1. Introduction

Optical coherence tomography (OCT) is a depth-resolved, high-resolution, noninvasive imaging technique that employs low-coherence interferometry to record the intensity of light back-scattered from a turbid sample [1]. Polarization-sensitive (PS) OCT is a variant that has been introduced to measure polarization alterations in birefringent tissues [2]; thus it provides an additional contrast to the technique. Equally significant, PS-OCT does not suffer from image degradations caused by tissue birefringence that can be problematic in conventional OCT images [3].

The complexity of polarization measurements and the emerging clinical and scientific need has resulted in a variety of approaches since PS-OCT was proposed. Several optical setups in bulk [2–5] were reported for characterization of polarization properties of tissues. Fiber-based systems, especially preferred for endoscopic imaging, were assembled by utilizing multiple measurements to compensate for random polarization transformations in conventional single mode fibers [6,7]. Multiple measurements could only be abandoned by sophisticated implementations.

With a constant unidirectional stress applied onto the core of a single mode fiber, two orthogonal linear polarization states can be maintained despite external perturbations. This type of fiber is known as polarization-maintaining fiber (PMF). The stress alters the dielectric constant along the axis of the fiber core; as a result, a lag occurs between two linearly polarized states propagating in the orthogonal channels (the fast and slow). PMF-based low-coherence interferometers [8–11] were reported for various applications. A time-domain PS-OCT based on PMF was demonstrated for reflectivity and birefringence imaging of tissue with a single measurement and a single detector [12]. The system retained the flexibility of fiber based systems with the direct and simple analysis of bulk systems, and the use of two detectors allowed phase-sensitive imaging of relative axis orientation. Time-domain systems, however, are slow for OCT imaging of tissues. The imaging speed can be increased by performing faster axial scans in the reference arm, but with a cost of degrading the signal to noise ratio [13,14].

Fourier-domain OCT does not require scanning of the reference arm; instead, a spectrometer in the detection arm (spectral-domain, SD) [15–17] or a wavelength sweeping laser source (swept-source, SS) [18–20] is utilized to record the optical spectrum. The spatial information along the axial direction is obtained by a Fourier transform of the interference pattern on the wavenumber (k-space) spectrum. Fourier-domain systems provide increased acquisition speeds for real-time and clinical applications with improved sensitivity [21–23]. Some of the theoretical and practical considerations on the performance of these techniques are also discussed [24].

Several Fourier-domain PS-OCT systems have been already introduced in bulk and based on single-mode non-PM fibers [25–31], retaining similar disadvantages in the corresponding time-domain implementations. Recently, PMF-based SDPS-OCT systems have been reported using single [32] and dual [33] camera setups. It is known that SD-OCT, however, compared to SS-OCT, suffers more from motion artifacts [34]. Due to the simplicity of its detection arm, SS-OCT has the potential to eliminate excess source noise [35,36], remove fixed parasitic lines due to laser source and autocorrelation in the interferometer [37], and achieve power-efficient interferometers [38].

Herein, we introduce a swept-source polarization-sensitive optical coherence tomography (SSPS-OCT) with a PMF-based interferometer. The system operates at 1.3 µm regime, suitable for imaging turbid tissues, and benefits the advantages of balanced-detection. A single sweep of the optical spectrum is sufficient to produce reflectivity and birefringence information along a depth profile (A-line) with straightforward algorithms. The circular polarization state illuminating the sample yields retardance measurement insensitive to sample orientation in the lateral plane. The SSPS-OCT produces relative optical axis orientation images as well. Here, we describe the system and compare its dynamic range and sensitivity with those of a non-PM fiber based SS-OCT. Retardance and relative axis orientation measurements of the SSPS-OCT are also characterized. The results demonstrate sensitive, accurate, and ghost-line-free performance. The imaging capabilities of the SSPS-OCT are presented by using biological tissues ex-vivo.

2. System description

Figure 1 shows a schematic of the PMF-based SSPS-OCT with a polygon mirror based wavelength sweeping source. Half of the output power of the internally-isolated booster optical amplifier is tapped out of the ring cavity using a 50:50 fiber coupler. The other half of the power is transmitted to a polygon scanning monochromator/filter through a circulator. This setup was described earlier in detail [39]. Our monochromator consists of a collimator, a 600 lines/mm grating, a confocal telescope (75 and 35 mm achromatic lenses), a rotating 72-facets polygon mirror (diameter of about 6.1 cm), and an end mirror. Determined by the rotational speed and the number of facets of the polygon mirror, the monochromator scans the optical spectrum at a sweeping rate of 28 kHz. Light directed to the amplifier through the circulator is amplified. Half of the amplifier’s output exits the ring cavity. Five percent of this is directed to a trigger generator for synchronizing data acquisition, and the rest measuring an average power of 45 mW is used in the interferometer.

 

Fig. 1 SSPS-OCT system. BD: balanced detectors, BOA: booster optical amplifier, Circ: circulator, G: grating, L: lens, T: achromatic telescope, M: mirror. P: polarizer, ND: neutral density filter, QWP: quarter-wave plate, WP: Wollaston prism. Dashed boxes show various stages as labeled. Bold lines represent the PMF. Zero degree splices are indicated in the reference and sample arms.

Download Full Size | PDF

The output of the swept source is fed into a circulator that prevents returning light from reentering the laser cavity. A polarization controller maximizes the polarization state transmitted through a polarization beamsplitter cube, which is positioned in a three-port fiber bench. The output port is aligned so that the polarized light couples into the slow axis of the PMF only.

A PM coupler transmits equal amounts of light into two arms. The sample arm is equipped with an achromatic QWP aligned at 45° with the input polarization state. As a result, circularly polarized light transmitted onto a sample yields retardance measurements independent from the optical axis orientation in the plane normal to ranging [40]. Lateral scanning of the sample is accomplished by a galvanometer scanner. Because birefringence of tissue alters the polarization state, the orthogonal polarization components of back-scattered light couple back into the fast and slow channels of the PMF. The reference arm consists of a polarizer with axis aligned at 45° with respect to the slow axis of the PMF, a neutral density attenuator, a 45 mm achromatic lens, and a stationary mirror. The polarizer ensures that the reflected light couples equally in the slow and fast axes of the PMF.

Due to its birefringence, light propagates at different speeds in the fast and slow channels of the PMF resulting in a lag between the polarization components. This lag is larger than the coherence length of the source after a few centimeters of propagation. Therefore, two decorrelated components exist for light returning to the PM coupler from the reference arm. They interfere with the corresponding components in the sample arm. The interference patterns on the orthogonal channels of the PMF make it possible to describe any polarization state of light returning from the sample.

In the detection arm, a Wollaston prism splits the interference patterns on these PMF channels and diverts the beams onto the corresponding optoelectric elements of the balanced photodetectors. Balanced detection is achieved for both channels by collecting the interference patterns returning to the input arm - the slow channel through the circulator and the fast channel through a fiber aligned to couple the reflected light of the polarization beamsplitter cube. The interference signals on the input and detection arms are 180° out of phase; therefore, balanced-detection doubles the signal and minimizes the common excess noise.

After amplification the electrical signals are digitized by a 12-bit analog to digital converter at a rate of 50 Msamples/sec. The depth profile of the sample can be calculated for each individual polarization channel by using these methods: doubling the number of points using a Fourier-transform-based interpolation [41] to map the interference patterns from wavelength domain to wavenumber domain (k-space) [17,41], apodization [42], numerical compensation for dispersion imbalance [43], reference arm’s background subtraction [21,37], and calculation of Fourier transform. As a result, spectral modulations in k-space are converted to complex spatial information of the sample along the axial direction.

For the two polarization channels the magnitudes (A₁ and A₂) and phases (ϕ ₁ and ϕ ₂) of the complex profiles are calculated. Reflectivity (R), phase retardance (δ) and relative optical axis orientation (θ) as a function of depth (z) are calculated as described earlier [5,12]: $R(z)\propto A_1{(z)}^2+A_2{(z)}^2$, $\delta (z)=\arctan (A_1(z)/A_2(z))$, and $\theta (z)=({\phi }_1(z)-{\phi }_2(z))/2$. Note that, in order to calculate these PS-OCT contrasts, the coherence functions on the two polarization channels need to be aligned as described in the Discussion section.

3. System characterization

3.1 Optical spectrum, coherence functions, and sensitivity

The time-varying output of the swept source is recorded by a photodetector. Figure 2(a) shows four consecutive cycles of the record. Using the monochromator characteristics calculated from the grating equation and the responsivity curve of the detector, a single cycle is rescaled to describe the optical spectrum [Fig. 2(b)]. A center wavelength (λ₀) of 1291 nm and a FWHM bandwidth (Δλ) of 78 nm suggest a FWHM axial resolution (≈0.44λ₀ ²/Δλ for a Gaussian source [13], ) of 9.4 µm.

 

Fig. 2 (a) Output of the swept source, (b) optical spectrum calculated from the monochromator characteristics

Download Full Size | PDF

Figure 3(a) shows the coherence function of the PMF based SSPS-OCT system in logarithmic scale at a depth of 500 µm, which is the optical path length difference between sample and reference arms. The vertical axis is the logarithmic reflectivity (10 × log(R)). With a neutral density filter (OD = 2.1) placed in the sample arm, the reflectivity profile (shown in black) is calculated after mapping the spectrum into k-space and compensating the dispersion in software. Figure 3(a) also shows the suppression of the broad side lobes by apodization using a hamming window, and rejection of the source line at 0.1 mm by reference subtraction.

 

Fig. 3 Coherence functions of (a) PMF-based SSPS-OCT and (b) a non-PM fiber-based SS-OCT. Black trace shows the coherence function after dispersion compensation. The red and blue traces are obtained after apodization, and reference subtraction in addition to apodization, respectively.

Download Full Size | PDF

To quantify the dynamic range and sensitivity of the reflectivity measurement, power returning from the reference and sample arms was measured at the fast channel of the detection arm as 12 µW and 220 nW, respectively. To avoid saturating the photodetectors, the source power was attenuated by loosening the FC/APC connector between the swept-source and the circulator. Using the peak reflectivity and the average noise level between 1 and 1.5 mm, a dynamic range of 64.2 dB was measured for the blue trace (with apodization and reference subtraction). The summation of the dynamic range and the sample arm attenuation (2 × 10 × OD, in double pass) yields a sensitivity of 106.2 dB, which is similar to or exceeding the numbers reported for SS-OCT systems [37,39,44]. For a shot noise limited system, the model developed earlier for non-PMF SS-OCT [37] yields a theoretical sensitivity of 120 dB with about 4.5 mW illuminating the sample and at 28 kHz A-line rate. The interferometer was optimized by reference arm attenuation [45] and balanced-detection; the latter has resulted in 6-7 dB improvement on the dynamic range of the SSPS-OCT. Without apodization, axial resolution is measured as 12.2 µm in air. The deviation from the theoretical value can be due to errors in mapping the optical spectrum, its non-Gaussian spectral distribution, and limited accuracy in the numerical processing. Although apodization reduces resolution to 15 µm, the contrast advantage is worthy.

The reflectivity measurement of our PMF-based SSPS-OCT is compared with that of a non-PM fiber-based SS-OCT. The PM coupler was replaced with a conventional (non-PM) 50:50 coupler. The input arm of the coupler was directly connected to the circulator output, and the fiber of the detection arm was connected to the balance detector. The polarizer and the quarter wave plate were removed from the reference and sample arms. The system parameters and the sample arm attenuation were in good agreement with the PMF setup. Degraded sensitivity was observed when polarization states in the sample and reference arms were not matched; therefore, fiber polarization controllers were added to these arms to maximize the fringe visibility. Figure 3(b) shows the reflectivity profile that is calculated as in the PMF case. The measured sensitivity is 112 dB. The discrepancy between the measured and theoretical sensitivity value (120 dB) can be attributed to residual thermal and excess noise [37]. The additional 5.8 dB sensitivity loss in the PM system is attributed to leakage between the polarization channels of the optical components. As shown in Fig. 3, the performances of the PM and non-PM systems for the reflectivity measurement are comparable.

The finite instantaneous line width of the swept source results in decay in the dynamic range, hence sensitivity, as a function of depth. The depth-dependent decay is characterized for both PM and non-PM fiber systems. The coherence functions of the two SS-OCT systems at different sample depths are shown in Fig. 4 . The performance of the PMF system shown in Fig. 4(a) is similar to that of the non-PM fiber system in Fig. 4(b). A decrease of about 11 dB was measured at a depth of 2.75 mm for both cases. The axial resolution measured at this location is degraded to 18 µm in air, which is attributed to numerical errors. With the current system parameters, imaging depth range [16] is 6 mm.

 

Fig. 4 Depth-dependent decay in (a) PMF-based SSPS-OCT, and (b) non-PM fiber-based SS-OCT. Differences between the plots are barely distinguishable.

Download Full Size | PDF

3.2 Birefringence measurements

To characterize the birefringence measurements of the SSPS-OCT system, a voltage-controlled variable retarder was used as a birefringent sample. It is placed in the sample arm between the quarter wave plate and the lens. The neutral density filter was also placed in the sample arm to optimize the system as reported in section 3.1. Voltage applied to the retarder was varied from 0 to 10 V with a step size of 10 mV. Figure 5 shows the retardance curve obtained from the SSPS-OCT measurement of phase retardance. The shape of the curve is consistent with the device specifications (shown in references [12] and [32]). The measured retardance, however, is slightly different due to the large discrepancy between our and manufacturer’s (849 nm) test wavelengths.

 

Fig. 5 Retardance measurement of a voltage-controlled variable retarder

Download Full Size | PDF

Because light incident on the sample is circularly polarized, the retardance measurement is insensitive to the optical axis orientation of the sample. This is illustrated by fixing the retardance of the variable retarder and rotating it over a range of 180° in rotational steps of 5°. Figure 6(a) shows the measured phase retardance (δ) in hollow blue circles. Indicating a repeatable measurement, δ measures 43.2° with 0.8° standard deviation. Figure 6(a) also shows the fast axis orientation (θ) in solid green circles. The measurement deviates from a unity slope by 0.02, and has a standard deviation of 2.5° around the best fit line.

 

Fig. 6 Measurements of retardance and axis orientation as a function of (a) rotation, and (b) depth

Download Full Size | PDF

To investigate the birefringence measurement at different depths, both retardance and orientation of the variable retarder were fixed and the optical path length difference between the reference and sample arms was varied. Figure 6(b) shows the consistency of the measurement over a depth of 2.75 mm. Standard deviations of 0.7° and 2.0° were calculated for phase retardance and axis orientation measurements, respectively. Small errors in Fig. 6 are attributed to misalignments and imperfections of the polarization components. The axis orientation values shown in Fig. 6 are absolute, because they are compensated by removing the phase offset described in Discussion.

4. Biological imaging

The capability of PMF-based SSPS-OCT to image biological tissues is demonstrated by imaging various chicken tissues. The sample objective is a telecentric scan lens. Illuminated by a collimated beam with 2.4 mm diameter, it provides a lateral 1/e ² spot size of 42 µm at focus. Figure 7 shows the reflectivity and retardance images of a leg muscle and a tendon. The dynamic range of the reflectivity images is 40 dB. The retardance images show banding patterns with a frequency proportional to tissue birefringence. The tendon was covered with a non-birefringent connective tissue that is hardly distinguishable in the reflectivity image, but these two layers are clearly visible in the retardance image. The frequency of the banding patterns shows that the tendon tissue exhibits more birefringence than the muscle tissue. Phase retardance at the tissue surface is zero (black), which changes as light propagates deeper in the birefringent tissue. The measurement returns back to zero after 180° of single-pass phase retardance. Using this, phase retardances of about 320 °/mm and 1280 °/mm are obtained for the leg muscle and tendon tissues shown in Fig. 7, respectively.

 

Fig. 7 Reflectivity (a,c) and phase retardance (b,d) images of a chicken leg muscle (a,b) and tendon tissue (c,d).

Download Full Size | PDF

To demonstrate imaging of the relative axis orientation, a rectangular slab of a muscle tissue harvested from chicken breast was diagonally cut into two pieces. The pieces were put together so that the muscle fibers had an angular separation. Figure 8 shows its SSPS-OCT images. The reflectivity and retardance images were not influenced from the divergence of tissue fibers. The relative optical axis orientation image, on the other hand, displays this variation as an additional contrast. The measured phase retardance for the imaged breast muscle tissue is 450 °/mm.

 

Fig. 8 Reflectivity (a), phase retardance (b), and relative optical axis orientation (c) images of two angled slabs of chicken breast muscle

Download Full Size | PDF

5. Discussion

PMF has high isolation between its orthogonal channels; as a result, the reflectivity profiles in Figs. 3 and 4 demonstrated a performance comparable to those of non-PM fiber systems. However, leakage between the polarization channels interferes with the sample or reference light. The cross-coupling in the PM coupler and the PM fiber splices, for example, can generate considerable ghost images. The separation between the ghost and main images is dictated by the length of the PMF segment after the leaking point. Therefore, the problem can be treated by adding long PMF segments (or other birefringent materials) to displace the ghost images out of the imaging range [12,32]. In SSPS-OCT, we added 30 meters long PMF segments to each of the sample and reference arms to displace the ghost lines. For a beat length B, and PMF length l, the displacement is given by l × λ ₀/2B, where 2 accounts for the double pass in free space of the sample arm. With B = 2.5 to 4 mm for the Corning’s PM 1300 fiber, the ghost line due to leakage in the PM coupler should be displaced by an amount between 4.9 to 7.8 mm away from the main line; we measured a displacement of about 6 mm. Leakage in either of the detection or input portions of the interferometer gives rise to fixed lines. These lines are minimized by fine alignments of the Wollaston prism and the polarization beamsplitter cube with respect to the PMF axes. Residuals, if any, can be eliminated by reference arm subtraction.

Birefringence of the PMF renders a lag between the polarization components propagating in the slow and fast channels. If this lag is matched in the two arms, the coherence functions of the polarization channels will coincide. Otherwise, an axial shift between the coherence functions of the two channels is observed. If not taken into account, this can degrade the resolution and generate errors. We positioned the coherence functions precisely by splicing fiber lengths matching the lags introduced in the sample and reference arms. A length mismatch of one beat length (about 3 mm) leads to a λ ₀ shift between the two coherence functions; an order of magnitude smaller than the axial resolution. Phases of the coherence functions, however, cannot be matched using this method. Hence, optical axis orientation images are relative due to a phase offset, which can be measured using a retarder with known orientation. Absolute orientation images can be obtained by removing the offset, provided that external disturbances do not change it. The applicability of this method was shown in Fig. 6. In case of unmatched fiber lengths, a numerical method can be deployed as demonstrated for Fourier-domain OCT [33].

Environmental disturbances such as fiber movement and temperature effects are minimized by placing the PMF interferometer in an enclosure. This excludes about 1 m of PMF in each arm that connects to collimators. A shift between the coherence functions due to fiber movement or time (day to day operations) was not observable, which suggests PMF-based probes for endoscopic studies in future. Relative axis orientation image can be influenced if the fiber is disturbed mechanically during image acquisition. At high imaging rates, the disturbances are usually too slow to affect a single frame.

6. Conclusion

We described and characterized a SSPS-OCT system based on PM fiber technology, and demonstrated its performance on imaging biological tissues. A single sweep of the optical spectrum is sufficient to yield the reflectivity and birefringence information along a depth-profile (A-line). The PMF-based PS-OCT combines the flexibility of fiber-based systems, the immunity to external perturbations and the computational simplicity of bulk PS-OCT systems. In this paper, we have established that PMF-based OCT can be constructed with sensitivity and dynamic range figures comparable to bulk or conventional (non-PM) fiber-based systems. For these reasons, in addition to the fact that PS-OCT accurately represents the reflectivity images of birefringent tissues, PMF-based PS-OCT has potential for considerable impact in scientific studies and clinical applications.