Abstract
purpose. The purpose of this study was to investigate Fourier domain optical coherence tomography (FD OCT) as a noninvasive tool for retinal imaging in the Rs1h-knockout mouse (model for X-linked juvenile retinoschisis).
methods. A prototype spectrometer-based FD OCT system was used in combination with a custom optical beam–scanning platform. Images of the retinas from wild-type and Rs1h-knockout mice were acquired noninvasively with FD OCT with the specimen anesthetized. At the completion of the noninvasive FD OCT imaging, invasive retinal cross-sectional images (histology) were acquired from a nearby region for comparison to the FD OCT images.
results. The retinal layers were identifiable in the FD OCT images, permitting delineation and thickness measurement of the outer nuclear layer (ONL). During FD OCT in vivo imaging of the Rs1h-knockout mouse, holes were observed in the inner nuclear layer (INL), and retinal cell disorganization was observed as a change in the backscattering intensity profile. Comparison of the ONL measurements acquired noninvasively with FD OCT to measurements taken using histology at nearby locations showed a degeneration of roughly 30% of the ONL by the age of 2 months in Rs1h-knockout mice relative to wild-type.
conclusions. FD OCT was demonstrated to be effective for noninvasive imaging of retinal degeneration and observation of retinal holes in Rs1h-knockout mice.
Significant inroads have been made into human retinal imaging by Fourier domain optical coherence tomography (FD OCT).
1 2 3 4 5 6 7 8 As the FD OCT imaging modality is gaining wider acceptance for its diagnostic utility, the investigation of FD OCT as a tool for basic visual science with animal models of retinal degeneration is also generating increasing interest. In contrast to human clinical imaging where only noninvasive imaging modalities can be used, research involving animal models can use invasive histology on enucleated eyes, permitting confocal imaging of immunologic stained sections to enhance the contrast of retinal features. Noninvasive imaging with FD OCT has the distinct advantage of facilitating longitudinal studies on the time course of diseases and has the potential to accelerate basic medical research by significantly reducing the number of animals required. Furthermore, noninvasive imaging with FD OCT has the ability to identify abnormal structural morphology in vivo, removing concerns of potential tissue processing artifacts introduced by histology.
Attempts to investigate retinal degeneration in mice with early time domain OCT systems were hampered by low resolution and slow image acquisition.
9 10 11 Investigation of retinal degeneration with FD OCT in rodents is at an early stage, and to date only wild-type (WT) specimens and simple models of retinal degeneration have been imaged.
12 13 14 Two techniques of focusing the light on the rodent retina have been investigated. In one approach,
13 a collimated beam was used, relying on the refractive elements of the mouse eye to focus the beam on a spot on the retina. In another approach,
12 14 a focused beam was used in conjunction with a contact lens to cancel out the refraction by the cornea. Irrespective of the technique used, the retinal layers were identifiable by their relative scattering intensity, similar to human retinal imaging. An investigation of FD OCT for a time course study of retinal degeneration in the
rd1 mouse was presented by Kim et al.,
14 who used a pulsed femtosecond titanium/sapphire laser providing 110 nm of bandwidth at an 800-nm center. Degeneration was quantified by comparing the ratio of the thickness of the outer retina (including the layers from the outer plexiform layer [OPL] to the outer segment [OS] of the photoreceptor layer) with the thickness of the whole retina (including the layers from the nerve fiber layer [NFL] to the OS).
In this study we built on the existing literature studying FD OCT as a noninvasive imaging modality for the study of retinal degeneration in mice in vivo. The mouse model for X-linked juvenile retinoschisis (RS), a form of genetic retinal degeneration in males, was investigated in this FD OCT research application. These knockout mice are deficient in the
Rs1h gene, the orthologue to the
RS1 gene in humans, and their eyes have been shown through histology to be characterized by holes in the inner nuclear layer (INL),
15 disorganization of the retinal cell layers, and slow progressive degeneration of the photoreceptors. We present a comparison of noninvasive retinal imaging in
Rs1h-knockout (KO) and WT mice acquired with the FD OCT with invasive histologic sections. Qualitatively, disorganization of the retinal morphology associated with the
Rs1h-KO mouse was observed with FD OCT, as well as the holes in the INL characteristic of RS. The application of FD OCT to provide quantitative measurements was also investigated. The metric used to quantify the amount of retinal degeneration was the thickness of the outer nuclear layer (ONL) which thins with increasing age in
Rs1h mice. Two-dimensional maps of the ONL thickness were extracted from the FD OCT volume data acquired on WT and
Rs1h-KO mice. From the standpoint of evaluating a research tool, the FD OCT system used in this study used a cost-effective and portable continuous superluminescent diode (SLED) light source, nominally one tenth the cost of an exotic femtosecond laser, and a custom sample arm constructed from bulk achromatic lenses.
A prototype FD OCT system was constructed for mouse retinal imaging. The standard interferometer topology used in this report, presented in
Figure 1 , consisted of a source, fiber coupler, and custom spectrometer–based detection. The SLED source (Superlum, Moscow, Russia) had a central wavelength of 826 nm and a spectral bandwidth full width at half maximum (FWHM) of 72 nm. The corresponding transform limited axial resolution was nominally 4 μm in tissue. During animal imaging, the optical power output from the source was reduced to 770 μW, the ANSI limit for maximum exposure of the retina to continuous light at this wavelength.
16 The interferometer was constructed from a 2 × 2 fiber coupler (AC Photonics, Santa Clara, CA) with an 80/20 splitting ratio. This configuration was used to provide 20% of the source light to the sample arm, but in the reverse direction, 80% of the collected light was directed to the detector to optimize the optical signal. The reference arm consisted simply of a collimating lens, attenuator, and mirror. Dispersion was corrected purely through numerical techniques,
4 and no additional optics were included in the reference arm. The sample arm consisted of a collimating lens followed by a pair of galvanometer mounted mirrors for raster scanning control of the beam. The beam expander and objective lenses after the scanning mirrors, as indicated in
Figure 1 , were constructed from antireflection-coated achromatic lenses. The calculated spot size given the optical configuration was nominally 13 μm with a depth of focus of 350 μm, designed to provide high-resolution lateral images, while spanning the full retinal thickness. The sample arm optics were mounted on a slit lamp biomicroscope stage for positioning of the beam relative to the mouse eye.
The high-speed spectrometer used was a custom design constructed with a 1200 l/mm transmission diffraction grating. The detector was a 1024-element, high-speed camera (Spyder3, GigE; Dalsa, Waterloo, ON, Canada) with 14-μm2 pixels. The camera can operate at a maximum line rate of 68 kHz, but was typically reduced to 20 kHz for imaging. Data acquisition was performed with custom software written in C++ for rapid frame grabbing, processing, and display of two-dimensional images. Processing performed in real time included resampling of the interferometric data from linearly sampled in wavelength space to linear sampling in wavenumber space, fast Fourier transform (FFT), and image contrast and brightness. Dispersion compensation was also performed by the software up to the third term, but was limited to real-time display of nominally 30 frames per second under imaging conditions of 512 lines per frame.
Although the FD OCT imaging is noninvasive, to keep the mice still during imaging they were anesthetized with an intraperitoneal injection of ketamine and xylazine mixture (0.1 mL per 10 g body weight). After anesthetization, the mouse was placed gently on a heating pad to maintain warmth, and simple manual manipulation was used to rest the head of the mouse in an orientation where the angle of the eye was properly coupled to the optical beam. The pupils were dilated with a topical solution (atropine sulfate 1%). Refraction of light at the cornea was cancelled by placing a flat coverslip generously coated with a generic artificial tear gel over the eye. Alignment of the optical system to the mouse retina required several minutes and was followed by rapid acquisition of data, requiring nominally 5 seconds per volume. During imaging, the software displayed the FD OCT B-scans at ∼30 frames/s. Registration of the location of the B-scan within the two-dimensional surface of the retina was performed by switching the display mode to a high-speed (60-kHz line rate), low-sampling-density area scan, representing a reconstructed fundus type image. All mouse handling adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research, and the experiments were performed under protocols compliant with the Canadian Council on Animal Care with the approval of the University Animal Care Committee at SFU.
The ONL thickness measurements were extracted by postprocessing saved data. A fundus-type image of the mouse retina was reconstructed by using the FD-OCT volumetric data
7 and was used to register the location of the ONL thickness measurement 400 μm from the optic nerve head where the retinal layers were of nominally constant cross section. The ONL boundaries were delineated by manually placing points at the boundary between layers and fitting a low-order polynomial to the points with a commercial mathematical software package (MatLab; The MathWorks, Natick, MA). The ONL thickness was subsequently measured along a line in the image and was calculated perpendicular to the curvature of the retina by using Snell’s law calculations to account for refractive index changes.
17 18 The estimated average refractive index of the retina was
n retina = 1.38. The recorded ONL thickness for a given eye was calculated as the mean of three adjacent depth profile frames to average out human error in delineation of the ONL boundaries. Automated computer segmentation of the depth profiles using advanced image processing algorithms is under development to complement the FD-OCT image acquisition and reduce interobserver variation. After the noninvasive FD OCT measurement, the mice were euthanatized, and the eyes were enucleated for histology.
Immediately after FD OCT measurements (hours), the mice were killed, and the eyes were enucleated and fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.2) for 15 hours. After fixation, the samples were washed and frozen in OCT compound, cut into 12-μm sections and stained with DAPI nuclear stain. The retina sections were visualized with a fluorescence microscope (Axioplan; Carl Zeiss Meditec, Inc., Dublin, CA), and the ONL thickness was measured by counting the number of nuclei at five locations in the DAPI-stained histologic sections.
For this study, we compared measurements of the ONL thickness in Rs1h-KO mice with those of their WT counterparts. The ONL thickness was measured in both eyes of each specimen used. The control group consisted of six WT mice for the FD OCT measurement, corresponding to a sample size of n = 12 eyes. The control group for the DAPI-stained histology control group consisted of five specimens, corresponding to a sample size of 10. Two mice were used for each age group of Rs1h-KO mice investigated, making a sample size of n RS1hXmo = 4 eyes (X represents the age of the specimen in months). Three ages of Rs1h-KO mice were used in this study: 2, 10, and 15 months. The reported ONL thicknesses represent the average and SD of n measurements from each group.
Examples of the fundus reconstruction and depth profile images acquired with the FD OCT are presented alongside DAPI-stained histology in
Figure 3 . The fundus image represents a volumetric data set; a FD-OCT depth scan of information is contained in each horizontal line.
The numerical results of the ONL thickness measurements comparing FD-OCT to DAPI-stained histology are summarized in
Table 1 . The ratio was calculated using the average ONL thickness measurement from each
Rs1h-KO group and dividing it by that of the corresponding control group. Standard error propagation was used based on the SD of the group averages. The measured differences in ONL thickness between
Rs1h-KO mice of neighboring age groups in the time-course study was statistically significant (
P < 0.01, by ANOVA).