1. Introduction

Optical coherence tomography (OCT) [1] is a noninvasive technology for defining anatomy in biological tissues. Through measuring singly backscattered light as a function of depth, OCT provides tissue structure with high resolution and sensitivity in vivo. OCT has become a widely used imaging method to image the eye [2,3] and skin [4,5]. Catheter-based OCT probes are developed to extend the OCT application for endoscopic imaging of internal organs [6–8]. OCT probes built in cardiology study can be used to visualize clinically important coronary plaque microstructures [9,10]. To improve the portability of OCT systems, handheld OCT probes have been developed for various applications [11–13]. In a fiber based spectral domain OCT (SD-OCT) system, light from a low coherence source is launched into the source arm of a Michelson interferometer, which consisting of a 2X2 optical beam splitter (BS). The optical coupler splits the source light between reference and sample arms of the interferometer. To optimize the interference signal, polarization state of the reference light should be matched with that of the sample light through using a polarization controller. However, in endoscopic or handheld OCT probe study, fiber perturbations in the sample arm may lead to the polarization state mismatching between sample and reference light in the OCT interferometer, and degrade performances of the imaging system.

Common-path OCT has the reduced sensitivity to vibration and is ideally suited for phase microscopy [14,15]. In a fiber based common-path OCT system developed for endoscopic imaging [16–19], sample and reference light travel through the same optical fiber. So the system is polarization insensitive with the reduced dispersion mismatch between its reference and sample light. However, in a SD-OCT system, signal sensitivity is dependent on depth within an image due to the resolution limit of the OCT spectrometer [20]. This limits the working distance of a common-path SD-OCT probe to several millimeters [17–19] and restrains its application coverage, such as in handheld OCT probe study in which long working distance is needed to achieve large scanning area. Long range OCT is essential for large volumetric field imaging [21]. In time domain OCT study, U. Sharma et al. [16] demonstrate a method to control the working distance of a common-path OCT system through sending the reference and sample light from the probing fiber into a Michelson interferometer to do the interference. But this method will introduce a 50% sample light loss in the Michelson interferometer due to 2X2 beam splitting. In this paper, we present a method to extend the working distance of a common-path SD-OCT system. Sample and reference light from the OCT probing fiber is directed into an optical wavefront splitting interferometer to generate OCT signal. The performance of our common-path SD-OCT system is tested through imaging biological tissues.

2. Methods

2.1 Common-path probing system

The experimental setup employed in this study is schematically illustrated in Fig. 1. It contains a power adjustable superluminescent diode (SLD) source with a center wavelength of 850 nm and a bandwidth of 33 nm (Inphenix Inc.). After passing through an optical isolator, light from the SLD source is guided into port 1 of a 2X2 fiber coupler BS with the beam splitting ratio of 25/75, and split into two portions. One portion of the light (ratio 25%) is delivered to the tissue sample via port 3 of the coupler BS. Fiber end P₁ of port 3 is polished at an angle of α ≈4.2 ± 0.2°, as shown in the dashed window (a) in Fig. 1. A small amount of light E_r is reflected back from fiber end P₁ and serves as a reference light. The measured ratio of back-reflected light to the coupler BS is 0.16% of the incident power at the fiber tip P₁, and the transmitted light efficiency is about 96.3%. The transmitted light is collimated by lens L₁ and travels to the sampling unit, which consisting of a pair of galvanometers for 2-axis beam steering, and a f = 30 mm objective lens L₂ which focusing probe beam onto the sample.

 

Fig. 1 Common-path spectral domain OCT setup. SLD, superluminescent diode; OI, optical isolator; BS, beam splitter. The dashed window (a) shows the fiber end P₁ which is polished at an angle α. The dashed window (b) shows the three dimensional diagram of the interferometer in which the common-path OCT light is split and recombined spatially. V and H represent the vertical and horizontal dimension respectively.

Download Full Size | PDF

Backscattered sample light E_s and reference signal E_r return back to port 2 of the coupler BS via the same physical path with an optical path difference of d₁ between E_r and E_s. Port 2 of the coupler BS is connected with a single mode fiber S_f by a fiber adaptor F_a. The combined sample and reference light output from fiber S_f is collimated by a f = 60 mm collimating lens L₃, and enters an optical wavefront splitting interferometer.

2.2 Interferometer and spectrometer

In Fig. 1, one arm of the interferometer consists of two silver coated D-shaped mirrors (Thorlabs Inc.) M₁ and M₄, which can be used to split optical beam spatially with its edged side. The other arm consists of two silver coated mirrors M₂ and M₃, and a K9 glass plate G_p for dispersion compensation. Mirrors M₂ and M₃ are mounted on a translation stage T_s which is used to adjust the optical path difference between two arms of the interferometer. The dashed window (b) in Fig. 1 shows the three dimensional (3D) diagram of light propagation in the interferometer. The input beam (consisting of E_r and E_s) with a round shape is split vertically into two segments by mirror M₁. The upper half-round beam segment (shown as red) is reflected to the output port of the interferometer by mirror M₄. The lower beam segment (shown as blue) propagating under mirror M₁ is reflected by mirrors M₂ and M₃, and transmits to the output port of the interferometer under mirror M₄. Therefore, input sample light E_s is split as E_s1 and E_s2 by mirror M₁, and reference light E_r is split as E_r1 and E_r2. At the output port of the interferometer, the recombined beam shape is round. But there is a phase difference between the upper and lower segment of the recombined beam which is induced by the optical path difference between two arms of the interferometer.

Light output from the interferometer is reflected by two AR coated dielectric mirrors M₅ and M₆, and detected by a custom spectrometer. The spectrometer consists of a diffraction grating G (1200 lp/mm), an IR achromatic doublet lens L₄ (f = 150 mm), and a CCD line-scan camera. The camera has a 12-bit resolution. Figure 2 shows the 3D diagram of the spectrometer. The upper (red) and lower (blue) segment of the light, which include sample and reference signals, are dispersed by the grating G and focused onto the CCD camera by the lens L₄.

 

Fig. 2 Three dimensional diagram of the spectrometer.

Download Full Size | PDF

As can be seen in Fig. 1, for the split light components E_r1 and E_r2 from different arms of the interferometer, when the optical path difference d₂ between them matches with the path difference d₁, light components E_s1 (sample light) and E_r2 (reference light) will have the same phase delay and interfere with each other to generate OCT signal [16]. For the light component E_r1 (or E_s2), frequency of the interference signal between it and another light component is too high to be resolved by the spectrometer due to the large optical path difference between them. Because E_r2 acts as the OCT reference signal, a glass plate G_p should be inserted in the arm $\overline{M_2{M}_3}$ of the interferometer to compensate the dispersion induced by lenses L₁ and L₂ in the sampling unit.

The spectrum, which contains the encoded depth reflectivity information, is measured by the CCD camera. The exposure time of the camera is set as 60 μs in this study. Data from the camera are transferred via the Cameralink interface to a computer. A VC + + program running on the computer coordinates the frame grabber and galvanometers, and provides data acquisition, processing, and image displaying.

2.3 Signal to noise ratio

In a SD-OCT system, at the shot noise limit, its signal to noise ratio (SNR) can be expressed as [22]:

2.4 System alignment

For the upper and lower beam segment output from the wavefront splitting interferometer in Fig. 1, they should be parallel with each other. This is accomplished through the adjustment of mirrors M₁ and M₄ coordinately. To facilitate system alignment, port 2 of the coupler BS is unplugged from the adaptor F_a firstly, and port 4 with higher light power is connected to the fiber S_f, as shown by the red dashed line in Fig. 1.

Without the insertion of mirrors M₁ and M₄, the optical path from the collimation lens L₃ to the CCD camera is aligned at first. When the spectrometer is aligned well, reflector M₁ is inserted vertically into the beam path to split the collimated light into upper and lower segments evenly in power. Then, mirror M₄ is inserted from upside with its bottom side closing to the lower beam segment reflected by mirror M₃. After that, mirrors M₁ and M₄ are adjusted coordinately to achieve balanced CCD signal strength for two arms of the interferometer. Finally, port 4 of the coupler BS is unplugged, and port 2 is connected with the fiber S_f to do OCT imaging. To evaluate the light loss induced by spatial beam splitting and recombining in the interferometer, light power between mirrors M₄ and M₅ in Fig. 1 is measured with and without the insertion of mirrors M₁ and M₄. The difference is about 2%. Therefore, loss induced by spatial beam splitting and recombining is small.

3. Results and discussions

Figure 3(a) shows the normalized interference fringe detected by the line-scan camera with the balanced sample and reference signal strength (E_s ≈E_r) using a silver mirror as the sample. As discussed above, due to the existence of E_r1 and E_s2 in the OCT interferometer which serve as DC background signal, modulation depth of the interference spectrum shown in Fig. 3(a) is lower than that of the classic SD-OCT system.

 

Fig. 3 (a) Interference spectrum detected by the line-scan camera. (b) Measured OCT point spread function in log scale.

Download Full Size | PDF

At the sampling power of 550 μW, a neutral density (ND) filter with 17.6 dB attenuation coefficient (measured in one path) is inserted between lens L₁ and the galvo-scanner to reduce sample light intensity (E_s << E_r), and the interference signal after removing of the background spectrum is sampled. This interference spectrum is then linearly interpolated in k-space. Fourier transformation of the interpolated result yields the point spread function (PSF) of the OCT signal. In Fig. 3(b), the blue curve represents the axial line (A-line) profile of a mirror reflector in log scale, which is obtained at a depth of approximately 0.16 mm from the zero delay. To determine the noise floor of the PSF curve, the silver mirror was removed and 400 A-lines were sampled. These 400 consecutive A-lines were then averaged and the result was plot in Fig. 3(b) as the red curve. In Fig. 3(b), the maximum signal intensity for the blue curve is 108.2 dB at the depth of 0.16 mm, and the noise level at that depth is 54.3 dB based on the red curve. So the SNR of the OCT signal is 53.9 dB. Considering the attenuation coefficient of the ND filter, the OCT sensitivity will be 89.1 dB. Through analyzing the PSF profile in linear scale, the axial resolution of our OCT system is 9.95 μm in air.

To demonstrate the feasibility of the common-path SD-OCT system investigated in this study, biological samples are imaged. The first sample is onion skin. The residual background signal, obtained through blocking the sample light prior to image acquisition, is subtracted from the interference signal to reduce the background noise. The imaging result is shown in Fig. 4(a), which comprises of 400 A-lines. The image size is 2.8 x 1.8 mm² (horizontal x vertical). The second sample used in this study is chicken trachea, which is imaged in vitro. The trachea was cut open from one side, and the OCT probe beam was directed onto its inner wall. Imaging result of the chicken trachea is shown in Fig. 4(b), in which the image size is 4.2 x 1.8 mm². In the image, structures of hyaline cartilage can be identified clearly. In the above experiments, distance from the fiber end P₁, which serves as the reference plane of the SD-OCT system, to the biological sample is about 140 mm. Thus, long range common-path SD-OCT imaging is achieved successfully.

 

Fig. 4 Images sampled using the OCT system illustrated in Fig. 1. (a) OCT image of onion skin; (b) OCT image of chicken trachea. HC, hyaline cartilage.

Download Full Size | PDF

In this study, the measured system sensitivity is 89.1 dB at the sampling power of 550μW. According to Eq. (1), OCT sensitivity can be improved through increasing the sampling power. But in the common-path probing system, a portion of light reflected by the fiber end P₁ acts as the reference signal, which strength is proportional to the light power onto biological samples. Further increasement of sampling power is limited by the possible CCD signal saturation. On the other hand, light strength reflected back by the end face of an optical fiber is inversely proportional to the angle polished at the fiber tip [18]. Thus, to improve the OCT sensitivity through applying higher sampling power, fiber end P₁ should be polished to a larger angle to sustain the intensity of reference light at a reasonable level.

In the wavefront splitting interferometer in Fig. 1, the input OCT light is split vertically into two parts by mirror M₁. For each beam segment, focused beam spot size at the CCD camera will be increased accordingly in the vertical dimension, because focused beam size is inversely proportional to the beam diameter onto the lens L₄. Variations of the focused beam size at the CCD array are analyzed with ZEMAX for input beams before and after spatial beam splitting. Figure 5 shows the result. For a round input beam in Fig. 5(a), at the focal plane of lens L₄, vertical PSF profile of the focused beam is shown as the solid curve in Fig. 5(c). For the lower half-round beam segment focused by lens L₄, as shown in Fig. 5(b), vertical PSF profile of the focused beam is represented by the dashed curve in Fig. 5(c). It can be seen that width of the dashed curve is larger. The calculated width ratio between the dashed and solid curve is 1.83, which means that the beam spot size at the CCD camera is increased 83% in the vertical dimension after the input beam is split vertically in the interferometer. There is no beam size increase in the lateral dimension which is orthogonal to the direction of beam splitting based on ZEMAX simulation [23]. So the resolution of the spectrometer is not affected by the vertical beam splitting. In Fig. 1, without the insertion of mirrors M₁ and M₄, the calculated beam spot size at the CCD array is 10.41 μm. Thus, for each half-round beam segment, focused beam size at the CCD camera will be 19.05 μm vertically. Pixel size of the CCD camera used in this study is 14 x 14 μm², which can barely cover the enlarged beam spot in the vertical dimension after spatial beam splitting. A line-scan camera with larger pixel height will be helpful to fully cover the enlarged beam spot at the CCD array to improve light collection efficiency of the spectrometer.

 

Fig. 5 (a) A round input beam is focused by lens L₄; (b) Lower half-round beam segment is focused by lens L₄; (c) Solid and dashed curves represent vertical beam profiles at the focal plane of lens L₄ for Fig. 5(a) and 5(b) separately.

Download Full Size | PDF

Common-path OCT is compact and eases of circumvention of the polarization mismatch problem. Long range common-path SD-OCT can be useful for the development of polarization insensitive hand-held OCT probe in ophthalmic and dermatologic studies, or endoscopic OCT probe for large volumetric imaging of internal organs. It may also be helpful in Doppler OCT study or parallel SD-OCT imaging with line-field illumination [24], which are sensitive to phase stability of the OCT system. To extend the working distance of a common-path SD-OCT system, reference and sample light from the probing fiber need to be sent to an optical interferometer to generate OCT signal. Instead of using a Michelson interferometer [25], a non-collinear optical interferometer is employed in this study. As can be seen in Fig. 1, the recombined beam output from the interferometer can propagate to the CCD detector without suffering additional light loss induced by beam splitting. If a Michelson interferometer is used to replace the interferometer in Fig. 1, half of the input sample light will not be able to reach the spectrometer due to beam splitting by a 2x2 coupler.

For the common-path OCT probe with a GRIN lens as the focusing element, such as the OCT probe in Yin’s paper [19], there is no optical element between the reference plane (distal surface of the GRIN lens) and biological samples. So the system has the reduced dispersion mismatch between its reference and sample light. But in the common-path OCT system investigated in this study, there are two lenses L₁ and L₂ employed between the reference fiber tip and biological samples to collimate and focus the probe beam. These two lenses will introduce additional dispersion to the sample light, which need to be compensated to optimize the axial resolution of the imaging system. In this study, a glass plate is used in the reference arm of the interferometer to balance the dispersion between sample and reference light. Because the reference signal E_r2 passes through the glass plate G_p only once, the dispersion of G_p should be one time larger than that of lenses L₁ and L₂. Meanwhile, for the long range OCT system in Fig. 1, its sensitivity will be 3 dB lower than that of the classic SD-OCT system with the same sampling power and CCD integration time. According to Eq. (1), this sensitivity drop can be compensated through increasing the integration time of the CCD canera. Thus, there will be a trade-off between the system sensitivity and imaging speed for the OCT configuration proposed in this study.

4. Summary

In conclusion, we present a method to extend the working distance of a common-path SD-OCT system. The common-path OCT light back from the probing fiber, which consisting of sample and reference signals, is launched into a free space optical interferometer and split spatially into two beam segments by a wavefront splitting mirror. Two parallel beam segments interfere non-collinearly to generate OCT interference signal. Results obtained in this study are promising. Measured OCT sensitivity is 89.1 dB at the sampling power of 550 μW, and biological samples are imaged successfully. Distance from the end of the probing fiber, which serves as the reference plane of the SD-OCT system, to the biological sample is about 140 mm. Method investigated in this work can be used to promote the development of polarization insensitive OCT probe for biomedical imaging applications.