Supercontinuum source-based multi-contrast optical coherence tomography for rat retina imaging

: This study proposed an ultrahigh-resolution multi-contrast optical coherence tomography system integrated with fundus photography for in vivo retinal imaging of rodents. A supercontinuum light source was used in the system, providing an axial resolution of less than 3 µm within 1.8 mm (in the tissue). Three types of tissue contrast based on backscattered intensity, phase retardation, and microvasculature at a capillary level can be simultaneously obtained using the proposed system. Pigmented Long-Evans, non-pigmented (albino) Sprague Dawley, and Royal College of Surgeons rats were imaged and compared. In vivo imaging results were validated with histology.


Introduction
Several retinal diseases, such as retinitis pigmentosa (RP) and age-related macular degeneration (AMD), were defined by the degeneration of retina [1,2], and these diseases are not completely understood. Animal eye models [3,4], particularly for mice and rats, are widely used in the preclinical and basic research of ophthalmology because of their short life cycles, structures similar to those of human eye, and various available disease models. Because of noninvasive, in vivo, and high-resolution tomographic imaging capabilities, optical coherence tomography (OCT) has become successful in clinical ophthalmology [5]. However, high axial resolution is required for rodent imaging to distinguish all retinal layers. Therefore, OCT for small animals is still relatively less prevalent in preclinical research.
The new development of broad bandwidth light sources generated ultrahigh-resolution (UHR) OCT [6][7][8][9], which could be favorable for assessing pathological changes in animal eyes. Advances in commercially available supercontinuum light sources [6,7] provide several features, including substantial optical bandwidth that allows UHR imaging, excellent spatial coherence, and high illumination power density, thereby providing higher sensitivity than conventional sources can provide. All of these attributes are essential parameters for UHR OCT for small animals.
Because of label-free and depth-resolved features, OCT angiography, which is a functional extension of OCT, has been widely applied in clinical research. Retinal degeneration is associated with retinal blood circulation because of the high metabolic demand of the neuroretina. Therefore, several clinical researches have investigated retinal diseases, such as RP, AMD, and diabetic retinopathy (DM) [10][11][12], by using OCT angiography. A multitude of functional-extension applications using a supercontinuum source in small animals has increased rapidly (e.g., visualizing microvasculature in mouse ears and monitoring acute stroke in a mouse model using OCT angiography [13,14] and measuring retinal oxygen metabolic rate noninvasively in rat eyes by spectroscopic OCT) [15,16].
Polarization-sensitive OCT (PS-OCT) is a functional extension of OCT, which uses polarization interferometry with broadband light sources to obtain structural and birefringence tomograms or other polarization properties of biological tissues. The phenomenon of birefringence is observed in media containing ordered arrays of anisotropic structures, such as in the retinal nerve fiber layer (RNFL) and sclera or depolarizing phenomenon caused by retinal pigment epithelium (RPE) and choroid. PS-OCT has been used to assess peripapillary sclera and RNFL in vivo in nonhuman primates [17], healthy albino rats [18], and pigmented and nonpigmented rats [19] with axial resolutions of 12, 7.6, and 5.1 µm, respectively. Changes in polarization states may indicate the presence of tissues in early stages of a pathological process, such as damaged RNFL by glaucoma, RPE-related changes in AMD, and hard exudates in DM [18][19][20][21].
The expansion of conventional OCT to multifunctional OCT (i.e., including intensity OCT, OCT angiography, and PS-OCT) enables the simultaneous acquisition of the structure, flow, and phase retardation information of tissues, which provides multiparametric, complementary information on tissue microstructure, blood vessel morphology, and anisotropic structures [22][23][24][25]. For example, Wang et. al [26] proposed multi-contrast OCTbased tractography for reconstructing micrometer-scale fiber pathways in the brain. Moreover, Jones matrix OCT was demonstrated for the Doppler and polarization-sensitive imaging of the posterior eye [27]. Furthermore, Augustin et al. [28] longitudinally observed spontaneous retinal-choroidal neovascularization in a mouse model with a threefold contrast.
This study aimed to develop a supercontinuum source-based multi-contrast OCT for the imaging of the retina degeneration in rats. Compared with previous threefold contrast setups, an all-fiber circular-state (CS) spectral domain (SD) PS-OCT with ultrahigh resolution was proposed. Bulk quarter wave plates (QWPs) were replaced by fiber optics polarization controllers (PCs) to realize a single-mode (SM) fiber optics PS-OCT. With the use of OCT angiography calculation, such a system enables the simultaneous imaging of three intrinsic contrasts: backscattered intensity, phase retardation (PR), and microvasculature at the capillary level. This study performed UHR multi-contrast OCT for Royal College of Surgeons (RCS−/−) rat retina imaging. The RCS rat is an acceptable photoreceptor degeneration model, which is used to determine the safety and efficacy of drugs or therapies used for treating several retinal diseases, and the genetic defect in RCS rats causes the inability of the RPE to phagocytose shed photoreceptor outer segments [29,30]. Moreover, pigmented Long-Evans (LE) and nonpigmented Sprague Dawley (SD) rats were imaged for comparison with RCS rats. The aforementioned in vivo imaging results were validated through histology and fundus photography. Figure 1 presents the schematic of the proposed multi-contrast OCT. A supercontinuum laser (NKT Photonics, Denmark) was used as a light source. The emission spectrum of the light was shaped by optical filtering to achieve a central wavelength and spectral bandwidth of 860 nm and 200 nm, respectively, with a theoretical axial resolution of 1.8 μm (in tissue). The light source was linearly polarized using a fiber optics polarizer. A fiber coupler with a coupling ratio of 90/10 placed beside the polarizer was arranged to split the light power into the two fiber arms of the interferometer, such that 90% was for the reference arm and 10% for the sensing arm.

Setup of multi-contrast OCT
The light reflected from the measured sample and the mirror of the reference arm to PC-1 and PC-2, respectively, was then propagated through the same SM fiber back to an SM where V and H denote the vertical and horizontal polarization channels, respectively. ( ) s R z is the sample reflectivity; the sin( ( )) z δ and cos( ( )) z δ terms are retardation moduli modulated by the birefringence of the measured samples. The amplitude( δ )) of interference signals was used to calculate the structure (Rs) and phase retardation ( δ ) images as prescribed for traditional bulk optics-type CS PS-OCT [32].
Vascular images were then acquired through complex signal calculations in an optical microangiography (OMAG) algorithm [33]. The flow signal can be written as follows: Here, , ( , ) x z is the complex value OCT data of the i-th frame at a lateral position of x and depth z in the vertical and horizontal channels, respectively. N = 5 is the number of frames to be calculated to one flow frame. Vascular contrast was achieved because flow regions and the bulk tissue provided several different values. To reduce the artifacts generated by laser jitter and eye motion, bulk motion compensation and phase compensation methods [34] were used in this study. After the compensation method, the OMAG algorithm could separate the flow tissue from the static tissue. Figure 2 presents an image processing flowchart. Interference signals were acquired using two spectrometers. After resampling, dispersion compensation, and Fourier transform, three contrast images were simultaneously obtained. The structure image was obtained using the following formula: , one structure image was averaged from five frames. Angiography performed using V S and H S , which were then combined together. A retardation image was calculated using To determine the microstructure of the retina, three-dimensional (3D) reconstruction performed using 2000 two-dimensional (2D) OCT images recorded to a 3D data set having a volume of 1 × 1 × 1 mm 3 corresponding to 400 × 400 × 2048 pixels in the X, Y, and Z directions of the retina. Manual retinal layer segmentation was then performed for the visualization of vascular structure through separate projections from different layers of the retina.

Animal model and histological analysis
In this study, the three strains of rats with different eye conditions were used, namely one 8-10-week old pigmented-eye LE rat, two 8-10-week old nonpigmented-eye (albino) SD rat, and two 12-month old hereditary retinal dystrophy rat RCS. All are male rats.
The RCS rat is a natural animal model of autosomal recessive RP. This rat strain harbors homozygous null mutation in gene encoding receptor tyrosine kinase (MERTK), which functions in the phagocytosis process. In the eye, MERTK is expressed only in the RPE. Therefore, the RPE of the RCS rat cannot eliminate the shed photoreceptor outer segment, which is accumulated as debris and induces cell loss and retinal degeneration. At approximately 18 days postnatal, the RCS rat rapidly develops photoreceptor cell loss and progressive degenerated retinas [35,36].
In this study, all animals were fed ad libitum and were kept in ventilation cages (2 rats per cage) with 12/12 h light/dark cycle, optimal temperature, and humidity-controlled conditions. All animals were maintained under anesthesia with an intraperitoneal injection of ketamine (100 mg/kg) + dexmeditomidine hydrochoride (dexdomitor) (1 mg/kg) throughout the experiment. The study was approved by the Institution of Animal Care and Use Committee at National Yang-Ming University. At the end of observation, the rats were sacrificed and enucleated. The eyes were collected and fixed in 4% paraformaldehyde for 4 h at 4°C. The fixed samples were then embedded in paraffin, sectioned, and stained with hematoxylin and eosin for a histopathological analysis. The histological change was assessed by an experienced pathologist. Figure 3 depicts the depth-dependent decay (logarithmic scale) for the mirror at different depths, where the measured signal to noise ratio (SNR) in dB was calculated to be 20 times the base 10 logarithm of the ratio of the A-scan peak height to the mean noise floor. An SNR of more than 40 dB was obtained for all the peaks less than 1.8 mm. The optimal axial resolution measured was 3.1 μm at a depth of 0.2 mm in air, which corresponds to 2.3 μm in the tissue (n = 1.33). For the performance verification of the proposed UHR multi-contrast OCT system for retardation measurement, a Berek's polarization compensator (BPC) placed in front of a mirror was used as a standard sample. For a short-term stability test, the BPC was set at 42° then measurements were continuously obtained for 5 min. Figure 4(a) indicates that the peakto-peak variation value was only approximately 0.5° within 5 min, which is considerably longer than the time required for acquiring a 3D retinal data set. In this study, the acquisition time for each 3D set was 77 s. For a long-term stability test, measurements were obtained for 1 min at intervals of 20 min for 2 h. Figure 4(b) illustrates a slight increase of approximately 2° within 2 h, indicating acceptable system stability. For the accuracy test, the BPC was set at various tilt angles for a retardation set step, increasing from 0 to 180° at an interval of 15°. The measured retardation (Fig. 4(c)) indicated satisfactory linearity with a deviation of 7.9°, which was the largest in the diagram. To validate that the retardation is independent of fastaxis orientation, retardation values were set at 0°, 15°, 30°, 45°, 60°, 75°, and 90° at different axis orientations, namely 0°, 22.5°, 45°, 67.5°, and 90°. The experimental results (Fig. 4(d)) indicate that retardation measurements were independent of fast-axis orientations, thereby validating that PC-1 and PC-2 exhibited equal effects from the bulk optics QWPs. A large deviation of retardation was observed at 90° of set retardation. This deviation may be attributed to the wavelength dependence of birefringence in the SM fiber and BPC [37].  , including approximately 10 A-scans. After laterally averaging Rs(z) within the ROI, a reflectivity signal was obtained as a function of depth as illustrated in Figs. 5(c) and 5(d) for SD and RCS rats, respectively. Therefore, retinal thickness was defined as the thickness from the NFL/GCL layer to the RPE, which was calculated using manually selected peak position; the pixel numbers between each layer were counted, where one pixel corresponded to 1.39 μm distance. Finally, the average thickness value in ten ROIs (i.e., from ten OCT images positioned 2 mm away from an optic disc) was 187.4 in the SD rat. The retinal thickness calculated using OCT images was close to that in histology (Fig. 5(e), with an average thickness of near 177.6 μm, measured with manual selection by Matlab software). Figure 5(b) illustrates retinal degeneration in a 55-week-old RCS rat. A significant reduction in retinal thickness was observed (i.e., only 98.0 μm from the surface to RPE, p < 0.05), which was almost half of that in Fig. 5(a). The OP, ON, and photoreceptor layer (IS, OS) were absent. The same result was observed in histology ( Fig. 5(f), 84.2 μm), but a smaller thickness in histology occurred due to the shrinkage during specimen preprocessing.   Fig. 6(a)) w t is more tortuo D volume of O ximum intensi scanning region f)). Before per e and elimina om ten A-scans of n SD (c), and RCS of (a) and ( (Fig. 7(c)), alidation throu ion of choroid e RCS rat due m RPE to the re a representative blood capillari ies and veins w er that supply yer that supply etinal vasculari The vessels w weak back-refle ere not able to (Figs. 6(h) and en face projecti aphy (a, f) and thre ; d: OP layer) an pth-color coded m L-GCL layer; green e OCT structur ent the extensio rast enhancem gence propertie tart the change whereas the b ugh histology ( and sclera, and e to old age. T etardation valu e retardation si ies. The roden within the NFL nutrients to th y nutrients to th ity within thes within NFL and ected intensity. find any vesse d 6(i)). Figures ion of the healt ee layers of OCT nd RCS rat (g: NF microvascular en-fa n: IP layer, red: OP re images of an on of the conv ment, which is es). In genera e from the scl boundary of sc Fig. 7(g) and 7 d the dysregula To quantitative ue reaches 45° ignal as a func nt retina can b L and GCL, the he inner retina, he outer retina se three layers d GCL in the R . Although the el distribution s 6(e) and 6(j) thy SD rat and angiography of a FL-GCL; h: IP la face projection of t P layer).

Results
n SD rat and an ventional struc s beneficial fo al, the pseudo lera, which is clera was blurr 7(h)), we specu ation of elastic ely evaluate th was measured ction of depth i be separated i e intermediate , and the deep a. Figures 6(b) of the retina i RCS rat (Fig. 6 e IP layer and within IP laye demonstrate th d RCS rat, respe healthy SD rat (b ayer; i: OP layer) the healthy SD rat n RCS rat, resp cture OCT to f or providing a o color represe the same as ry in the RCS ulated that it w city or connect he retardation c d. Figures 7(e) in the SD and R into three capillary capillary -6(d) and in the SD (g)) were OP layer er and OP he depthectively. imaging, whi reflectivity-on SD rat are de illustrated in 8(g), and 8(j) (Fig. 8(c)), l considerably Moreover, be the RPE layer (Fig. 8(e) ontrast OCT f ond that of t ardation, and a tively. The res an LE rat are p D) (Fig. 8(b)) n a healthy L t in the retina of melanin, the (g)) compared from RPE to t m (RCS rat), dation signal as a ), and (f), and their he boundary of the from RPE to the for small anim the convention angiography m sults for an RC presented in F and abnormal LE rat (Fig. 8  a of   s from NFL/GC m in Fig. 8(b)) graphy demon D (Fig. 8(h) Fig. 8(d)). Fur l microvascula ). ptors of the RC er. Figure 7(c) ed stratificatio ion imaging tions of the dis μm (SD), 5.3 Fig. 8(i)) confi ity of intermed the OP layer. n the SD (36.9 olism in the re ): structural image ction (blue: NFL-G rat (c, f, and i); and ness were simi rthermore, dep ar distribution CS rat were deg ) demonstrates on and thinner in Fig. 8

Discussion and conclusion
The expansion of conventional OCT to multi-contrast OCT enables the multiparametric analysis of pathological processes, such as in wet AMD, choroidal neovascularization can lead to macular edema and retinal hemorrhage, and the development of fibrosis is very common and associated with neovascularization [38]. In DM, retinal and choroidal vascular circulations both are altered, such abnormal retinal capillaries leak extracellular lipids; this phenomenon then leads to leakage of hard exudates that depolarize the light. In this study, an in vivo RCS−/− rat with the rapid loss of photoreceptor cells applied is used as a model for RP. Compared with the previous threefold contrast setups, this study proposed to use a supercontinuum source-based CS SD PS-OCT with a resolution of less than 3 µm in the tissue. All SM fiber optic setup ensures minimally stringent requirements on the lens alignment in this scheme, and therefore easily perform a compact and portable system. Moreover, three intrinsic contrasts-backscattered intensity, phase retardation, and microvasculature at the capillary level-can be simultaneously obtained. High resolution for rodent imaging enables the system to distinguish each layer of the retina clearly and to delineate the IS/OS of the photoreceptor layer (such as in Fig. 5(a)), whose degeneration is primarily responsible for reduction in total retinal thickness. The retardation contrast can provide the information of the changes in choroid and sclera, and depolarization properties from melanin. In the RCS rat, uneven retardation change indicates degeneration in choroid and sclera. Fundus photography can rapidly perform 2D en-face inspections, presenting shallow vascularity and vessel branches: however, the underlying microvessels are not visible. Blood vessels in various depth layers of the retina in the SD and LE rats can be observed by depth-color-coded OCT angiography where vascular contrast was achieved from the flow signal in vascular regions. In the RCS rat, the vessel within NFL and GCL was still observed; however, no vascularity was found within IP layer and OP layer. The previous report using fluorescein angiography has demonstrated the patches of hypofluorescence in the RCS rat retina. The severe capillary of the nonperfusion area was identified in the aged RCS rat (older than 6 months) and correlated to devoid capillaries observed through NADPH-d histology [39]. Capillary nonperfusion is caused by insufficient tissue perfusion, which the crucial blood delivery process in the retina. Several attempts have been made to illustrate the role of photoreceptor cells in the retinal vascular degeneration, particularly in diabetic retinopathy and genetic-mediated photoreceptor degeneration [40][41][42][43][44]. A mechanism by which photoreceptor degeneration could mediate retinal vasculature degeneration was explained in animals with opsin-deficiency-induced photoreceptor degeneration [40,42]. These animals have significantly reduced vascular density. Particular opsin and rhodopsin expressions are specific to photoreceptor cells. Therefore, opsin deficiency from the photoreceptor loss may be the probable cause of retinal vascular degeneration in the RCS rats.
In conclusion, UHR multi-contrast OCT system enables the simultaneous acquisition of the microstructure, blood vessel morphology, and anisotropic properties of tissues, which may provide multi-parametric, complementary information in a lesion.