted inside the tip of the needle was inserted through the sclera such that the tip was just behind the ocular lens. The thermocouple was protected with a thin layer of elastomer (SYLGARD 184, Dow Inc., Midland, MI, USA). Readings from this thermocouple were recorded at 10 Hz using a thermocouple reader (4-Channel Temperature Meter SDL200; Extech Instruments, FLIR systems, Wilsonville, OR, USA). Prior to each experiment, the thermocouple reading was calibrated by a single-point calibration to ambient temperature, measured by a standard scientific alcohol thermometer. Immediately prior to beginning an imaging sequence, a thin layer of OptiSol-GS was placed on the cornea to prevent corneal dehydration during image acquisition. As the globes remained in air at ambient temperature, the lens temperature slowly increased, and consequently the cold cataract opacity reversed. Images were acquired periodically (approximately 1 M-B scan per minute) until the cold cataract was no longer discernible in the real-time OCT images. 

Figure 1.

 

Imaging the reversal of cold cataract. (A) Diagram of porcine globe experimental system with IOP control and temperature monitor under OCT imaging system. (B) Porcine globe immediately after removal from cold storage, demonstrating nuclear cataract (white arrow). IOP needle (purple arrow) and needle-mounted thermocouple (green arrow). (C) Porcine globe after warming to >15°C, when the nuclear cataract has reversed and is no longer apparent.

Phase Decorrelation OCT

The principle of phase-decorrelation optical coherence tomography (PhD-OCT)²⁷ is based on the theory of DLS. When coherent light illuminates randomly moving particles suspended in fluid, the amplitude and phase of the scattered light field fluctuate in a stable manner in time.²⁸ Since 1998, various methods have been devised which allow OCT systems to make DLS measurements.²⁹ 

In PhD-OCT, an OCT M-B scan is used to collect spatially resolved information on Brownian motion. No additional hardware is required beyond a phase-stable OCT system of sufficient speed to measure the temporal fluctuation. The time-dependent complex-valued autocorrelation, g⁽¹⁾, of this signal follows an exponential decay for a diffusive, non-flowing sample. PhD-OCT fits the initial, linear portion of this curve.  

\begin{equation}{g^{\left( 1 \right)}}\left( {{\rm{\Delta t\;}}} \right) = {e^{ - {\rm{\Gamma \Delta t\;}}}} \approx 1 - {{\rm{\Gamma }}_T}{\rm{\Delta t\;}}\end{equation}

(1)

where Δt is the time lag and Γ_T is the time constant of the exponential decorrelation. In the case of dilute particles in solution undergoing Brownian motion, the Stokes-Einstein equation applies:  

\begin{equation}{{\rm{\Gamma }}_T} = Q\frac{{Tn_T^2}}{{{R_T}{\eta _T}}},\;Q = \frac{{8{\rm{\pi }}{k_b}}}{{3{\lambda ^2}}}\end{equation}

(2)

where T is temperature in Kelvin, n_T is refractive index of the sample, R_T is radius of the protein aggregate particles, assumed to be spherical, and η_T is the viscosity of the sample, k_b is the Boltzmann constant, and λ is the wavelength of the imaging system. Where Q is constant for an imaging setup. This equation can be used in DLS to calculate the hydrodynamic radius of polymer aggregates.³⁰ However, this equation assumes the protein solution to be dilute. In high-concentration solutions, such as the lens cytoplasm (20–50% protein by mass),³¹ interactions between the polymers themselves (polymer-polymer interactions) dominate, rendering the viscosity and radius poorly suited metrics.^32–34 Because the lens cytoplasm has high protein density, −Γ_T is reported in this study without further calculation of particle size. 

In this work, a custom 1310 nm 47 kHz spectral-domain OCT system with resolution of 12.5 µm axially and 20 µm laterally was used. Further details on the spectrometer were presented previously.³⁵ The M-B scans were acquired with the following parameters: 1000 A-lines in each M-scan (0 mm), and 500 M-scans in each B-scan (over a 5 mm extent). To account for noise-induced decorrelation, each decorrelation curve was normalized by the g⁽¹⁾ (Δt = 1 A-line) value. Effectively, this is a signal-to-noise ratio (SNR) correction.^36,37 Additionally, it is assumed that all scatterers in the volume are dynamic (i.e. no scatterers are wholly static) and that all stay within the imaging voxel.³⁸ The initial 10 time-lag points in each decorrelation curve were omitted from further calculation due to decorrelation artifact caused by high-frequency galvanometer scanner jitter. Therefore, the decorrelation slope (−Γ_T, Equation 1) was calculated from the time lags of 10 to 30 A-lines (0.213 to 0.638 ms). To generate cross-sectional decorrelation images of the sample, the slope of this decorrelation line is calculated and reported for each B-scan pixel (px) location. For calculating average values and curves, a 100 px by 100 px region of interest (ROI) was selected manually from the center of each nuclear cataract decorrelation image. The data workflow for the PhD-OCT processing is demonstrated graphically in Figure 2. 

Figure 2.

 

Graphic illustration of the PhD-OCT workflow. M-B scans are acquired and processed in the conventional manner. The resulting complex-valued M-B data, where each frame represents a single depth scan over time, are analyzed for temporal fluctuations. For a given pixel depth, this is accomplished by taking vertical spatial windows of 6 pixels from within each M-scan, and computing the average temporal decorrelation curve. This curve is normalized to the t = 2 point. Then, the initial slope of this curve is calculated. The slope of the curve, −Γ_T, is then plotted for each point in the cross-sectional B-scan, resulting in the PhD-OCT image of the sample.

Results

As each globe warmed in ambient temperature, a gradual increase in temperature was measured by the thermocouple (Fig. 3). Correspondingly, the nuclear opacity was reversed, as evident in the OCT intensity images (see Fig. 3) and visual observation (see Fig. 1). The PhD-OCT images show a gradual increase in magnitude of the decorrelation slope which occurs relatively evenly across the nucleus (see Fig. 3). Although intensity OCT and PhD-OCT mapping appear to correspond, closer inspection shows that the correspondence is not strict. 

Figure 3.

 

Cold cataract reversal. (Top panel) Temperature over time as the lens warms, dotted lines correspond to image acquisitions, shown below (middle row) OCT intensity images of a lens during warming. Images shown are taken from the M-B scans and magnitude-averaged across the M-scan direction. Color bar units are arbitrary (a.u.) but consistent across the images. (Bottom row) Corresponding PhD-OCT images of the lens nucleus during warming. Field of view is 3.9 mm wide (325 px) × 0.6 mm high (130 px). PhD-OCT images were filtered with a 3 × 3 median filter. Units of color bar are s⁻¹.

Within each image set, a 100 px by 100 px ROI was selected manually from the approximate center of the nucleus. Then, average decorrelation curve was calculated from this ROI to represent the state of the nucleus at a given temperature. These curves for one example lens, at all measured temperatures, are shown in Figure 4. There is a significant increase in the rate of decorrelation with increasing temperature. 

Figure 4.

 

Temporal decorrelation curves as a function of temperature . Decorrelation curves, averaged over a 100 px by 100 px ROI approximately centered in the nucleus of an example lens, displayed over a range of temperatures as the cold cataract protein aggregation reverses. Curves have been normalized such that the first data-point of each curve is 1. Note that there is an instrumentation artifact between 0 ms and 0.2 ms.

The results from all 6 imaged lenses are shown in Figures 5A and 5B. Decorrelation slope and intensity values averaged from the central ROI are presented as a function of temperature. For each individual sample, there is good correlation between temperature and both decorrelation and intensity (r² = 0.944 ± 0.095 and r² = 0.950 ± 0.057, respectively). However, the relationship between intensity and temperature is not consistent across samples. This variability results in a poor overall r² value of 0.019 for intensity. In contrast, the relationship between decorrelation and temperature is fairly consistent across samples, resulting in an overall r² value of 0.880. 

Figure 5.

 

Comparing intensity and decorrelation to temperature. (A) Decorrelation slope versus temperature and (B) intensity (arbitrary units) versus temperature from an ROI (100 px by 100 px) centered in the nucleus of each lens. Whereas, for each individual sample, decorrelation and intensity correlate equally well with temperature (average r² values of 0.944 ± 0.095 and 0.980 ± 0.057, respectively; P = 0.9), when considering all samples together decorrelation exhibits a more consistent relationship to temperature than intensity does. (C, D) The same data as in A and B are plotted with axes reversed. These plots highlight the aggregation prediction error in the case that either C) decorrelation slope or D intensity is taken as the independent measure and then used to predict temperature, a surrogate measure of protein aggregation.

Figures 5C and 5D illustrate the challenge of using image intensity to predict temperature. The same data are shown as in Figures 5A and 5B, however, with x and y axes reversed to place more emphasis on the sensitivity of each measurement to temperature, which here is taken to be a surrogate measurement of particle size – the characteristic of clinical interest. Whereas a given decorrelation value corresponds to a 90% prediction interval of 2.2°C, a given intensity value corresponds to a 90% prediction interval of 6.2°C. 

Discussion

These results suggest that PhD-OCT is sensitive to protein aggregation in the lens, as induced by the cold cataract phenomenon. Additionally, it appears that PhD-OCT may be a more reliable measurement of protein aggregation in the lens than traditional backscatter-intensity OCT imaging. Historically, backscatter-based methods of quantifying early-stage cataract-associated protein aggregation have suffered from lower-than-desired sensitivity and specificity.^12,13 This is likely because backscatter intensity is easily confounded by other factors, such as reflectance or absorbance of tissues and structures anterior to the lens and instrumentation issues. By isolating the dynamic component of the backscattering, DLS-based techniques are able to reject measurement noise which arises from variations in overall light intensity or coupling. It is possible that changes within the cornea, as it changes temperature, may have contributed to the decorrelation signal of the lens. Future work will use a more localized model of cataract formation. 

In addition, whereas the ex vivo porcine cold cataract model is a readily accessible cataract model, it does not recapitulate all aspects of human age-associated nuclear cataracts. Future studies will use human lenses and larger sample sizes. It was found that, despite efforts to ensure even cooling of all globes, the globes developed cataracts at different rates while held at 4°C. For this work, only globes which developed visually apparent cataract were selected for the experiment. The difference in cataract formation may be attributable to different ages at the time of death, or differences in the quantity or character of the γ-crystallin naturally present in each lens. This difference in cataract development means that it is possible that temperature is not a perfect surrogate for protein aggregation. Additionally, due to the nature of the experimental workflow, some globes were held at 4°C for longer than other globes. This difference could similarly explain some variability in the relationship between physical properties of the lens and temperature. However, because of the consistency of the relationship between decorrelation and temperature, it is likely that these differences are relatively small contributors to the overall effect of cold cataract reversal, once a cold cataract has formed. 

Thus, future work studying the development and prevention of cataracts may find more repeatable results if a dynamic light scattering measurement, such as PhD-OCT, is used rather than a traditional static light scattering metric, such as Scheimpflug densitometry or OCT signal intensity quantification. 

The development of a more precise, sensitive, and accessible cataract quantification method would have compounding benefits. First, a more sensitive method could decrease the time needed to conduct clinical trials proving effectiveness of pharmaceutical therapies, as smaller changes could be reliably detected. Second, it would enable the study of cataracts in their early stages, where intervention may be more effective. Third, if a pharmaceutical intervention is found to be successful and entered into widespread use, a robust method of assessing cataract formation and indicating treatment would be a prerequisite. 

Acknowledgments

The authors thank E.R. Boliantz Co. (Ashland, Ohio, USA) for providing whole porcine globes for research use. 

Supported by NIH R01EY028667, R01HL126747, T32EB007509, T32EY007157, an Unrestricted Grant Award from Research to Prevent Blindness to the Department of Ophthalmology Cole Eye Institute (RPB1508DM), Foundation Fighting Blindness Center Grant to the Cole Eye Institute (CCMM08120584CCF), NEI/NIH P30 Core Center Grant (IP30EY025585) While the research reported in this presentation was supported by the NIH, the content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. 

Disclosure: B.J. Blackburn, Alcon (P); M.T. McPheeters, None; M.W. Jenkins, Alcon (P); W.J. Dupps Jr, Alcon (P, C); A.M. Rollins, Alcon (P)