2.1.

Hardware Configuration

A schematic of our FF-OCT system is shown in Fig. 1 . The system is based on a white light interference microscope with a Linnik-type configuration. The radiation from a 150-W tungsten halogen lamp is incidented into a flexible fiber bundle (not shown), and it is conducted to a Köhler illumination (KI) system that consists of four aspheric lenses and two variable apertures. The KI system is intended to be used for a uniform illumination to the sample, as in a conventional microscope. The output quasi-parallel beam from the KI system is launched into the Michelson interferometer and then split into the signal and reference beams by a 50:50 beamsplitter. Two identical infinity-corrected water immersion microscope objectives ( $40\times$ , $0.8\mathrm{NA}$ , LUMPLFL40XW/IR2, Olympus, Tokyo) with a working distance of $3.4\mathrm{mm}$ are placed in both arms of the interferometer. An achromatic quarter wave plate (QWP) and a linear polarizer are inserted in the reference arm. The light beam reflected back by a reference mirror is again passed through the linear polarizer and the QWP so that the light beam returning to the BS becomes a circularly polarized light. A drop of water (W3500 cell cultured water, Sigma-Ardorich, St. Louis, Missouri) is placed as an immersion fluid between the tip of the objective and the total reflection mirror. A neutral density (ND) filter whose transmittance is experimentally optimized to 13% is set in the reference arm to increase the visibility of the interferometric image. On the other hand, the sample is illuminated with a randomly polarized light. To compensate for a residual dispersion mismatch between the signal and the reference arms, a BK7 glass plate is placed in the reference arm so that the dispersion is initially matched. The recombined output beam from the interferometer is detected by the microscope setup, where the interference light is divided into two orthogonal parts by a polarization beamsplitter (PBS) before reaching the two identical CCD cameras (MC-512PF, Texas Instruments, Tokyo). The CCD cameras employed have $500\times 500$ pixels with a 12-bit resolution at an output rate of $30\mathrm{Hz}$ .

Fig. 1

Ultrahigh-resolution FF-OCT system based on a white light interference microscope using a dual-channel detection technique. BS: beamsplitter; QWP: quarter wave plate; RM: reference mirror; PZT: piezoelectric transducer; AS: aperture stop; PBS: polarization beamsplitter.

Provided that the sample exhibits no birefringence, the dual CCD outputs exhibit a phase difference of $\pi ∕2$ and can be expressed as follows:

To extract the sine and cosine components from the CCD outputs, a subtraction of the background component is needed. In the present setup, we measure an additional pair of images $C_{A2}(x,y)$ and $C_{B2}(x,y)$ with a phase shift of $\pi$ by using a piezoelectric transducer (PZT) that is glued to the reference mirror. An FF-OCT image is then derived by summing the squares of the sine and cosine components of the interference signal as follows:

To synchronize the dual-detection channels, common pulse signals are applied to the CCD cameras through a programmable camera control line in the camera link format. The alternative voltage changes applied to the PZT, which correspond to phase shifts of 0 and $\pi$ , are triggered by the frame-sync signal from one CCD camera so that every two consecutive CCD frames exhibit a relative phase difference of $\pi$ in the interference image.

The validity of Eq. 3 requires that the two CCD cameras capture exactly the same field. In other words, a pixel-to-pixel correspondence between the CCD cameras is needed. Mechanically, one camera is stationary, while the other camera that is mounted on a xyz translation stage is manually aligned to its counterpart with a subpixel accuracy. The alignment employs either home-written template matching or image-correlation software. This calibration is done only once in the initial setting of the system.

2.2.

Image Acquisition Software

We wrote a software program that performs three main functions: monitor mode for real-time observation, time-lapse mode for 2-D video-rate imaging, and z-stack mode for 3-D imaging. In monitor mode, a software sequence including image acquisition, processing, and display is repeated at a 5-Hz rate. Second, in time-lapse mode, FF-OCT images are continuously detected at a fixed depth as a function of time to achieve a time-sequence image. Assuming that the consecutive frames with a phase difference of $\pi$ are $C_1$ , $C_2$ , $C_3,\dots ,C_{n-1}$ , and $C_n$ , the frame subtractions of $C_1-C_2$ , $C_2-C_3,\dots ,C_{n-1}-C_n$ may effectively remove the background component to derive the sequential FF-OCT images using Eq. 3. Using such a rolling manner, video-rate $(30\mathrm{frames}∕\mathrm{s})$ imaging is made possible. To avoid the possible fringe averaging during in vivo imaging, the exposure time of the cameras is set to $6.4\mathrm{ms}$ , while the frame interval is fixed at $33\mathrm{ms}$ so that a single FF-OCT image can be acquired as short as $40\mathrm{ms}$ . Third, in z-stack mode, the sample is placed in a motorized z-translation stage (ALV-600-H0M, Chuo Precision Industrial, Tokyo) and is translated along the optical axis to yield a stack of FF-OCT images. In this mode, the exposure time is set to $33\mathrm{ms}$ . Using such a 3-D data set, a longitudinal cross section (xz- or yz-), which is of the common form of the conventional OCT tomogram, can be sectioned out by software. Illumination flux is also adjusted so that the camera pixels are close to the saturation level in both imaging modes.

2.3.

Basic Performance

Although a half-micron axial resolution can be achieved by fully exploiting the extremely broad bandwidth of the halogen lamp, only a partial band in the near infrared ranging from $600\mathrm{nm}$ to $1000\mathrm{nm}$ has been chosen. Optical filters are inserted at the exit of the KI system to cut off the visible and long wavelength components. To avoid any thermal injuries of the sample, a short pass filter (passband of $<1.1\mu \mathrm{m}$ ) is inserted as well. This wavelength region has been chosen by considering the weaker absorption and better penetration in most of the biological tissues, and also the wavelength dependency of the polarization optics employed. To reduce unwanted reflections, all optical components are antireflection coated for the wavelength range employed. The resultant spectrum is shown in Fig. 2a . The spectrum was measured at the exit of the interferometer by blocking the sample arm so that it contains a transmittance dependency of the overall optics. In the present study, a spectrally filtered source having a full-width at half maximum (FWHM) of $180\mathrm{nm}$ centered around $700\mathrm{nm}$ was employed, yielding a theoretical axial resolution of $0.9\mu \mathrm{m}$ in water. It should be noted that the effective spectrum for imaging is a product of the source spectrum and the spectral response of the CCD chip. Figure 2b shows the axial response profile of one pixel measured by scanning a total reflection mirror along the optical axis. As seen in Fig. 2b, the axial point spread function detected by FF-OCT exhibits a symmetrical form, with negligible sidelobes. The measured value of $0.8\mu \mathrm{m}$ (FWHM) is slightly better than the theoretical value, mainly because the spectrum is slightly different from the Gaussian distribution.

Fig. 2

(a) Spectrum of the filtered output of a halogen lamp employed for FF-OCT imaging, and (b) the axial response profile of one pixel measured by scanning a total reflection mirror along the optical axis.

The theoretical transverse resolution ( $\mathrm{\Delta }x$ , $\mathrm{\Delta }y$ ) is determined by the well-known formula $\mathrm{\Delta }x=0.66\lambda ∕\mathrm{NA}$ . An 0.8-NA objective employed yields $\mathrm{\Delta }x=\mathrm{\Delta }y=0.58\mu \mathrm{m}$ at the wavelength of $\lambda =700\mathrm{nm}$ , while we experimentally confirmed a transverse resolution of $0.7\mu \mathrm{m}$ using the edge response (10 to 90%) of a silicon wafer as a sample. In tissue imaging, however, aberration may degrade the transverse resolution when imaging at a deeper position due to multiple scattering.

The theoretical detection sensitivity for FF-OCT is described in the literature.^{20, 34} We followed the equation by considering a four-frame phase shift detection technique. A full-well capacity of the camera is an important factor to achieve a high sensitivity measurement. The CCD camera employed has a full-well capacity of 400,000 electrons, and the theoretical detection sensitivity is calculated to be $\sim 80\mathrm{dB}$ . The measured detection sensitivity was approximately $75\mathrm{dB}$ . Unwanted back-reflection from the optical components in the FF-OCT setup and an uneven ploarization splitting ratio by the PBS may account for a sensitivity loss.

A typical FF-OCT image acquired by our system consists of $500\times 500$ pixels covering an area of $215\mu \mathrm{m}\times 215\mu \mathrm{m}$ . The illumination flux to the biological tissue is approximately $3\mathrm{mW}∕{\mathrm{mm}}^2$ . In the measerement, where the exposure time of the CCDs was shortened to $6.4\mathrm{ms}$ , the incident power was adequetely increased.

2.4.

Sample Preparation

A Xenopus laevis (African frog) tadpole of stage 40 to 45 was chosen for the present study. Xenopus laevis is a widely used and well-characterized developmental biology animal model and is commonly used in OCT measurement to demonstrate the visualization of the morphological structure ^{35, 36, 37, 38} as well as the cardiovascular activity.^{39, 40}

Xenopus laevis tadpoles were bred in an aquarium tank at 23 to $25°\mathrm{C}$ until measurements were performed. The first group of specimens were anesthetized by immersion in diluted ketamine solution for $10\min$ until they no longer responded to touch. The second group of specimens were fixed in 4% buffered solution of formaldehyde for an hour until no cardiac activity was observed. The tadpoles were transferred to a Petri dish and then submerged in water at room temperature. Specimens were oriented for imaging with the optical beam incident from either the dorsal or ventral sides. No staining and contrast agents were used. All animal handling was performed according to protocols approved by the Committee on Animal Care, Yamagata Promotional Organization for Industrial Technology.

3.1.

Time-Lapse Imaging

To demonstrate the video-rate imaging capability of our FF-OCT system, Fig. 3 shows examples of the FF-OCT images of an anesthetized tadpole at approximately 0.3-s time interval. Figures 3a and 3d are detected where the speed of the blood flow is minimum. It can be seen from these images that individual blood cells are visible as well as the cellular structure of the tadpole. Time-lapse imaging results can be seen in 1Video 1 . The movie shows a total of 300 FF-OCT images recorded during a period of $10\mathrm{s}$ . Outputs from the CCD cameras were continuously captured and stored in the computer main memory for post-processing. FF-OCT imaging was performed with the tadpole ventral side up, and the time-lapsed images were taken near the exit of the aorta. In the movie, the pulsatile blood cells are clearly visualized. It is noteworthy that both the cell nuclei and the blood flow can be observed in a single image, where the expansion and contraction of a blood vessel can be seen in accordance with a pulsation of the blood flow. It is estimated from the movie clip that the pulsation occurs at a rate of $2.2\mathrm{Hz}$ . The $30\mathrm{frames}∕\mathrm{s}$ imaging speed and the ultrahigh spatial resolution of our system make it possible to identify most of the individual blood cells during diastole of the heart. During systole, however, the blood cells move beyond the temporal resolution of the present system, so that the images of blood cells appear to be blurred [Fig. 3b] and, in the severe case, averaged out [Fig. 3c].

Fig. 3

(a) to (d): Examples of time-lapse FF-OCT imaging results of an anesthetized tadpole recorded at a video rate. Images are selected at approximately 0.3-s time intervals. A movie clip can be seen in 1Video 1. The image size in the movie clip is half size to reduce the file size.

Video 1

Time-lapse imaging results recorded during a period of 10 s (AVI, 5.7 MB). 1

3.2.

Z-Stack Imaging

In the feasibility study of FF-OCT for ultrahigh-resolution 3-D morphological imaging, a series of FF-OCT images was acquired from the dorsal side to the ventral side of a fixed tadpole with a depth interval of $0.75\mu \mathrm{m}$ . Four FF-OCT images were averaged at each depth position to increase the image contrast, and the measurement of 3-D volume involving FF-OCT imaging at 200 depths took approximately $70\mathrm{s}$ . The lower parts of Figs. 4a and 4b show the representative FF-OCT images at different depths of $300\mu \mathrm{m}$ and $370\mu \mathrm{m}$ , respectively, while the upper parts of (a) and (b) depict the longitudinal cross section reconstructed from the 3-D OCT volume. The cyan line indicates the depth where the FF-OCT image shown in the lower part of the figure was measured. Z-stack imaging results can be seen in 2Video 2 .

Video 2

Z-stack imaging results (AVI, 5.1 MB). 2

Fig. 4

OCT images of a fixed tadpole. The lower parts of (a) and (b) depict the FF-OCT images measured at different depths, while the upper parts show the longitudinal cross sections reconstructed from the z-stack imaging. A movie clip can be seen in 2Video 2. The image size of the movie clip is half size to reduce the file size. For reference, (c) depicts a longitudinal cross section, where the broken line indicates the initial depth from which the z-stack measurement was started.

Features of the internal architectural morphology, such as cell membranes, boundaries of the cell cytoplasm, and cell nuclei, are distinctly visible from the FF-OCT images. It is interesting to note that inside the cell nuclei, a bright spot is always observed. Such bright spots at the individual cells are indicated by arrows in Fig. 4a. These bright spots might correspond to the nucleolus, and the fine fibers connected to the cell membrane might correspond to the cytoskeleton. A similar structure has also been reported in Ref. 38. Identification of the details of the other complicated internal morphology are under investigation. It can be seen from the movie clip that FF-OCT provides enriched information about cellular structure along the horizontal plane.

It is noteworthy that the present z-stack imaging started from a depth approximately $250\mu \mathrm{m}$ below the sample surface. A longer-range yet more time consuming z-scan [Fig. 4c] reveals that the OCT images shown in Figs. 4a and 4b reflect the details of a second layer under the sample surface. In Fig. 4c, the broken line indicates the initial measurement depth of the z-scans in Figs. 4a and 4b. The results in Figs. 3 and 4 show some of the advantages of the present FF-OCT scheme over conventional OCT. FF-OCT is a nonscanning method capable of capturing an ultrahigh-resolution horizontal cross-sectional image that is constantly matched to the focal plane. Video-rate FF-OCT imaging offers the opportunity of continuous (time-lapsed) observation of cellular dynamics at a fixed depth. While speckles are prominent in conventional OCT and adversely affect the image contrast, they are minimal in Figs. 3 and 4, demonstrating another advantage of FF-OCT using a spatially incoherent thermal source.