2.1.

Setup of Dynamic-Focus SS-OCT

Figure 1 shows the schematic of the dynamic-focusing SS-OCT system. The swept laser has a central wavelength of 1310 nm, a 100 nm full sweep bandwidth, and a sweeping rate of 50 kHz (Axsun Technologies Inc.). The light from the laser is split by an 80:20 fiber splitter into the reference arm and the sample arm. In the sample arm, the collimated light passes through a telescope formed by an electrical tunable lens (EL-10-30-NIR, Optotune, Dietikon, Switzerland) and a chromatic doublet L2 (focal length 70 mm). After passing through a dual axial galvanometer mirror scanner, the beam diameter was expanded by another telescope (L3 with effective focal length of 30 mm and L4 with effective focal length of 50 mm). After reflecting by a dichroic mirror, the beam is focused by another achromatic double L5 with a focal length of 75 mm onto the sample. The light backscattered from the sample is collected by a circulator and interferes with a reference light reflected from the reference mirror and collected by another circulator. The final interference signal is detected by a balanced photodetector with a bandwidth of 200 MHz (PDB460C, Thorlabs Inc., Newton, New Jersey). The final signal is then digitalized by a high-speed digitizer (ATS9350, Alazar Technologies Inc., Pointe-Claire, QC, Canada). The laser power at the cornea is measured to be 4.59 mW and is within the safety limit set by the American National Standards Institute.¹⁴ The measured system sensitivity is 104 dB around the zero path delay location. The laser provides a clock with a frequency of 91 to 134 MHz. A frequency doubler is used to double the laser clock so that the axial imaging range was extended to around 10 mm in the current setup.¹⁵ The OCT system has a lateral resolution of $\sim 15\mu \mathrm{m}$ and axial resolution of $8.52\mu \mathrm{m}$ in tissue. The controlling software can perform automatic focusing, and the operator can switch between dynamic-focusing mode and normal mode. In dynamic-focusing mode, the electrical current applied on the tunable lens was changed automatically for 10 consecutive B-scans, and these 10 consecutive B-scans form a tuning cycle. Within a tuning cycle, the current is tuned from 100 to 390 mA with an increase of 32 mA for each B-scan.

2.2.

In Vitro Characterization of Dynamic Focus

In order to test the dynamic-focusing performance of the system, 0.1% intralipid was imaged. The 0.1% intralipid solution was obtained by diluting the 20% stock solution (Sigma-Aldrich) with deionized water. The intralipid solution was held in a cuvette and imaged with the dynamic-focusing mode by the dynamic-focusing SS-OCT system. The results [Figs. 2(a)–2(d)] extracted four B-scan images in a dynamic-focusing tuning cycle, and they show that the system can dynamically change the axial focusing plane. These images show the results when the focal planes are, respectively, at the surface, upper middle, lower middle, and bottom of the sample. Figure 2(f) shows the four normalized A-line profiles extracted from the four representative B-scan images in Figs. 2(a)–2(d). It can be seen that the peak intensity location (which corresponds to the focal plane location) moves with the changing of electric current to control the lens. The images of the intralipid acquired at different axial focal plane locations are filtered by Gaussian windows to remove the out-of-focus regions part and then averaged to generate a composite image. The final composite image is shown in Fig. 2(e). The axial intensity profile of the composite image is shown in Fig. 2(g), and it can be seen that a strong OCT signal of from 1 to 7 mm is obtained along the axial direction. The composite image in Fig. 2(e) shows the better image quality (higher signal) along the whole image depth.

Fig. 1

Schematics of dynamic-focusing swept-source optical coherence tomography (SS-OCT) system. Cir, circulator; MP, miniprojector; EL, electrical lens; and L1–L7 as the lens number from 1 to 7.

Fig. 2

SS-OCT images of 0.1% intralipid solution obtained at different focal depths. (a)–(d) Images obtained with focal planes at the surface, upper middle, lower middle, and bottom region of intralipid solution. (e) The composite image generated by fusing the Gauassian-weighted window filtering of images (a)–(d). (f) The axial intensity profile for images in (a)–(d). (g) The axial intensity profile for the composite image in (e).

The sensitivity of the OCT system is proportional to the power returning from the sample. The tissue around the focal plane location will reflect (and scatter) more light due to the high laser power density in the region. So the focal plane region in the OCT image will show better sensitivity. By refocusing the beam into different parts (depths in this setup) of the sample, a higher sensitivity region is tuned accordingly. By combing the high-sensitivity parts of different images acquired from the same sample, a higher sensitivity image is obtained. From Fig. 2(g), it can be found that the composite image combines all the high-sensitivity regions in Figs. 2(a)–2(d) and shows an overall higher signal and better sensitivity. The enhanced sensitivity will allow the OCT system to penetrate deeper. Compared with Fig. 2(a), the composite image in Fig. 2(e) shows increased penetration depth. This increase is a result of using the DOF region in Fig. 2(d). Quantitative analysis shows an increase of more than 20% in penetration depth. This increase is dependent on the focal plane location. The relationship between focal plane depth and the electric current applied to the electrical lens is shown in Fig. 3.

Fig. 3

Relationship between focal plane depth and current applied to the electrical lens.

To demonstrate the improvement of lateral resolution along the whole axial imaging range by the dynamic-focusing SS-OCT system, microbeads are imaged. The polymer microbeads (Duke Scientific Corporation, Lot No. 26386) used in this study have a diameter of $9.6\mu \mathrm{m}$, and they were embedded into 5% (mg/mL) agarose (Sigma-Aldrich, A3893-25G). The sample was held in a cuvette and imaged using the dynamic-focusing mode with the SS-OCT system. Figures 4(a)–4(d) demonstrate the four frames extracted from the images acquired in a focus tuning cycle. The dotted yellow lines in the images indicated the locations of the focusing plane. Due to the limited DOF of the imaging optics, the regions that were out of the DOF in these images were severely blurred, and the microbeads appeared larger or unresolvable in those regions. By filtering of the out-of-DOF regions using Gaussian windows (white curves in the white dashed box overlaid on images) in Figs 4(a)–4(d) and then averaging these filtered images, a composite image as shown in Fig. 4(e) was obtained. A total of nine images were used for averaging to get the composite image in Fig. 4(e), and the white curves overlaid on the image show the nine Gaussian windows that were used to filter the images. Microbeads were clearly resolved along the whole depth range in the composite image. This shows that dynamic-focus OCT can extend the DOF and maintain optimized lateral resolution along the whole imaging depth.

Fig. 4

Imaging microbeads by dynamic-focusing SS-OCT system. (a)–(d) The images of microbeads obtained when the focal planes were set from the surface to the bottom of the cuvette. (e) The composite image by averaging the Gaussian-filtered image frames. Microbeads diameter: $9.6\mu \mathrm{m}$. The scale bars are 1 mm.

By maintaining the optimum lateral resolution along the whole axial imaging range, the dynamic-focusing SS-OCT improves the lateral resolution for the regions that are out-of-DOF of the focusing optics. The degree of improvement on the lateral resolution for the out-of-DOF region is related to focal plane location and the axial distance of the region to the focal plane. For a Gaussian beam, the beam spot size changes with the propagation of the beam along the beam axis and the beam spot at an axial distance $z$ from the beam focal plane is

2.3.

In Vivo Imaging and Image Process

This study followed the tenets of the Declaration of Helsinki and was in accord with the Health Insurance Portability and Accountability Act of 1996. The study protocol was approved by the Institutional Review Board (IRB) of OHSU. The subject in this study was a 25-year-old male with healthy eyes and dark iris pigmentation (brown eyes). Written informed consent was obtained.

The SS-OCT system was used to image the human anterior chamber in vivo. For dynamic-focusing imaging of the sample, 150 repeated B-scan images were obtained in 3 s. Each B-scan frame contained 1024 A-scans and covered a lateral range of about 19 mm. A postprocessing phase stabilization method was used to process the OCT data to eliminate the horizontal line artifacts in the images.¹⁶ The 150 B-scan images were further processed to obtain a high-resolution image. Figure 5 shows the flowchart of the image-processing steps. First, images with large eye motion artifacts were removed from the image set. Then the 10 consecutive frames of one tuning cycle were registered by calculating the cross-correlation between these images [Fig. 5(a)]. The DOF region of each image was selected, and Gaussian window was performed on the region [Fig. 5(a)]. The 10 images were then combined to form a composite, high-resolution image [Fig. 5(a)].^8,9 The width of the Gaussian weighting function was 1 mm in our setup. These steps were repeated for different tuning cycles and a few composite images were obtained. A maximum total of 15 composite images could be obtained. Finally, these composite images were registered and averaged to generate a final image [Fig. 5(b)]. A dewarping algorithm was used on the final image to correct the distortion caused by the index transition at the air–tissue interface [Fig. 5(b)].¹⁷

Fig. 5

Flowchart of dynamic-focusing OCT image process. (a) Procedure to get the composite image from the images in a focus tuning cycle. (b) The procedure in (a) is repeated for different tuning cycles to get different composite images. The composite images are further registered and averaged to get a high-resolution image with reduced speckles. The high-resolution image was dewarped to get the final image.

2.4.

Compare the Dynamic-Focusing SS-OCT with Time-Domain OCT, Spectral OCT, and UBM

To evaluate the results from our custom-built SS-OCT system, the anterior segment of the normal human subject eye was also imaged using commercially available imaging systems: a TD-OCT system (Visante, Carl Zeiss Meditec Inc., Doublin, Caliafornia), a spectral OCT system (Avanti, Optovue Inc., Fremont, California), and a UBM (Eyecubed, Ellex Inc., Minneapolis, Minnesota), respectively. The TD-OCT system has a central wavelength of 1310 nm, a speed of 2048 A-scan/s, a lateral resolution of $60\mu \mathrm{m}$, and an axial resolution of $18\mu \mathrm{m}$. In this study, the anterior segment of the subject was imaged with the built-in scan mode (enhancement anterior segment single scan) in which four B-scan images are registered and averaged. The width of the B-scan was 16 mm, and the axial imaging range in the tissue was 6 mm. The number of A-scans per B-scan for the TD-OCT system was 256. The spectral OCT system used in this study has a central wavelength of 840 nm, a lateral resolution of $22\mu \mathrm{m}$, and an axial resolution of $5\mu \mathrm{m}$. The width of the B-scan was 8 mm, and the axial imaging range in the tissue was 3 mm. The number of A-scans per B-scan was 1020. A total of 48 B-scan images were registered and averaged. The UBM used in this study has a lateral resolution of $33\mu \mathrm{m}$ and an axial resolution of $23\mu \mathrm{m}$. The scanning angle was 30 deg and the axial imaging range in the tissue is 11.9 mm. The acquisition rate was 12.5 B-scans/s. A total of 13 B-scans were used for registration and averaging. The parameters used for different devices in this study are typical clinic setting parameters.

3.1.

Whole Anterior Chamber Imaging

The anterior chamber of a healthy human subject was first imaged by manually changing the focal plane. The results are shown in Fig. 6. Focal planes were manually tuned from the anterior surface of the cornea to the anterior surface of the crystalline lens as indicated by solid arrows in Figs. 6(a)–6(c). The crystalline lens and the periphery part of the cornea (white dashed arrows) were not clearly visible when the focusing plane was at the anterior corneal surface [Fig. 6(a)]. These parts were better visualized when the focusing plane moved to the middle of the anterior chamber [Fig. 6(b)]. When the focusing plane was set at the anterior surface of the crystalline lens, the anterior corneal surface signal intensity was low and the crystalline lens signal intensity became higher [Fig. 6(c)]. It can be seen that the top corneal surface was best visualized in Fig. 6(a). The periphery part of the cornea was best shown in Fig. 6(b). The ciliary body was best demonstrated in Fig. 6(c).

Fig. 6

Imaging of the anterior segment of a normal human subject with manual focus SS-OCT. (a) Focus on anterior surface of the cornea, (b) focus on the middle of anterior chamber, and (c) focus on the anterior surface of crystalline lens. Focal planes are indicated by the solid arrows. The crystalline lens and the periphery part of the cornea (dashed arrows) were not clearly visible when the focusing plane was at the anterior corneal surface (a). These parts were visualized better when the focusing plane moved to the middle of the anterior chamber (b). When the focusing plane was set at the anterior surface of the crystalline lens, the anterior corneal surface signal intensity was low and the crystalline lens signal intensity became higher (c). The scale bar is $500\mu \mathrm{m}$ for all images.

The anterior segment eye of the subject was then imaged using dynamic-focus SS-OCT. The composite image of the anterior segment of a normal human subject was shown in Figs. 7(a) and 7(b). Anterior chamber angle recess, ciliary sulcus, and adjacent structures could be visualized from the image. The sulcus-to-sulcus distance was measured to be 10.93 mm in the 1310 nm SS-OCT image [Fig. 7(b)].

Fig. 7

Imaging of the anterior segment of a normal human subject. (a) and (b) Dynamic-focus SS-OCT images without and with annotation, (c) images obtained with 40 MHz ultrasound biomicroscopy (UBM), and (d) image obtained with Visante TD-OCT. Both SS-OCT and Visante TD-OCT operate at the 1310-nm wavelength. CCT, central corneal thickness; ACD, anterior chamber depth; and PD, pupil diameter. Scale bars are $500\mu \mathrm{m}$ for all images.

The angle recess and ciliary sulcus were visible on the UBM image [Fig. 7(c)]. The crystalline lens was also visible in the UBM image. The sulcus-to-sulcus distance was measured to be 11.01 mm on the UBM. The angle recess was clear on the TD-OCT, however, the crystalline lens and ciliary sulcus were not visible on the TD-OCT [Fig. 7(d)].

3.2.

Transcleral Ciliary Body Imaging

The iris root, trabeculum, ciliary body, ciliary process, and ciliary sulcus could be visualized on the UBM [Fig. 8(a)]. The iris root, trabeculum, and most of the ciliary body could be visualized on the 1310 nm dynamic-focus SS-OCT [Fig. 8(b)] and the 1310 nm TD-OCT [Fig. 8(c)], but the ciliary process and sulcus could not be seen. The iris root, trabeculum, and ciliary body cannot be visualized on the 840 nm spectral OCT [Fig. 8(d)]. The contour of the ciliary body epithelium was better defined by the dynamic-focus SS-OCT compared with the UBM.

Fig. 8

Imaging of the ciliary body of a normal human subject to UBM and three different OCT systems. (a) Image obtained with 40 MHz UBM, (b) image obtained with dynamic-focus SS-OCT at 1310-nm wavelength, (c) image obtained with Visante TD-OCT at 1310-nm wavelength, and (d) image obtained with spectral OCT at 840-nm wavelength. Scale bars are $500\mu \mathrm{m}$ for all images.