2.1.

Phase-Variance SSOCT System

A schematic diagram of our $1\text{-}\mu \mathrm{m}$ SSOCT system used for anterior segment imaging has been previously reported.²³ Here, we briefly describe changes to our SSOCT system (Fig. 1) that allowed posterior segment imaging. The light source was an external cavity tunable laser, SS laser (Axsun Technologies), with a central wavelength of 1060 nm, sweep bandwidth of 110 nm, repetition rate of 100 kHz, 46% duty cycle and average output power of $\sim 23\mathrm{mW}$. A spectrally balanced interferometer configuration was used with three 50/50 fiber couplers (AC Photonics), and a balanced Mach–Zehnder fiber interferometer.⁸ The sample port was attached to a probe unit, which comprises a fiber collimator lens (12.38 mm, focal length, TC12APC-1064, Thorlabs, Inc.), two single-axis ($X\text{-}Y$) scanners (Cambridge Technology), and two achromatic doublets of focal length 50 and 30 mm. The reference port of the fiber coupler was attached to a reference arm unit comprising a collimator lens, an achromatic doublet lens, dispersion compensating block, and a static silver-coated mirror.

Fig. 1

Schematic diagram of phase-variance-swept source optical coherence tomography (pv-SSOCT) system. Swept source laser 1060-nm Axsun ECTL. FBG: fiber Bragg grating, FC: fiber coupler, GS: galvanometric scanning mirrors, M: mirror, BPD: balanced photodiode receiver (DC to 350 MHz, PDB130C, Thorlabs), PC: polarization controller, LP: linear polarizer, RF Amp: radio frequency amplifier (Minicircuits $\mathrm{ZFL}\text{-}1000+$), DB: dichroic beamsplitter, FT: fixation target, BD: beam dump, UP: unused port, OSA: optical spectrum analyzer, DAQ: digitizer (ATS9350, AlazarTech, 12-bit resolution and sampling rate of $500\mathrm{MS}/\mathrm{s}$), function generator (NI PCI-6363 DAQ, National Instruments, Texas).

2.2.

Image Acquisition and Data Processing for pv-SSOCT

The galvanometric scanners that steer the probe beam were controlled by a data acquisition (DAQ) card (PCI-6363, National Instruments, Texas). The probing beam power at the sample was set at 1.85 mW, which is lower than the ANSI limit for safe light exposure (ANSI Z136.1 standard). The reflected beams from both arms of the interferometer were combined at the couplers and detected by a balanced photodetector with a bandwidth of DC to 350 MHz (PDB130C, Thorlabs Inc.) followed by an RF amplifier and high pass filter (Minicircuits $\mathrm{ZFL}\text{-}1000+$). The OCT interference signal was digitized by a DAQ board at 12-bit digital resolution at a sampling rate defined by the external trigger provided by the Axsun SS $k$-clock (ATS9350, AlazarTech, QC, Canada).

Here, a simple method²³ for phase stabilization was implemented. It also reduces A-line trigger fluctuations from the laser source. A fixed wavelength reference signal generated by an FBG (OE Land, Quebec, Canada, ${\lambda }_0=988.9\mathrm{nm}$, $\text{reflectivity}=99.91\%$, $\mathrm{\Delta }\lambda =0.4\mathrm{nm}$) inserted in a transmission mode between the source and the OCT interferometer (Fig. 1) was used to align the start time of each A-line in postprocessing. Due to slight mismatch in the intensity of dual balance detection channels of our intereferometer, the reference signal generated by the FBG can be easily detected. Note, however, that in a perfectly balanced OCT detection system, this method would not work as the FBG reference signal could not be detected. The necessary shifts^21,23 in the acquired SSOCT signal can be calculated based on the first falling slope of the signal around the FBG signal. This phase stabilization method was adopted before pv calculation. The accuracy of phase-stabilization is discussed in Sec. 2.3. This results in measurement of phase-shifts at a noise level down to the theoretical limit (see Sec. 2.3.1), thereby allowing complete removal of fixed-pattern noise and artifact-free pv contrast images.

Synchronization between data acquisition and the scanners is achieved in the build sweep trigger signal of the laser source, which is utilized for triggering the data acquisition of the DAQ for a single sweep and additionally as an update clock for the waveform generation for the galvo scanners. The subject’s gaze was reduced by using a custom-built fixation target reflected through a dichroic beamsplitter (FF775-Di01, Semrock). The fixation target is a mini LCD screen (MIMO) controlled by computer to allow movement for precalibrated fixation light positions. Data acquisition with LabVIEW software (National Instruments) required 1 to 2 s depending on the scanning pattern settings. The raw data are saved in the PC workstation (HP xw8600, 3.2-GHz dual processors). Data sampling was performed using an optically generated clock signal with equi-spacing in wavenumber, namely, a $k$-clock from the laser source. Due to this $k$-linear sampling, no rescaling process is required before Fourier transform of the spectral data. To validate the performance of our system on different eyes, the pv-SSOCT data sets were obtained in the Vision Science and Advanced Retinal Imaging laboratory (VSRI) at the University of California Davis Medical Center on two normal subjects with normal ocular media (Subject 1: 62 year-old male and Subject 2: 27 year-old male) and four AMD patients (P1: 62 year-old male; P2: 57 year-old female; P3: 55 year-old male; P4: 73 year-old male). Written informed consent was obtained prior to imaging using a protocol approved by the UC Davis Institutional Review Board. Subject’s head position was fixed during acquisition using a custom bite-bar and forehead rest. 1% Tropicomide was used for cycloplegia and pupil dilation. A BM-scanning mode [acquisition of multiple B-scans at the same position, similar to M-mode (motion mode) in ultrasound] was used to acquire the phase-change measurements required to calculate the pv contrast. For the $1.5\times 1.5\text{-}{\mathrm{mm}}^2$ scanning pattern, $4.2\text{-}\mu \mathrm{m}$ spacing between both consecutive A-scans and BM-scans was used. Due to the spatial oversampling between consecutive BM-scans, phase changes from within two adjacent BM-scans were used for the variance calculations. Three B-scans were acquired within each BM-scan, producing two phase change measurements for each BM-scan. The A-line exposure time was $4.2\mu \mathrm{s}$ and the spectral data were saved in a binary file format for postprocessing in LabVIEW software. All images shown in this manuscript were acquired in vivo at a rate of 100,000 axial scans (A-scan) per second.

Each OCT data set (A-scan, already sampled linearly in $k$-space) was first aligned “in time” based on the reference FBG signal position. After finishing this step, fixed-pattern noise was removed by subtracting the averaged B-scan spectrum from each individual A-line spectrum. The resultant spectra were windowed to minimize side-lobes of the coherence function by apodization with a flat top Gaussian window. Residual mismatch between sample and reference arm dispersions was numerically compensated by automatic dispersion compensation based on an entropy minimization method.¹⁶ Finally, a complex SSOCT image with intensity and phase information was obtained after Fourier transformation of the processed and zero padded (to 4096 points) data.

OCT signal sensitivity fall-off profiles for our 1060-nm system were measured using standard procedures^6,7 and reported in our previously published report.²³ Each B-scan consisted of 440 A-scans with 375 A-scans acquired over a 1.5-mm lateral scanning range on the retina and 65 “fly-back” A-scans. The dynamic range of the intensity image was 35 dB and the phase values varied from $-\pi$ to $+\pi$. The basic concept of the pv-based motion contrast method has been described in our previous papers.¹⁹ The acquired and preprocessed OCT data sets were postprocessed by software developed at the Biological Imaging Center at the California Institute of Technology to extract pv signals. The cross-correlation function between consecutive intensity images aligns each B-scan from a BM-scan in an axial direction. Phase unwrapping on phase difference values, bulk motion removal, and histogram-based thresholding were implemented to remove phase shifts caused by eye and head movements. To visualize depth-resolved vasculature networks projections in the chorioretinal complex, the three–dimensional (3-D) pv data sets were manually segmented into three layers as per prior literature. The en face projection view of the segmented data produced two-dimensional (2-D) vascular perfusion maps, which are similar to the ones obtained with fundus FA. For comparison, FA and indocyanine green angiography (ICGA) images were acquired with the Topcon (TRC-50IX) fundus camera and Heidelberg Spectralis™ $\mathrm{HRA}+\mathrm{OCT}$. The images consisted of $1536(\mathrm{H})\times 1536(\mathrm{V})$ pixels over a 30-deg field-of-view (FOV) on the retina. All of these clinical images were taken with fully dilated pupils. Pseudocolor [red-green-blue (RGB)] depth coding of the processed pv-OCT volumetric data sets was used to better visualize 3-D vessel networks and its projection view created 2-D color vasculature maps linearly color coded by axial depth location of the vessels.

2.3.

Phase Stability Analysis

The accuracy of pv measurements and the removal of fixed-pattern noise both depend on the phase stability of the SSOCT system. Phase stability can be measured by analyzing the phase difference between successive A-lines as measured at the same location on a stationary mirror.^11–13 The phase noise, ${\sigma }_{\mathrm{\Delta }}$, defined as the standard deviation of phase difference values, follows a Gaussian probability distribution. The phase stability performance (as a function of path-length difference between sample and reference arms) was measured by phase difference values from the stationary mirror. At each depth (mirror position), 4000 subsequent A-lines were taken which were split into four data sets of 1000 A-lines. The phase-noise ${\sigma }_{\mathrm{\Delta }\varnothing }$ of each data set, plotted in Fig. 2, was calculated by taking the standard deviation of the phase difference values as well as the mean and standard deviation over ${\sigma }_{\mathrm{\Delta }\varnothing }$ from all four data sets. The prestabilized data were calculated without FBG reference signal correction between the A-lines. The measured values for ${\sigma }_{\mathrm{\Delta }\varnothing }$ are plotted against depth as red symbols for the case-with and as blue dots for the case-without phase stabilization. The measured standard deviations of these measurements are displayed as vertical error bars. We can conclude from the figure that the phase noise is considerably lower (up to about 40 times at greater depths) when phase stabilization is used. In Fig. 2(b), a zoomed part of Fig. 2(a) (red), phase noise measured with phase stabilization is shown.

Fig. 2

(a) The phase noise in radians with (red) and without (blue) phase stabilization is plotted against depth of the mirror placed in the sample arm. Error bars denote the standard deviation of the measurements. The phase noise is significantly lower when phase stabilization is used, especially at greater depths. (b) The zoom-in view of the phase noise measured with phase stabilization.

A mean phase noise value of 30 mrad was measured at a depth of 1.5 mm. An SNR value of 50 dB was used to calculate the phase-noise in Fig. 2. This is comparable to previously published results for SSOCT systems phase noise (e.g., Hendargo et al.¹³ reported 18 mrad for 53-dB SNR, and Baumann et al.¹⁰ reported 97 mrad at 100 kHz and 47 mrad at 200 kHz for a 35-dB SNR).

2.4.

Imaging Performance

Here, we present images demonstrating the effects of fixed pattern noise removal on intensity-based OCT and pv-OCT images.

2.4.1.

Fixed-pattern noise removal

The effect of fixed-pattern noise removal is shown in an intensity image of the macula of a healthy volunteer (Fig. 3). The image has a width of 5 mm consisting of 1500 A-lines and a displayed depth of 2.8 mm.

Fig. 3

Fixed-pattern noise removal on an SSOCT intensity image of the macula of a healthy volunteer (Subject 1). (a) The image without fixed-pattern noise removal when the phase stabilization method is not used. Artifact (coherent noise) lines are clearly visible above the retina. (b) Fixed-pattern noise removal on a phase-stabilized image. Scale bar: $500\mu \mathrm{m}$.

The image quality is degraded at several depths by fixed-pattern noise lines [yellow arrow Fig. 3(a)] when no phase stabilization correction is used. The phase-stabilization method allowed DC removal to suppress these artifact lines (fixed-pattern noise) very effectively [Fig. 3(b)].^10–12 With a mirror as an object in the sample arm, the fixed-pattern noise suppression was measured at different depths and showed a reduction of 40 dB at the zero-delay point (data not shown).

2.4.2.

Fixed-pattern phase-noise artifact removal in flow mapping

The need to apply the phase-stabilization method before pv-OCT processing is shown in Fig. 4 on B-scans acquired at 6-deg temporal retina. The intensity images are shown in Figs. 4(a) and 4(d). Each intensity B-scan image covers a lateral area of 1.5 mm consisting of 375 A-lines, a displayed depth of 1.5 mm, and a dynamic range of 35.0 dB. The corresponding pv images [Figs. 4(b) and 4(e)] have the same dimensions. A 2-D median filter ($3\times 10\text{pixels}$) was used for noise reduction. In Fig. 4(b), the pv image is shown when the phase-stabilization method was not used, which resulted in horizontal (yellow arrow) as well as vertical (green arrow) phase-artifacts throughout the B-scan, obscuring a clear view of the detected blood flow. These artifacts are, however, absent when the phase-stabilization method was used as shown in Fig. 4(e). The same effect can be observed on en face projection images in Figs. 4(c) and 4(f).

Fig. 4

Top and bottom panels show images before and after phase stabilization, respectively. Average intensity image of three B-scans (a and d). pv processed image using phase data from the same three B-scans (b and e), en face projection of retinal vasculature (c and f); yellow dashed line in panel (f) shows the location of B-scan. Vertical phase artifact lines (yellow arrows) and horizontal lines (green arrow) can be seen throughout the images processed without phase stabilization (a and b). Scale bar: $300\mu \mathrm{m}$.

3.1.

Three-Dimensional Imaging of the Chorioretinal Vascular Network

To extract chorioretinal vasculature from BM-scans, pvOCT data were acquired over a $1.5\times 1.5\text{-}{\mathrm{mm}}^2$ area of the human retina with an A-scan sampling spacing of $4\mu \mathrm{m}$ (between consecutive A-scans) for clear visualization of microcirculation. The scanning beam was focused on the RPE to improve visualization of the outer retinal layers and the blood supply in the choroid and choriocapillaris. An average intensity image composed of three successive B-scans from a single BM-scan is shown in Fig. 5(a). Figure 5(b) shows the image postprocessed with the pv method from these data. The composite image with averaged intensity (gray image) and pv (red image) is shown in Fig. 5(c). There was no significant eye motion during BM-scan acquisition due to the short acquisition times of each B-scan (approximately 3.5 ms). The composite image validates that the pvOCT identifies most perfused vessels as identified on the averaged intensity images where vessels can be classified based on the change in intensity (bright dots located near inner and outer plexiform layers). One can also observe a shadow artifact that generates residual pv values below vessels. We think that this is caused by variation in the optical path-length (phase) of photons passing through blood flowing in these vessels. This effect creates nonzero pv values calculated in retinal layers where no perfusion should be observed, including in the retinal-pigmented epithelium and deep choroidal layers. There is also loss in the pv signal from deeper choroid (larger vessels in Haller’s layer) due to very fast blood flow; they appear as dark areas in pv images [yellow circles in Fig. 5(b)]. More pv signal in deeper choroid was observed compared to our previously reported SD-OCT-based pv.¹⁹ This was possible due to the higher OCT sensitivity in SSOCT, and the use of a 1060-nm central wavelength increased penetration depth, thus improving contrast below the RPE. Additional benefits include extended measurement range in depth due to short instantaneous line-width of the source and permitting OCT imaging at distances further away from the zero path-length difference than is possible with spectrometer-based OCT systems. Thus, even with modest axial head motion of the patient from the forehead rest during automatic acquisition of several volumes, image quality was not impaired by relative changes of the location of the imaged structures with respect to the reference arm position.

Fig. 5

Images of human retina acquired over 1.5 mm. (a) Average intensity of three B-scans from one BM-scan sequence. (b) pv image coded as red. (c) Composite B-scan image showing flow signal in different layers overlaid over intensity image (gray) (a) and pv image (red) (b). The imaging acquisition time of a single BM-scan containing three B-scans is $\sim 10.5\mathrm{ms}$. Yellow ovals are drawn around vessels in Haller’s layer. Scale bar: $300\mu \mathrm{m}$.

3.2.

Pv-SSOCT Vascular Network Imaging of Chorioretinal Complex in Normal Eyes

Eyes from two normal subjects were imaged with the pv-SSOCT system. The images illustrated in Fig. 6 were acquired over $1.5\times 1.5{\mathrm{mm}}^2$ at two eccentricities (fovea and 6-deg temporal retina). We compared maximum projection views of pvOCT in Figs. 6(c) and 6(f) to the fundus FA in Figs. 6(b) and 6(e), respectively, acquired at the same location. Figures 4(b) and 4(e) show dense microcapillary networks around the foveal avascular zone and temporal zone, not distinctly recognizable in the zoomed fundus FA image. The size of the smallest resolvable capillary in this acquisition scheme was $\sim 10$ to $12\mu \mathrm{m}$. Figures 6(d) and 6(g) show a depth color-coded projection view. The pseudocolor (RGB) map codes the linearly scaled axial location of the vessels.

Fig. 6

In vivo human retinal vasculature images ($1.5\times 1.5{\mathrm{mm}}^2$) from a normal subject at foveal (upper panel) and 6-deg temporal retina (lower panel). (a and d) Fluorescein angiography (FA) images. (b and e) Projection view of retinal contributions to pvOCT. The numbers of A-scans and B-scans are 375 and 360, respectively, (c and f). RGB depth color-coded projection view of pvOCT data. Upper right shows depth scale bar with red color denoting a retinal layer, green showing capillaries as choroidal vascular bed, whereas blue represents capillaries in deeper choroidal layers. The imaging acquisition time was 3.6 s. Scale bar: $500\mu \mathrm{m}$. Video 1 shows C-scan fly through the three-dimensional pvOCT data set (MOV 4066KB) [URL: http://dx.doi.org/10.1117/1.JBO.19.12.126010.1]..

The superficial capillary network is observed mostly within the nerve fiber layer or the ganglion cell layer. In addition, the depth-resolved intermediate capillary plexus, between the inner plexiform layer and the inner nuclear layer as well as the deeper vascular plexus located between the outer plexiform layer and the outer nuclear layer, can be distinctly observed. The attached gray-scaled video (Video 1) demonstrates the depth morphology of the vasculature using the pvOCT data from the nerve fiber layer to the deeper choroidal layer.

A large FOV of the retinal layer was also presented to compare the corresponding FA image in Fig. 7. To maintain high-resolution in a single volume acquired in a few seconds, the scanning patterns were set to cover the $1.5\times 1.5\text{-}{\mathrm{mm}}^2$ scanning area. Several volumes were acquired at different retinal locations. Figure 7 presents the large FOV image overlaid on the fundus FA. This image was made from ten $1.5\times 1.5{\mathrm{mm}}^2$ pvOCT volumes acquired by shifting the position of light emitting diode fixation points. After pvOCT processing, each projection image was manually aligned with linear translation to create a large FOV mosaic.

Fig. 7

Large field-of-view (FOV) stitched pvOCT imaging overlaid on a fundus FA. (a) FA. (b) Projection view of retinal contributions to pvOCT imaging with the 10 volumes.

Several groups have previously demonstrated images of choricapillaris and Sattler’s layer with different OCT flow contrasting methods.^22,24–26 Our results are comparable to those reports. As an example, Fig. 8 shows images of the chorio-retinal perfusion of another normal eye from a different subject. A high sampling density angiogram of retinal, choriocapillaris, and Sattler’s layer were shown in Figs. 8(a), 8(b), and 8(c), respectively. Figures 8(f) and 8(g) show a dense vascular network right below the fovea. Choriocapillaris is usually described as a monolayer vascular network [Figs. 8(d) and 8(e)] with a mesh-like structure with pore size ranges from 5 to $40\mu \mathrm{m}$ (images adapted from Olver et al.²⁷). Choriocapillaris angiograms generated by our pv-SSOCT method [Figs. 8(f) and 8(g)] show a similar mesh-like network with small black pores where the pv-SSOCT signal is absent. Several pores were indicated by yellow arrows. The average cross-sectional diameters of the pores are $26\mu \mathrm{m}$. This value is similar to our pore size. The pores beyond $17\mu \mathrm{m}$ are not visible due to limitations of the lateral resolution of our current system. From the above results and similarities between the pv-SSOCT angiogram of the choriocapillaris and the scanning electron microscopy image, we can conclude that Fig. 8(f) represents choriocapillaris perfusion.

Fig. 8

Mosaic view of pvSSOCT imaging of a normal subject (2) from central foveal region. (a) Projection view of retinal layers (b) Stitched OCT angiograms of the choriocapillaris, and (c) Sattler’s layer. Scale bar: $400\mu \mathrm{m}$. (d) In-vitro scanning electron microscopy image from a methyl methacrylate cast of an excised eye27 which shows the subfoveal choriocapillaris microstructures in high-resolution for a $1.4\times 1.4\text{-}{\mathrm{mm}}^2$ area. (e) A magnified view on a $450\times 450\text{-}{\mu \mathrm{m}}^2$ area of the choriocapillaris angiogram [marked yellow in (d)] to show individual pores in detail (indicated with three yellow arrows). (F) In vivo angiogram of the choriocapillaris with pvSSOCT showing a similar mesh-like network of small vessels with small black pores for a $1.5\times 1.5\text{-}{\mathrm{mm}}^2$ area (volume section highlighted with red dotted lines in (b). (g) A magnified view from (f) for a $450\times 450\text{-}{\mu \text{m}}^2$ area [marked yellow in (f)] in which yellow arrows indicate pores with a similar size as those observed in (e).The choriocapillaris angiogram of (d) shows similar microstructures as in (f). The yellow scale-bars are $300\mu \mathrm{m}$ in length; the black scale-bars are 50 μm in length.

3.3.

Pv-SSOCT Vascular Network Imaging of Optic Nerve Head (ONH) Area

In vivo imaging of vasculature around the optic nerve head region was also tested with our pv-SSOCT system. A depth-resolved pv-SSOCT and corresponding FA images are compared in Fig. 9. Video 2 demonstrates depth-resolved en face OCT angiography and virtual C-scan fly through. Several blood vessels are visualized with the progression of scans through the optic nerve head. pv shadows created by blood vessels are clearly seen in the movie, tracking with the vessels entering the optic nerve head. The retinal capillary network (within the nerve fiber layer to the outer nuclear layer) is coded as red. The 3-D vascular network can be visualized further by generating a 3-D reconstruction of the optic nerve head (Video 3). A better visualization of branches of the vascular tree is observed compared to FA.

Fig. 9

Retinal capillary network at optic nerve head area ($1.5\times 1.5{\mathrm{mm}}^2$). (a) FA. (b) Total projection view of pvOCT (c) RGB depth color-coded projection view of pvOCT data. Upper right shows depth scale bar with red color denoting a superficial layer, green showing capillaries in an intermediate vascular bed, and finally blue represents microcapillaries as a deeper vascular layer. Scale bar: $250\mu \mathrm{m}$. Video 2 shows C-scan fly through the data and Video 3 shows volumetric visualization of pvOCT data set (MOV 9058KB) [URL: http://dx.doi.org/10.1117/1.JBO.19.12.126010.2].; (MOV 2182KB) [URL: http://dx.doi.org/10.1117/1.JBO.19.12.126010.3].

3.4.

Comparison between ICGA and pv-SSOCT Vascular Network Imaging of Chorioretinal Complex

A comparison between pv-SSOCT projections and corresponding ICGA images is presented in Fig. 10. The flow signal from different retinal and choroidal layers has been superposed on the intensity image shown in a composite B-scan image [Fig. 10(a)]. A volumetric manual segmentation [Fig. 9(a)] was performed to extract each vascular layer based on the existing literature.^12,13,22,24 The retinal vascular network (within the nerve fiber layer to the outer nuclear layer) is coded as red [Fig. 10(c)]. Choriocapillaris marked at different depths ($6\mu \mathrm{m}$ below Bruch’s membrane coded as gray) were generated from the OCT pv volume [Fig. 10(d)]. Here, shadow artifacts from large retinal vessels (marked with yellow arrows) are observed. They are not a part of the choroidal vasculature. Figure 10(e) shows an en face angiogram extracted from the same volumetric data sets from 26 to $34\mu \mathrm{m}$ below Bruch’s membrane, where feeding arterioles and draining venules in Sattler’s layer are clearly visualized. Haller’s layer, consisting of larger feeder vessels located from 61 to $90\mu \mathrm{m}$ below Bruch’s membrane, is shown in Fig. 10(f).

Fig. 10

pv-SSOCT volumetric scan over $1.5\times 1.5{\mathrm{mm}}^2$ at 6-deg temporal retina (from subject 1). (a) Composite B-scan image showing flow signal in different retinal and choroidal layers overlaid over intensity image, and (b) corresponding FA image, segmented depths for en face projection images are: (c) retinal layers, (d) $6\mu \mathrm{m}$ below Bruch’s membrane for choriocapillaris (e) 26 to $34\mu \mathrm{m}$ below Bruch’s membrane for Sattler’s layer; (f) 61 to $90\mu \mathrm{m}$ below Bruch’s membrane for Haller’s layer; (g) corresponding indocyanine green angiography (ICGA) image, (h) total en face projection of choriodal layer only in gray scale, and, (i) depth color-coded en face projection. Yellow arrows in (d) show the shadow artifact from large retinal vessels projected onto the choriocapillaris layer. Scale bars, $300\mu \mathrm{m}$. Morphological features here resemble histological images by corrosion vascular casts and scanning electron microscopy.²⁷

To generate an en face map of Haller’s layer, the following method was used. The segmented C-scans from Haller’s layer were projected with minimum intensity, and then the image intensity scale was inverted. The reason for this image processing step was explained in Sec. 3.1. Large vessels in Haller’s layer appear black (no detected flow signal) in our pv-SSOCT data. A minimum threshold was applied in the above process; hence, the displayed pvOCT signal is generated by large vessels in Haller’s layer. For comparison, the ICGA [Fig. 10(g)] and corresponding pv-SSOCT total choroidal projection in gray scale [Fig. 10(h)] are presented. The pseudocolor (RGB) depth-coded projection is also presented in Fig. 10(i). There is excellent correspondence of structural features between the OCT angiograms and the electron micrograph corrosion casts reported by Olver et al.²⁷ Deep penetration pv-SSOCT clearly permits in vivo visualization of the chorioretinal complex.

To illustrate clinically applicable images of choroidal flow, a large FOV pvOCT from different choroidal layer en face projections overlaid on the fundus ICGA image are presented in Fig. 11.

Fig. 11

Large FOV images of stitched pvSSOCT scans overlaid on a fundus ICGA (subject 1). (a) ICG angiography. (b) Projection view of retinal layers from 10 volumes, (c) stitched OCT angiograms of the choriocapillaris, (d) Sattler’s layer, (e) Haller’s layer, and (f) depth color-coded imaging. Upper right shows depth scale bar with red color denoting retinal vasculature, green showing Sattler’s layer, and blue denoting Haller’s layer.

At different retinal eccentricities, variations in choriocapillaris density and pattern are observed. A tightly packed choriocapillaris with a mesh-like structure is visualized at the central fovea [Fig. 11(c)]. However, the density is reduced at a higher eccentricity [Fig. 11(c)]. Similar kinds of variations in vessel diameter and pattern are also observed in Sattler’s layer [Fig. 11(d)] as well as Haller’s layer [Fig. 11(e)]. The density of the feeding vessel diameters in Sattler’s layer increases as we move out from the central foveal region. Larger choroidal vessels below the choriocapillaris layer are visible in pvSSOCT en face images because of the deeper penetration, high sensitivity, and high phase stabilization in our SSOCT system.

A major advantage of the high penetration $1\text{-}\mu \mathrm{m}$ pv-SSOCT technique over fluorescein or ICG angiography, similar to other OCT-based flow contrasting methods, is that it does not require any exogenous contrast agents that are uncomfortable or have occasional adverse reactions. Thus, our method is noncontact and noninvasive. Additionally, depth-resolved fine vascular structure located below the RPE can be clearly seen. The depth color-coded visualization of the retinal and choroidal vasculature shown in Fig. 11(f).which summarizes vascular maps segmented from retinal and chroidal layers: below the inner limiting membrane to the outer limiting membrane (retinal layer)—red, Sattler’s layer—green, and Haller’s layer—blue. Vessels present in the intermediate layer have a yellowish color.

3.5.

pv-SSOCT Vascular Network Imaging of Chorioretinal Complex in Diseased Eyes

Four AMD patients were imaged with high penetration $1\text{-}\mu \mathrm{m}$ pv-SSOCT. Figure 12 shows the pv-SSOCT imaging from patient 1 (AMD). The position of atrophy can be seen from Fig. 12(c) yellow dotted circle. A close look of composite flow images [Figs. 12(a), 12(b), and 12(b1)], provides detailed information under the RPE. In Figs. 12(b) and 12(b1), inconsistent choriodal flow was observed due to AMD and atrophy, and flow signals by microcirculation disappeared underneath Bruch’s membrane due to AMD (blue arrows [Figs. 12(b) and 12(b1)], whereas motion contrast signals by blood perfusion in the remaining choriocapillaris exists in the region pointed out by the yellow arrowhead. The green arrow head shows the position of the deformed RPE. Manual segmentation positions were presented as a dotted curved green line in Fig. 12(a). It was also observed that there is an absence of flow signals in the degenerative region (green arrows). The FA image [Fig. 12(c)] and corresponding mosaic view of the retinal projection layers in Fig. 12(d) are similar. Yellow circles on the angiograms of Sattler’s layer [Fig. 12(e)] represent the positions of ischemic flow regions. Haller’s layer was shown in Fig. 12(f).

Fig. 12

pv-SSOCT imaging of P1. (a) OCT composite B-scan image at volume scan over $1.5\times 1.5{\mathrm{mm}}^2$ at 3-deg nasal retina of the patient with age-related macular degeneration (AMD) with deformation of retinal pigment epithelium (RPE) (c, yellow dotted circle), with a retinal scan location as identified on the horizontal yellow dashed line (d). Dotted curved green line represents manual segmentation position (I-retinal layer, II-degeneration region, III-Sattler’s layer and IV-Haller’s layer). Yellow arrow shows deformed RPE; (b and b1) Two OCT zoomed composite B-scan images from the location of the red and green dotted lines from image (d), respectively. As indicated by the blue arrow, pv signals by microcirculation disappeared underneath Bruch’s membrane due to AMD, whereas motion contrast signals by blood perfusion (remaining choriocapillaris) exist in the region pointed by the yellow arrowhead. Green arrow head shows position of deformed RPE; (c) FA image at central foveal region; (d) projection view of retinal layers; (E) stitched OCT angiograms of Sattler’s layer, yellow dotted circle represents ischemic flow region; (f) Haller’s layer. (Scale bars, $300\mu \mathrm{m}$ and for b and b1 $100\mu \mathrm{m}$).

Another patient with reticular pseudodrusen (deposits above RPE) was imaged. The position of the drusen and corresponding flows were presented in Fig. 13(a). A cropped view (yellow rectangle in [Fig 13(a)] provides detailed information about the flow within and beneath the drusen area. It was observed that the microcirculation disappeared underneath Bruch’s membrane due to choriocapillaris atrophy [blue arrow [Fig. 13(b)], but blood perfusion in the remaining choriocapillaris exist in the region pointed out by the yellow arrowhead. The green arrow head shows the position of the RPE. A well-defined capillary network in the retina [Fig. 13(d)] was observed. However, a focal disruption/loss of choriocapillaris within the area of underneath drusen and atrophy region the inner choroidal layer [Fig. 13(e)] was clearly visible for the same subject. The region was highlighted with red dotted lines and compared with the corresponding FA image [Fig. 13(c)]. However, there is no flow signal observed within the drusen area.

Fig. 13

pv-SSOCT imaging of P2. (a) OCT composite B-scan image at volume scan over $1.5\times 1.5{\mathrm{mm}}^2$ at 3-deg temporal retina of the patient having reticular pseudodrusen (c, red dotted circle), with a retinal scan location as identified on the horizontal yellow dashed line (d). (b) an OCT zoomed composite B-scan images from the location of yellow dotted rectangle from image (a). As indicated by the blue arrow, pv signals by microcirculation disappeared underneath Bruch’s membrane due to choriocapillaris atrophy, whereas motion contrast signals by blood perfusion (remaining choriocapillaris) exist in the region pointed by the yellow arrowhead. Green arrow head shows position of RPE; (c) FA image at central foveal region; (d) projection view of retinal layers; (e) stitched OCT angiograms of Sattler’s layer, yellow dotted circle represents ischemic flow region; (Scale bars, $300\mu \mathrm{m}$ for a, b; for c, d, e $400\mu \mathrm{m}$).

Alterations in chrorioretinal perfusion were also observed in layers from two other AMD patients (Fig. 14, upper panel and lower panel). There was deformation of the RPE layer for both patients [green arrow, Figs. 14(a) and 14(d)]. A well-defined flow in the retinal layer was present, as observed in Figs. 14(b) and 14(e). However, there is a loss of flow signals observed in Sattler’s layer [Figs. 14(c) and 14(f)]. Missing and abnormal vessel perfusion areas were highlighted with red dashed circles in Figs. 14(c) and 14(f).

Fig. 14

pv-SSOCT imaging of P3 (upper panel) and P4 (lower panel). (a and d) OCT composite B-scan image at volume scan over $1.5\times 1.5{\mathrm{mm}}^2$ center foveal retina of the patient having AMD, with a retinal scan location as identified on the horizontal yellow dashed line (b) and (e), respectively; green arrow shows position of atrophy and RPE deformation; (b and e) Projection view of retinal layers; (c and f) Stitched OCT angiograms of Sattler’s layer, red dotted circle represents ischemic flow region; scale bars, $400\mu \mathrm{m}$.