Open Access
Add to BibSonomy
Abstract
Three-dimensional (3D) imaging of the human retina is instrumental in vision science and ophthalmology. While interferometric retinal imaging is well established by optical coherence tomography (OCT), non-interferometric volumetric imaging in the human retina has been challenging up to date. Here, we report confocal oblique scanning laser ophthalmoscopy (CoSLO) to fill that void and harness non-interferometric optical contrast in 3D. CoSLO decouples the illumination and detection by utilizing oblique laser scanning and oblique imaging to achieve ∼4x better axial resolution than conventional SLO. By combining remote focusing, CoSLO permits the acquisition of depth signals in parallel and over a large field of view. Confocal gating is introduced by a linear sensor array to improve the contrast and resolution. For the first time, we reported non-interferometric 3D human retinal imaging with >20° viewing angle, and revealed detailed features in the inner, outer retina, and choroid. CoSLO shows potential to be another useful technique by offering 3D non-interferometric contrasts.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Imaging the human retina is unique in its ability to create 3D microscopic images through the clear ocular media and is significant for its translational values. To date, volumetric imaging in the living human retina has almost exclusively been performed by optical coherence tomography (OCT) based on low-coherence interferometry. However, there are important optical contrasts that are not directly detectable by interferometry because the detected signal is incoherent with the excitation. For example, fluorescence as used in fluorescein angiography (FA) and indocyanine green angiography (ICGA) are non-interferometric signals [1], as well as autofluorescence from photoreceptors, the retinal pigmented epithelium (RPE) and choroid contributed by the molecular composition of bisretinoids, lipofuscin, and melanin [2–6]. In addition, by detecting a portion of the transmitted light, multiple scattering and differential phase contrast were reported to enhance the contrast for transparent cells [7–11]. These scattering contrast can be challenging for interferometric imaging method that mainly detects back-scattered light.
While non-interferometric imaging methods exist, 3D retinal imaging for non-interferometric contrast in the human retina is still challenging. Fundus photography can give easy access for imaging both incoherent scattering and fluorescence [1,12,13]. However, it doesn’t provide tomographic view. Scanning laser ophthalmoscopy (SLO) and confocal SLO (cSLO) use a scanning laser [14,15] to improve the resolution and contrast. To reduce the aberration of the eye, they underfill the pupil, equivalently limit the numerical aperture (NA), and the axial resolution is in the order of several hundred microns [16,17]. The topography around the optic nerve head can be produced by SLO, yet the depth resolution is insufficient to produce tomographic retinal images. Adaptive optics SLO (AOSLO) pushes the resolution limit by fully utilizing the pupil size and correcting wavefront aberration [18]. Diffraction-limited resolution can be achieved in the order of 1 micron in the lateral plane and 30–40 microns in depth [19]. By acquiring several en-face retinal images, tomographic retinal cross-sections can be generated by an optical incoherence tomography (OIT) [20]. As the OIT generally belongs to AOSLO family, adaptive optics is introduced to achieve diffraction-limited resolution. While SLO, cSLO, and AOSLO may compile a volumetric dataset by z-stacking as confocal microscopy [19,21–26], the practical challenge is those sequentially acquired z-stacks might be affected by eye motion during long acquisition. The data throughput by the point scanning approach also limits the imaging speed. For example, to generate high-definition depth cross-sections, it takes several minutes for the OIT to acquire enough en-face images.
To increase data throughput, line scanning ophthalmoscopes with or without AO have been developed in which an excitation line is projected on the retina, and a linear detector is used to record the line image [27]. The increased speed may provide opportunities to compile a quasi-static volume by rapid Z-stacking. On the other hand, z-stacking can be avoided by adopting oblique line illumination [28,29], where oblique cross sections in the depth can be recorded simultaneously. However, without proper 3D reimaging and descanning of the oblique images, 3D data with consistent resolution throughout the volume is not feasible. Due to the above technical challenges, non-interferometric volumetric imaging in the human retina is sparsely presented in the literature.
To facilitate 3D non-interferometric retinal imaging in the human eye, we report a confocal oblique SLO (CoSLO) to collect depth signals in parallel so that z-stacking is avoided. We adopt the idea of oblique illumination and reengineered the concept of single objective light sheet microscopy (SOLSM) [30,31] with human ocular optics, where the ocular lens serves as a low NA objective lens. A demagnifying optical design was implemented to overcome the limitation of small NA for the dilated human eye while still achieving tomographic imaging with practical light collection efficiency [32]. A confocal detection was implemented to reject the diffusive light and improve the image contrast. We de-scanned the oblique laser line onto a linear detector array, so that the A-line depth signal is acquired simultaneously. Besides enabling 3D imaging ability in SLO, CoSLO improves the axial resolution of SLO significantly by decoupling illumination and detection, and utilizing partial pupil apertures for both illumination and detection (Fig. 1(a)).
Fig. 1. The concept of non-interferometric volumetric imaging in the eye by using CoSLO. (a) Axial resolution comparison between coaxial detection and off-axial detection with the same light collection angle. The full width at half maximum (FWHM) of the coaxial detection is much larger than that of off-axial detection. (b) The framework of CoSLO; (c) The virtual confocal array in the intermediate image space (IIS). (d) The acquisition of A-line and B-scan.
For the first time, we demonstrated non-interferometric volumetric imaging over a large FOV of 6 × 6 mm2 area (i.e., ∼20° viewing angle) in the living human retina. Without the need for z-stacking, the entire volumetric data can be acquired within 5 seconds. The axial resolution is ∼45 µm, 4-times better than that of SLO [16,17] and approaching to AOSLO [19]. Different FOVs centered at the macular region and optic disc were demonstrated. We also showed that spectral contrast could be achieved with minimal change of the optics.
2. Materials and methods
2.1 Concept of the CoSLO
The concept of the CoSLO is illustrated in Fig. 1. Human ocular optics is used for both oblique illumination and detection. Instead of co-axial excitation and detection in conventional SLO scanning scheme, the illumination (pink color) in CoSLO is off-axis and creates a scanning oblique A-line (Fig. 1(b)). The non-interferometric signals (blue color) are de-scanned such that they can be mapped onto a stationary A-line image, which is also at an angle with respect to the optical axis. A tilted remote focusing system then images the stationary A-line image with a linear detector to record the depth-resolved signals. At the same time, confocal gating is implemented by the pixel in the linear detector, effectively posing a virtual confocal array (Fig. 1(c)). By scanning the oblique illumination line along the fast axis in X direction (A-lines), a B-scan is created in the retina (Fig. 1(d)). Because the depth-resolved non-interferometric A-line is acquired in parallel, no refocusing and z-stacking are needed as compared to the point scanning scheme or the horizontal line scanning [15,27]. Therefore, the data throughput is dramatically improved, facilitating non-interferometric volumetric retinal imaging.
2.2 System description
The 3D model and the schematic of the proposed CoSLO setup are shown in Fig. 2. The light sources (LS) could be switched between near-infrared (NIR, SLD830S-A20W, Thorlabs) and visible light (SuperK Extreme OCT, NKT photonics) with other components unchanged. The NIR channel had a center wavelength of 830 nm with 55 nm bandwidth, and the visible channel had a center wavelength of 600 nm with 35 nm bandwidth. Both of them were broadband low coherence sources, with micron-level coherent length. The powers on the cornea for the NIR and visible light channels were 200 µW and 60 µW, respectively.
Fig. 2. The 3D model and the schematic of the CoSLO system. (a) The 3D model of the CoSLO setup. (b) The system schematic of CoSLO: LS, light source; L: lens; M: mirror; GM: galvanometer mirror; IIS: intermediate image space; OL: objective lens; BS: beam splitter.
After the light was collimated by (L4: f = 10 mm), a pair of mirrors (PF10-03-P01, Thorlabs) was used to adjust the attitude and orientation of the beam. A right-angle mirror (MRA05-F01, Thorlabs) was moved by a translational stage so that the offset of the beam with respect to the optical axis could be adjusted. To reduce the number of relay lenses and the aberrations, a virtually conjugated galvanometer pair (VCGP) consisting of three galvanometer mirrors (GM1& GM2: QS-20X-AG, Thorlabs; GM3: QS-20Y-AG, Thorlabs) was used as the slow-axis mirror and the fast-axis mirror [33,34]. GM2 and GM3 scan the beam along the same axis (slow axis) so that GM3 is conjugated with on GM1, which scans in orthogonal axis (fast axis). The exit pupil of the VCGP was then mapped onto the human eye’s pupil by a relay lens group (L3: Two ACT508-200-A, Thorlabs; L1: Two ACT508-100-A, Thorlabs). The scanning protocol is shown in [Supplementary Materials, Fig. S1].
As the dilated human pupil usually has a diameter of ∼7 mm, the offset of the illumination is adjusted to ∼3 mm away from the center of the pupil to maximize the oblique angle of the excitation line in the retina. The beam radius was ∼0.5 mm when projected onto the cornea of the eye. The approximated focal length for the human eye is ∼16 mm according to its optical power of 62.3 D [35]. Given the offset of the beam and the focal length of the eye, the oblique angle of the excitation line created on the retina is estimated to be ∼10°. For the NIR light, the beam waist and the Rayleigh range of the oblique Gaussian beam focused on the retina were estimated to be ∼ 8.5 µm and ∼270 µm, respectively. As for the visible light, the beam waist and Rayleigh range were approximately 6 µm and 196 µm, respectively. The beam splitter (BS: BP208, Thorlabs) and lens (L2) were used for a pupil camera (C1) so that the position of the eye could be adjusted accordingly by a motorized chin rest. The photograph of the device construct is in supplemental Fig. S2.
The scattered light from the retina was directed through the same relay lens group into the VCGP unit, where the oblique scanning laser was de-scanned to be stationary in the intermediate imaging space (IIS) after OL1 (Olympus UplanSApo 20X, 0.75NA). The tilted final imaging system consisted of the objective lens (OL2, LMPL50XIR, Olympus), camera lens (L5: MVL50M23, Navitar), and line scan camera (OCTOPLUS, Teledyne e2v), all of which were mounted on a 5-axis stage (X-Y-Z, yaw, and pitch).
One challenge with CoSLO is that human ocular optics has low NA ∼0.23, in contrast to high NA objective lens in SOLSM [30,31]. The intersection angle between excitation and detection is small. If the conjugate image of the oblique line at IIS were to maintain the same 10° excitation angle with unity magnification, the remote focusing system would have dismal collection efficiency [32]. To overcome this challenge, we used a demagnifying design from the retina to IIS, so that the angle of the stationary line image can be increased with M1 < 1 [36] by
where β is the angle of the stationary line image with respect to the optical axis of OL1, α (10°) is the angle of the oblique excitation laser line on the retina, and M1 is the lateral magnification from the retina to IIS.Another different design in CoSLO from our previous work [36–38] is to use a confocal detection scheme in the remote focusing system by utilizing a line sensor array instead of a planar camera. The narrow line width of the linear photon detector allowed it to serve as the slit aperture. Although this type of confocal line scanning system only has confocal effects in one dimension [39], it still can effectively reject out-of-focus signals [Supplemental Fig. S5].
2.3 3D reimaging of an oblique object under low NA by a demagnification design
To reduce spherical aberration when replicating the 3D distribution of object space into the focal region of a second objective lens, the magnification is usually designed to match the refractive index of the media across the object space and the image space when a high NA objective lens (e.g., NA = 1.4) is used [40]. However, the spherical aberration of reimaging 3D object under low NA objective, such as human ocular optics, was not investigated. The ray-tracing model shown in Fig. 3(a) was established to evaluate the optical aberration of our demagnifying design. The oblique A-line illumination and its conjugated stationary A-line image are shown in panel Fig. 3(a). An eye model was used to simulate the human eye [41]. The lens data for the tube lens and the objective lens is provided by the vendor or can be found in the patent [42]. Although the lens data in the patent may not fully match with that of the actual objective lens, the ray-tracing results are only representative of basic trend. Pupil size of ∼5 mm in the eye model was utilized in the ray tracing. Considering the amount of light that can be detected by the refocusing system, only a partial back aperture of the objective lens was utilized in the simulation.
Fig. 3. The simulation of the aberration and light loss under varying M1. (a) The ray-tracing model from the retina to intermediate imaging space (IIS). L1 and L2 are the relay lenses. OL1 is the objective lens used to form the stationary A-line image in IIS. The angles of the oblique illumination and stationary A-line image with respect to the optical axis are denoted as α and β, respectively. (b) and (c) The examples of the calculation of light loss under varying magnification M1. in panel a from the retina to IIS. The green and blue triangles represent the light cone of OL1 and OL2. OL2 is the primary objective lens in the tilted remote focusing system. (d) The Seidel spherical aberration and light loss under different M1. (e) and (f) The spot diagrams under different M1.
By changing the focal length of the tube lens in Fig. 3(a), both the light loss and the Seidel spherical aberration under different M1 could be analyzed. Figure 3(b) and 3(c) include two examples showing how the demagnification improved the collection efficiency. The improvement of light loss under different M1 is shown in Fig. 3(d). However, the spherical aberration increases with further demagnification at the same time (Fig. 3(d)). Figure 3(e and f) demonstrated that smaller M1 led to deteriorated spot diagram. According to the aberration curve shown in Fig. 3(d) and the spot diagram shown in Fig. 3(e and f), an optimized M1 (e.g., M1 = 0.28) can achieve ∼32% light collection efficiency while maintaining the spherical aberration within ∼3.5 µm and keeping the spot diagram at near diffraction-limited resolution. M1 is designed to be 0.28 by choosing proper tube lenses (L1, L2) and objective lens (OL1: UplanFL N 20 × /0.75, Olympus). After M1 is determined, β is calculated to be ∼32° by Eq. (1). With a practical light collection efficiency and limited aberration, the proposed demagnification design overcomes the limitation of 3D reimaging oblique objects in the human eye.
2.4 Generalized 3D confocal theoretical framework to evaluate resolution in CoSLO
In order to evaluate the diffraction-limited resolution by CoSLO, we adopted the framework for confocal laser scanning imaging with a generalization in 3D, the theoretical expression for CoSLO resolution can be written as (see supplementary materials for detailed derivation)
Having established the theoretical framework, we set out for a numerical simulation to evaluate the diffraction-limited resolution in CoSLO. While Eq. (2) is expressed in spatial domain, it is more convenient to calculate the 3D frequency support in the Fourier domain and convert it to spatial domain. The Fourier domain model is shown in Fig. 4(a). The coordinate is reciprocal with the Cartesian coordinate (X, Y, Z) in Fig. 1, and can be expressed with a wave vector $\vec{k}$,
where $\hat{k}$ is the directional vector with unit magnitude, and λ is the wavelength. Based on the Fourier model, the frequency ranges of the illumination and detection in CoSLO are specified. The blue circle S1 and the black circle S2 represent the spatial frequency range for the illumination and the dilated human pupil (NA = 0.23), respectively. The dash green circle S3 shows the equivalent frequency range of the remote focusing system that is mapped on the pupil. The overall spatial frequency range for the detection is illustrated by the intersection area of S2 and S3. kI and kD are used to describe the frequency offset resulting from the oblique alignment of both the illumination and the detection. After the description of the 3d frequency support, hi and hd can be obtained by performing 3D Fourier transformation (Fig. 4, b-c). The specific calculations of hi and hd can be found in [Supplementary Materials].Fig. 4. The Fourier domain model of the CoSLO. (a) The 3D frequency support of the illumination and detection in the 3D Fourier domain. (b and c) The Y-Z cross-sections of the illumination and detection PSFs. (d) The spatial relationship of the virtual confocal array aligned with the oblique illumination path on the retina. The virtual aperture is a demagnified image of the actual sensor pixel.
To evaluate how confocal aperture would impact the system PSF, the dimension of the virtual confocal aperture must be considered. Figure 4(d) illustrates the alignment of the virtual confocal aperture with the oblique illumination PSF. The size of the virtual confocal aperture, A, is described by W and H (Unit: microns), which have inequivalent magnifications from the physical pixel size in the linear detector (See Supplementary Materials for detailed calculation).
2.5 Human retinal imaging procedure
The institutional review board at Johns Hopkins University has reviewed and approved the study. The study was Health Insurance Portability and Accountability Act-compliant and adhered to the tenets of the Declaration of Helsinki.
One drop of 1% Tropicamide was used to dilate the pupil. As the proposed confocal detection scheme allows the detection of individual A-line at any scanning position, we designed a cross scanning pattern in the alignment mode to display both the X-Z and Y-Z simultaneously. After focus adjustment and beam alignment at the pupil, satisfactory image quality was obtained in both cross-sectional images. Then a raster scanning was used to acquire the volumetric data.
The powers on the cornea for the NIR and visible light channels were 200 µW and 60 µW, respectively. The power was well within the maximum permissible exposure power that allowed for laser safety [43]. The line acquisition rate was set to 50K Hz, and the fast axis has a duty cycle of 0.8.
2.6 Image processing
The remote focusing system acquired one A-line image of the scanning oblique excitation each time. Sequentially scanned A-line images across fast axis were used to build the B-scan, which is an oblique planar image. Because of the oblique illumination and detection, directly stacking B-scans does not reflect the realistic 3D geometry. Therefore, affine transformation [32,44,45] consisting of scaling and shearing can be used to transform the 3D volume to a standard cartesian coordinate to recover the geometry. The geometric transformation function (imwrap) in MATLAB (MathWorks) was used to implement the affine transformation. After the correction, the en face retinal images can be obtained by calculating maximum intensity projections (MIPs) of different layers along the depth dimension. All the image processing was performed in MATLAB.
3. Results
3.1 Theoretical resolution
The numerical simulation based on the Fourier domain model described in Materials and methods was performed to estimate the diffraction-limited resolution in CoSLO. Figure 5(a-b) illustrates how a finite aperture size leads to a relaxed overall system resolution as expected. Figure 5(c) describes the change of system resolution in three dimensions with varying W at fixed H = 50 µm. The axial resolution deteriorates when W increases along the excitation light path. At the same time, Y resolution becomes worse. Figure. 5(d) shows the change of the system resolution with varying H at W = 7.5 µm. The axial resolution is maintained within 32-35 µm, and increasing H only impacts X resolution. In our setup, the line camera pixel size (200 × 10 µm) is equivalent to H = 51 µm, and W = 7.7 µm. Therefore, the theoretical resolution for the NIR light is 13.5, 5, and 35 µm in X, Y, and Z. As the wavelength becomes shorter, the visible light has better resolution of 10, 3.5, and 24 µm in X, Y, and Z.
Fig. 5. The theoretical resolution of CoSLO. (a and b) The Y-Z cross-sections of the overall system PSFs with infinite small and finite aperture size. (c and d) The change of the resolution under varying width, W, and Height, H, of the aperture. The scale bar is 100 µm.
3.2 Experimental resolution characterization
To characterize the resolution and FOV, a schematic eye model [36] was imaged. A thin layer of agarose gel with dispersed polystyrene microspheres was attached to the bottom of the eye model to mimic the retina. The eye model was first imaged by optical coherent tomography (OCT) [46] to obtain the dimension of the agarose gel (See supplementary materials). Figure 6(a-c) show three cross-sectional projections of the non-interferometric volumetric data with NIR illumination, with additional live volume data provided in the flythrough in Visualization 1. The circular boundary in Fig. 6(b) indicates the 3D-printed mold to confine the gel. The diameter of the mold is 10 mm, demonstrating the large FOV by our CoSLO. Vignetting began to appear at the edge of the FOV, limiting more in the X direction than Y. The bright signal from the bottom of the gel marks the inner boundary of the schematic eye model. The overall optical magnification was calibrated to ∼1.2 and ∼3.5 for the depth and lateral directions, respectively. The calibrated results are close to the above theoretical calculations. The zoom-in view of the squared area in Fig. 6(b) is shown in Fig. 6(d). Figure 6(e) is the X-Z cross-section taken from the area between the dashed lines in Fig. 6(d). As the thickness of the agarose gel is within twice the Rayleigh range, Fig. 6(e) exhibits consistent image quality in the depth direction. The profiles of three representative microspheres are shown in Fig. 6(f). To quantitatively analyze the resolution over the full FOV, the full width at half maximum (FWHM) along each direction of different microspheres was calculated. As shown in Fig. 6(g), the resolutions for NIR light along X, Y, and Z direction are 15.4 ± 1.3 µm, 11.3 ± 1.1 µm, and 45 ± 3 µm, respectively. The same characterization was carried out for visible light. Due to the shorter wavelength, the visible light has better resolution, which are 14.9 ± 1.5 µm, 10.2 ± 1.6 µm, and 39.6 ± 5 µm in X, Y, and Z. Furthermore, we evaluated the sensitivity of CoSLO by contrasting signals from the weakly scattering gel to the signal above gel (Fig. 6(h)), and compared that from OCT (Fig. S4c). The intensity profile in Fig. 6(h) was calculated by averaging 8 pixels. The value of the signal was calculated by averaging 20 pixels in the gel along the direction of the intensity profile while that of the noise was calculated by averaging 20 pixels above the gel along the direction of the intensity profile. As for OCT, the measurements were calculated by averaging the intensity of pixels representing the same physical area (∼20 × 8 pixels for CoSLO and ∼150 × 8 pixels for OCT). The ratio was calculated to be ∼14 and ∼4.4 in CoSLO and OCT, respectively.
Fig. 6. The resolution characterization using a model eye by CoSLO. (a) X-Z cross-section generated by calculating MIP of the area between two green lines in panel (b); (b) The en-face view generated by MIP of the area between the purple lines in panel (c); (c) The Y-Z cross-section generated by MIP of the area between the brown lines in panel (c); (d) The zoom-in view of the square area in panel (b). (e) The X-Z view of the area between the white dash lines in panel (d). (f) The profile of three representative microspheres. (g) The resolutions along X, Y, and Z directions; Error bar = SD, n = 50; Scale bars in (a to c), (d to e), and (f) are 500 µm, 100 µm, and 25 µm, respectively. (h) The intensity profile is taken from the position indicated by the white dash line in panel c. The signals above the gel represent the noise floor. The signals from the gel (without the presence of beads) represent the smallest reflectivity. The white dash line has a width of 8 pixels along X dimension.
We also evaluated the effect of confocal gating in CoSLO, and demonstrated that shorter sensor heights (H) effectively suppressed diffusive background and improved image contrast [Supplementary Materials, Fig. S5].
3.3 Human retina imaging by non-interferometric scattering contrast in 3D
With the resolution characterization, we proceeded to image the living human retina. Two healthy human volunteers participated in this study. Subject 1 has myopia of −2.5 diopters, and subject 2 has hyperopia of 1 diopter and astigmatism of 3 diopters in their imaged eyes respectively. To demonstrate the FOV of the CoSLO, volumetric data covering an area of ∼6 × 6 mm2 in the retina, i.e., 20° viewing angle, was first acquired using the NIR channel (Fig. 7, a-e) for subject 1. The large FOV enabled imaging across the optic nerve head (ONH) and macular region. The entire volumetric data had a pixel density of 512 × 512, acquired in ∼5 s by a 50k Hz A-line rate. Depth-resolved signal was acquired in parallel as compared to sequential z-stacking in conventional point scanning methods. Figure 7(a-c) are maximum intensity projections (MIPs) which were taken from around 0–50 µm, 130–170 µm, and 300–330 µm from the retinal boundary that indicated by the yellow line in Fig. 7(d). Due to the depth sectioning ability of CoSLO, different anatomical features can be identified. Figure 7(a) shows nerve fiber bundles in the inner retina close to ONH and retinal vasculatures. Figure 7(b) shows the shadows of the vasculatures as well as spotty signals in the background. By extracting a deeper layer as shown Fig. 7(c), the choroidal vessels can be observed. The cross-sectional views in Fig. 7(d) demonstrates distinct features along the depth of the retina. An interesting observation is the double image of vascular shadow in the en-face MIPs from the outer retina. This is due to the blood attenuations for both the oblique illumination and the angled detection (Fig. 7(e)).
Fig. 7. The demonstration of FOV in imaging human retina by CoSLO. (a-c) are the maximum intensity projection (MIP) of the X-Y planes taken from three layers with positions indicated by the black, blue, and brown markers in panel (e), respectively. (d) The X-Z cross-section taken from the position indicated by the red vertical line in panel (a). The scale bars for X, Y, and Z direction are 1000 µm, 1000 µm, and 100 µm, respectively. (e) The explanation of the blood vessels with double edges.
To observe the details of the retina in 3D, high-definition volume data with a FOV of 2.85 × 4 mm2 and a pixel density of 370 × 512 was acquired within 4 s. The volume data from subject 2 is shown in Fig. 8 (Visualization 2 and 4). The data from subject 1 was collected under the same experiment parameters and is shown in supplemental Fig. S6 (Visualization 3). En-face MIPs from the inner retina, outer retina, and choroid were plotted in Fig. 8(a-c). The depth ranges to generate the en-face MIPs were indicated in the cross-sectional image in Fig. 8(d). The yellow and blue dash lines indicated the inner and outer boundaries of the retina, serving as depth references. Individual nerve fiber bundles in the superficial retina layer can be clearly observed in Fig. 8(e and f).
Fig. 8. High-definition cross-section images of the human retina by CoSLO. (a-c) are the MIPs of the X-Y planes taken from the positions indicated by the black, blue, and brown markers in panel (d), respectively. (d) The X-Z cross-section taken from the position indicated by the red dash line in panel (a). (e-f) The zoom-in view of the squared area in panel (a). (g-h) The zoom-in view of the squared areas in panel (b). (i and j) The intensity profiles of adjacent bright spots marked by squares in panel (g and h). (k) The zoom-in view of the squared area in panel (d). (l and m) The XZ cross-sections taken from the location indicated by the dashed lines in (c). (n) The intensity profiles of adjacent bright spots in the squared areas in panel k. (o-p) The intensity profiles of the adjacent bright spots in panel (l-m). The scale bars for X, Y, and Z direction are 1000 µm, 1000 µm, and 100 µm, respectively.
The bright spotty signals in Fig. 8(g and h), which were taken from the outer retina layer, were likely contributed by cone photoreceptors [19]. Intensity profiles of adjacent bright spots shown from the locations marked by the black and blue rectangles in Fig. 8(g and h) are plotted in Fig. 8(i and j). The full width at half maximum (FWHM) of both curves is ∼15 µm which matches the characterized resolution by the schematic eye. The intensity profiles in two peripheral areas of the FOV support a well maintained lateral resolution throughout the whole FOV.
To better illustrate the depth resolution in CoSLO, an enlarged view from Fig. 8(d) is shown in Fig. 8(k). The discrimination of bright contrast is direct evidence of the depth sectioning ability of CoSLO. Figure 8(n) shows the normalized A-line intensity in Fig. 8(k) confirmed that the small feature with a distance of ∼45 µm can be well resolved, which matches the previous resolution characterization in the eye model. To further investigate whether the axial resolution is well maintained in the whole FOV, two B-scans (Fig. 8(l and m)) with a distance of ∼2.7 mm were extracted from the locations indicated by the dashed lines in Fig. 8(c). The normalized A-line intensities are shown in Fig. 8(o and p), demonstrating that the axial resolution is also well maintained within ∼45 µm over the whole FOV. To validate the obtained cross-sections, the images from OCT and CoSLO were compared (Fig. S8). OCT shows better axial resolution as expected. The comparison confirms that the features shown in CoSLO are also present in OCT.
In addition to image orientation with the central FOV around macula, we tested whether the performance is maintained when imaging an off-center FOV on ONH about 6mm away from the fovea (Fig. 9(a-d)). The cross-sectional view (Fig.9a) displayed similar contrast through the retina as in Fig. 8(d). En-face projection views were generated within the inner retina, outer retina, and choroid in Fig. 9(b-d). The nerve fiber bundles are clearly visualized within the inner retina where the nerve fiber layer thickens in the peripapillary region (Fig. 9(b)). The vessel shadow and the rim of the nerve head in Fig. 9(c) are enhanced features in the outer retina, and we are able to image choroidal vessels in the choroid (Fig. 9(d)) as dark contrast.
Fig. 9. The comparison of the volume data sets acquired by NIR light (a-d) and visible light (e-h). (a and e) are the X-Z cross-sectional views taken from the positions indicated by the red arrows in panels (b) and (f), respectively. (b-d) are the MIPs of X-Y planes taken from the depth positions indicated by the black, blue, and brown markers in panel (a). The yellow and blue dash lines in panel (a) were the boundaries of the retina that were used as depth references for the segmentation. (f-h) are the MIPs of X-Y planes taken from the area indicated by the black, blue, and brown markers in panel (e), respectively. (i and j) are XZ cross-sections taken from the locations indicated by the dashed lines in panels (d) and (h). The scale bars for X, Y, and Z direction are 1000 µm, 1000 µm, and 200 µm, respectively.
By simply changing the light source, CoSLO is able to image volumetric spectroscopic contrast in different wavelengths. Figure 9(e-h) shows the CoSLO image with a visible wavelength (600 ± 30 nm) illumination on the same ONH FOV. The anatomical features in the cross-section and three en-face projections are similar to those with NIR illumination. The signal is relatively weaker in the outer retina and beyond when comparing the baseline Fig. 9(e) to Fig. 9(a). A similar comparison of the cross-sections from the rim of the nerve head is provided in Fig. 9(i) and Fig. 9(j). The difference could be due to the stronger absorption by photoreceptor pigmentation or RPE in the visible wavelength. Another distinct spectroscopic contrast between NIR and the visible channel is the choroidal vessels, which appear in dark contrast in Fig. 9(d) with NIR illumination, but bright contrast in Fig. 9(h) with visible illumination. A Dual-wavelength imaging from a different FOV in subject 1 under the same experiment parameters is shown in supplemental Fig. S7.)
3.4 Comparison of theoretical and measured resolutions
A summary of resolutions obtained by numerical simulation (section 3.1), beads imaging (section 3.2) experiment, and human eye imaging experiment (section 3.3) is shown in Table 1 for comparison. All the resolutions were obtained under the NIR channel. The resolution of human eye imaging was estimated by evaluating the FWHM of intensity profiles of the small features in the acquired data. Considering the aberration of the optical system, which is not included in the theoretical framework for resolution characterization, the experimental resolution generally matched well with that of the theoretical resolutions.
Table 1. Resolution comparison at 830nm
4. Discussion
A new 3D imaging modality termed CoSLO for the human retina is presented. CoSLO has two major improvements as compared to the previous version [36–38]. Firstly, a confocal line detection scheme is used to improve the image contrast. Secondly, a 3D reimaging scheme for low NA optics is utilized such that the application of oSLO can be extended to the human eye with a small NA of ∼0.2. This study demonstrates the first time that large-scale volumetric non-interferometric imaging has been made possible in the living human retina. CoSLO avoids z-stacking and achieves adequate volumetric imaging speed for the human eye by utilizing oblique illumination, remote focusing, and line de-scanning. By utilizing different light sources, wavelength-dependent retina features have been revealed.
We overcame the limitation of using oblique illumination and oblique detection in low NA human ocular optics by a demagnification design. We derived the analytical expression for CoSLO, and analyzed how the confocal gating impacts the 3D resolution. At the current setup, numerical simulation of the NIR light suggests 35 µm axial resolution, 13.5 and 5 µm in X, Y directions. The experimental resolution using microspheres in the gel is ∼45 µm axial resolution, 15 and 11 µm in X and Y. The axial resolution is about four times better than the conventional SLO reported in [16,17] and comparable with AOSLO [19]. Experimental results of imaging the human eye also reveal the same axial resolution which is well maintained over a distance that is larger than the FOV of the existing AOSLO [47]. The depth resolution in CoSLO is ultimately determined by the accessible spatial frequency range, as shown in the 3D frequency support (Fig. 4(a)), particularly in the kz direction. Although CoSLO only collects the signal existing through the half side of the pupil, the kz range is largely covered, and therefore the axial resolution is in the similar order of AOSLO. The experimental resolution is worse than the simulation, presumably due to the aberration. However, since only partial pupil is utilized, CoSLO could be less sensitive to aberration than using the entire pupil and thus still maintains reasonable axial resolution. With comparable axial resolution and much larger FOV than AOSLO, CoSLO will be helpful to explore non-interferometric contrast in the human retina.
The effect of confocal gating with the linear detector array is similar to that in conventional confocal line scanning microscopy [39], only that the confocal gating is aligned at an angle due to the tilted remote focusing system. Beyond the impact on the diffraction-limited resolution in Fig. 4(e and f), one major benefit of confocal gating is to reject diffusive signal (Supplementary Materials, Fig. S5) and improve the imaging contrast and resolution. This is particularly useful when imaging highly scattering tissue.
With the successful demonstration of CoSLO in the living human retina, we aim to improve the current system on three major aspects in our ongoing work. First, the imaging speed of our current implementation is mainly limited by the galvanometers which have a large aperture and were used to ensure light collection efficiency. With ∼5-s long acquisition time, it may take several tries for the current CoSLO to avoid significant eye motion. Future improvements can be achieved by using a faster scanner. Second, both the lateral and the axial resolution can be improved by utilizing adaptive optics. Third, the alignment of the eye pupil will impact the image quality. A motorized chinrest can be used to track and compensate the movement of the pupil in real-time such that the offset of the excitation beam projected on the pupil can be well maintained during the acquisition. This paper serves as an important proof-of-concept of CoSLO for non-interferometric volumetric imaging in human retina. Our future plan includes imaging a range of non-interferometric contrasts, including exogenous fluorescence contrast (i.e., Fluorescein, indocyanine green) to characterize vascular permeability [48]; intrinsic autofluorescence from photoreceptor [2], RPE [3,4], and choroid [5,49] that has important implications in macular degeneration [50] and other retinal degeneration conditions (e.g. retinitis pigmentosa [51]). Beyond fluorescence, we will also explore 3D imaging on non-interferometric scattering-based contrasts in the human retina. For example, the absorption contrast by the photothermal effect can be used to characterize melanin and hemoglobin content by a pump-probe approach [52,53], as well as multiple scattering contrast by misaligning the line camera cells [7,9,11].
In conclusion, CoSLO is a novel method for non-interferometric volumetric imaging for the human retina. By rapid 3D imaging capability, CoSLO bridges the gap between rich non-interferometric contrasts and the limitation of the current 2D retinal imaging modalities. We expect CoSLO will be another important imaging method with significant translational values.
Funding
National Institutes of Health (R01NS108464); BrightFocus Foundation (2018132).
Acknowledgments
The authors would like to thank Dr. Zhenpin Guan and Dr. Yahui Wang for their assistance during the experiment. The authors also appreciate the help of Dr. Weiye Song in providing the OCT data of the schematic eye.
Disclosures
All other authors declare they have no competing interests.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
Supplemental document
See Supplement 1 for supporting content.
References
1. H. R. Novotny and D. L. Alvis, “A method of photographing fluorescence in circulating blood in the human retina,” Circulation 24(1), 82–86 (1961). [CrossRef]
2. T. Ach, C. Huisingh, G. McGwin, J. D. Messinger, T. Zhang, M. J. Bentley, D. B. Gutierrez, Z. Ablonczy, R. T. Smith, K. R. Sloan, and C. A. Curcio, “Quantitative autofluorescence and cell density maps of the human retinal pigment epithelium,” Invest. Ophthalmol. Vis. Sci. 55(8), 4832–4841 (2014). [CrossRef]
3. T. Liu, H. Jung, J. Liu, M. Droettboom, and J. Tam, “Noninvasive near infrared autofluorescence imaging of retinal pigment epithelial cells in the human retina using adaptive optics,” Biomed. Opt. Express 8(10), 4348–4360 (2017). [CrossRef]
4. J. I. W. Morgan, A. Dubra, R. Wolfe, W. H. Merigan, and D. R. Williams, “In vivo autofluorescence imaging of the human and macaque retinal pigment epithelial cell mosaic,” Invest. Ophthalmol. Vis. Sci. 50(3), 1350–1359 (2009). [CrossRef]
5. S. Schmitz-Valckenberg, D. Lara, S. Nizari, E. M. Normando, L. Guo, A. R. Wegener, A. Tufail, F. W. Fitzke, F. G. Holz, and M. F. Cordeiro, “Localisation and significance of in vivo near-infrared autofluorescent signal in retinal imaging,” Br. J. Ophthalmol. 95(8), 1134–1139 (2011). [CrossRef]
6. V. Vaclavik, S. Vujosevic, S. S. Dandekar, C. Bunce, T. Peto, and A. C. Bird, “Autofluorescence imaging in age-related macular degeneration complicated by choroidal neovascularization: a prospective study,” Ophthalmology 115(2), 342–346 (2008). [CrossRef]
7. T. Y. P. Chui, D. A. VanNasdale, and S. A. Burns, “The use of forward scatter to improve retinal vascular imaging with an adaptive optics scanning laser ophthalmoscope,” Biomed. Opt. Express 3(10), 2537–2549 (2012). [CrossRef]
8. D. Scoles, Y. N. Sulai, and A. Dubra, “In vivo dark-field imaging of the retinal pigment epithelium cell mosaic,” Biomed. Opt. Express 4(9), 1710–1723 (2013). [CrossRef]
9. E. A. Rossi, C. E. Granger, R. Sharma, Q. Yang, K. Saito, C. Schwarz, S. Walters, K. Nozato, J. Zhang, T. Kawakami, W. Fischer, L. R. Latchney, J. J. Hunter, M. M. Chung, and D. R. Williams, “Imaging individual neurons in the retinal ganglion cell layer of the living eye,” Proc. Natl. Acad. Sci. 114(3), 586–591 (2017). [CrossRef]
10. T. Laforest, M. Künzi, L. Kowalczuk, D. Carpentras, F. Behar-Cohen, and C. Moser, “Transscleral optical phase imaging of the human retina,” Nat. Photonics 14(7), 439–445 (2020). [CrossRef]
11. A. Guevara-Torres, D. R. Williams, and J. B. Schallek, “Origin of cell contrast in offset aperture adaptive optics ophthalmoscopy,” Opt. Lett. 45(4), 840–843 (2020). [CrossRef]
12. P. A. Liebman and R. A. Leigh, “Autofluorescence of visual receptors,” Nature 221(5187), 1249–1251 (1969). [CrossRef]
13. G. K. Frampton, N. Kalita, L. Payne, J. L. Colquitt, E. Loveman, S. M. Downes, and A. J. Lotery, “Fundus autofluorescence imaging: systematic review of test accuracy for the diagnosis and monitoring of retinal conditions,” Eye 31(7), 995–1007 (2017). [CrossRef]
14. R. H. Webb, G. W. Hughes, and F. C. Delori, “Confocal scanning laser ophthalmoscope,” Appl. Opt. 26(8), 1492–1499 (1987). [CrossRef]
15. K. V. Vienola, M. Damodaran, B. Braaf, K. A. Vermeer, and J. F. de Boer, “Parallel line scanning ophthalmoscope for retinal imaging,” Opt. Lett. 40(22), 5335–5338 (2015). [CrossRef]
16. W. J. Donnelly and A. Roorda, “Optimal pupil size in the human eye for axial resolution,” J. Opt. Soc. Am. A 20(11), 2010–2015 (2003). [CrossRef]
17. J. J. Hunter, C. J. Cookson, M. L. Kisilak, J. M. Bueno, and M. C. W. Campbell, “Characterizing image quality in a scanning laser ophthalmoscope with differing pinholes and induced scattered light,” J. Opt. Soc. Am. A 24(5), 1284–1295 (2007). [CrossRef]
18. A. Roorda, F. Romero-Borja, W. J. D. Iii, H. Queener, T. J. Hebert, and M. C. W. Campbell, “Adaptive optics scanning laser ophthalmoscopy,” Opt. Express 10(9), 405–412 (2002). [CrossRef]
19. R. Lu, N. Aguilera, T. Liu, J. Liu, J. P. Giannini, J. Li, A. J. Bower, A. Dubra, and J. Tam, “In-vivo sub-diffraction adaptive optics imaging of photoreceptors in the human eye with annular pupil illumination and sub-Airy detection,” Optica 8(3), 333–343 (2021). [CrossRef]
20. P. Mecê, E. Gofas-Salas, M. Paques, M. Paques, K. Grieve, K. Grieve, and S. Meimon, “Optical Incoherence Tomography: a method to generate tomographic retinal cross-sections with non-interferometric adaptive optics ophthalmoscopes,” Biomed. Opt. Express 11(8), 4069–4084 (2020). [CrossRef]
21. J. Lu, B. Gu, X. Wang, and Y. Zhang, “Adaptive optics parallel near-confocal scanning ophthalmoscopy,” Opt. Lett. 41(16), 3852–3855 (2016). [CrossRef]
22. Z. Liu, J. Tam, O. Saeedi, and D. X. Hammer, “Trans-retinal cellular imaging with multimodal adaptive optics,” Biomed. Opt. Express 9(9), 4246–4262 (2018). [CrossRef]
23. T. DuBose, D. Nankivil, F. LaRocca, G. Waterman, K. Hagan, J. Polans, B. Keller, D. Tran-Viet, L. Vajzovic, A. N. Kuo, C. A. Toth, J. A. Izatt, and S. Farsiu, “Handheld adaptive optics scanning laser ophthalmoscope,” Optica 5(9), 1027–1036 (2018). [CrossRef]
24. M. Azimipour, R. S. Jonnal, J. S. Werner, R. J. Zawadzki, and R. J. Zawadzki, “Coextensive synchronized SLO-OCT with adaptive optics for human retinal imaging,” Opt. Lett. 44(17), 4219–4222 (2019). [CrossRef]
25. P. Mecê, E. Gofas-Salas, C. Petit, F. Cassaing, J. Sahel, M. Paques, K. Grieve, and S. Meimon, “Higher adaptive optics loop rate enhances axial resolution in nonconfocal ophthalmoscopes,” Opt. Lett. 44(9), 2208–2211 (2019). [CrossRef]
26. K. Hagan, T. DuBose, D. Cunefare, G. Waterman, J. Park, C. Simmerer, A. N. Kuo, A. N. Kuo, R. P. McNabb, J. A. Izatt, J. A. Izatt, S. Farsiu, and S. Farsiu, “Multimodal handheld adaptive optics scanning laser ophthalmoscope,” Opt. Lett. 45(17), 4940–4943 (2020). [CrossRef]
27. D. X. Hammer, R. D. Ferguson, T. E. Ustun, C. E. Bigelow, N. V. Iftimia, and R. H. Webb, “Line-scanning laser ophthalmoscope,” J. Biomed. Opt. 11(4), 041126 (2006). [CrossRef]
28. R. Zeimer, M. Shahidi, M. Mori, S. Zou, and S. Asrani, “A new method for rapid mapping of the retinal thickness at the posterior pole.,” Invest. Ophthalmol. Vis. Sci. 37, 1994–2001 (1996).
29. M. Shahidi, J. Wanek, N. P. Blair, and M. Mori, “Three-dimensional mapping of chorioretinal vascular oxygen tension in the rat,” Invest. Ophthalmol. Vis. Sci. 50(2), 820–825 (2009). [CrossRef]
30. C. Dunsby, “Optically sectioned imaging by oblique plane microscopy,” Opt. Express 16(25), 20306–20316 (2008). [CrossRef]
31. M. B. Bouchard, V. Voleti, C. S. Mendes, C. Lacefield, W. B. Grueber, R. S. Mann, R. M. Bruno, and E. M. C. Hillman, “Swept confocally-aligned planar excitation (SCAPE) microscopy for high-speed volumetric imaging of behaving organisms,” Nat. Photonics 9(2), 113–119 (2015). [CrossRef]
32. W. Shao, K. Kilic, W. Yin, G. Wirak, X. Qin, H. Feng, D. Boas, C. V. Gabel, and J. Yi, “Wide field-of-view volumetric imaging by a mesoscopic scanning oblique plane microscopy with switchable objective lenses,” Quant. Imaging Med. Surg. 11, 98397–98997 (2021). [CrossRef]
33. N. J. Sofroniew, D. Flickinger, J. King, and K. Svoboda, “A large field of view two-photon mesoscope with subcellular resolution for in vivo imaging,” eLife 5, e14472 (2016). [CrossRef]
34. M. Pisanello, D. Zheng, A. Balena, F. Pisano, M. D. Vittorio, and F. Pisanello, “An open source three-mirror laser scanning holographic two-photon lithography system,” PLoS ONE 17(4), e0265678 (2022). [CrossRef]
35. J. Polans, B. Jaeken, R. P. McNabb, P. Artal, and J. A. Izatt, “Wide-field optical model of the human eye with asymmetrically tilted and decentered lens that reproduces measured ocular aberrations,” Optica 2(2), 124–134 (2015). [CrossRef]
36. W. Shao, W. Song, and J. Yi, “Is oblique scanning laser ophthalmoscope applicable to human ocular optics? A feasibility study using an eye model for volumetric imaging,” J. Biophotonics 13(6), e201960174 (2020). [CrossRef]
37. W. Song, L. Zhou, and J. Yi, “Volumetric fluorescein angiography (vFA) by oblique scanning laser ophthalmoscopy in mouse retina at 200 B-scans per second,” Biomed. Opt. Express 10(9), 4907–4918 (2019). [CrossRef]
38. L. Zhang, W. Song, D. Shao, S. Zhang, M. Desai, S. Ness, S. Roy, and J. Yi, “Volumetric fluorescence retinal imaging in vivo over a 30-degree field of view by oblique scanning laser ophthalmoscopy (oSLO),” Biomed. Opt. Express 9(1), 25–40 (2018). [CrossRef]
39. K.-B. Im, S. Han, H. Park, D. Kim, and B.-M. Kim, “Simple high-speed confocal line-scanning microscope,” Opt. Express 13(13), 5151–5156 (2005). [CrossRef]
40. E. J. Botcherby, R. Juškaitis, M. J. Booth, and T. Wilson, “An optical technique for remote focusing in microscopy,” Opt. Commun. 281(4), 880–887 (2008). [CrossRef]
41. “OpticStudio models of the human eye – Knowledgebase,” https://support.zemax.com/hc/en-us/articles/1500005575082-OpticStudio-models-of-the-human-eye.
42. Y. Fujimoto, “Dry-type microscope objective lens,” United States patent US7982961B2 (July 19, 2011).
43. American National Standards Institute, Light Hazard Protection for Ophthalmic Instruments, Light Hazard Protection for Ophthalmic Instruments (2016).
44. M. Hoffmann and B. Judkewitz, “Diffractive oblique plane microscopy,” Optica 6(9), 1166–1170 (2019). [CrossRef]
45. M. Kumar, S. Kishore, J. Nasenbeny, D. L. McLean, and Y. Kozorovitskiy, “Integrated one- and two-photon scanned oblique plane illumination (SOPi) microscopy for rapid volumetric imaging,” Opt. Express 26(10), 13027–13041 (2018). [CrossRef]
46. W. Song, W. Shao, W. Yi, R. Liu, M. Desai, S. Ness, and J. Yi, “Visible light optical coherence tomography angiography (vis-OCTA) facilitates local microvascular oximetry in the human retina,” Biomed. Opt. Express 11(7), 4037–4051 (2020). [CrossRef]
47. M. Laslandes, M. Salas, C. K. Hitzenberger, and M. Pircher, “Increasing the field of view of adaptive optics scanning laser ophthalmoscopy,” Biomed. Opt. Express 8(11), 4811–4826 (2017). [CrossRef]
48. B. F. Hochheimer, “Angiography of the retina with indocyanine green,” Archives of Ophthalmology 86(5), 564–565 (1971). [CrossRef]
49. A. von Rückmann, F. W. Fitzke, and A. C. Bird, “Fundus autofluorescence in age-related macular disease imaged with a laser scanning ophthalmoscope,” BLL 113(02), 114–116 (2012). [CrossRef]
50. S. Schmitz-Valckenberg, F. G. Holz, A. C. Bird, and R. F. Spaide, “Fundus autofluorescence imaging: review and perspectives,” Retina 28(3), 385–409 (2008). [CrossRef]
51. T. Duncker, M. R. Tabacaru, W. Lee, S. H. Tsang, J. R. Sparrow, and V. C. Greenstein, “Comparison of near-infrared and short-wavelength autofluorescence in retinitis pigmentosa,” Invest. Ophthalmol. Vis. Sci. 54(1), 585–591 (2013). [CrossRef]
52. M. Lapierre-Landry, A. L. Huckenpahler, B. A. Link, R. F. Collery, J. Carroll, and M. C. Skala, “imaging melanin distribution in the zebrafish retina using photothermal optical coherence tomography,” Trans. Vis. Sci. Tech. 7(5), 4 (2018). [CrossRef]
53. R. V. Kuranov, S. Kazmi, A. B. McElroy, J. W. Kiel, A. K. Dunn, T. E. Milner, and T. Q. Duong, “In vivo depth-resolved oxygen saturation by dual-wavelength photothermal (DWP) OCT,” Opt. Express 19(24), 23831–23844 (2011). [CrossRef]
