Optical coherence tomography (OCT) is a non-invasive imaging method that has found widespread applications in biomedicine [1]. Providing real-time volumetric imaging with micrometer-scale resolution, OCT has become an indispensable diagnostic tool in ophthalmology and more recently also in catheter-based applications such as intravascular imaging. As OCT is based on interferometry of broadband light, most modern OCT layouts use light sources such as superluminescent diodes (SLDs) [2] or short cavity lasers [3] which emit highly polarized light. OCT is based on signals originating from interference between light backscattered from the sample and a reference light beam. Thereby, only the co-polarized portion of the light returning from the sample contributes to the interference signal representing the structure. Here co-polarized describes the polarization component matching the polarization state of the reference beam. Conversely, the cross-polarized portion of the sample signal—i.e., light polarized orthogonal with respect to the reference beam, which can be created for instance by sample birefringence—cannot be detected and hence does not contribute to the OCT signal. In OCT imagery, artifacts ranging from a signal reduction to a complete absence of the structural signal are observed at locations associated with—at least partly—cross-polarized light. OCT images of birefringent tissues such as tendons, muscles, cartilage, or nerves thus exhibit a seemingly striped appearance since the polarization meanders between co- and cross-polarized states as the beam propagates into the sample (see for instance [4,5]). Polarization artifacts like the aforementioned stripe pattern cannot be avoided when a polarized light source is used and therefore may be misconceived as actual morphology of the imaged material. Additional polarization artifacts leading to a degraded image resolution, signal reduction, or even annihilation are often observed when the polarization state of the light source is changed within the sample arm optics and wave guides in many fiber-based OCT schemes and catheter-based OCT systems [6,7]. One rather costly and complex mitigation approach has been the implementation of polarization-diverse detection schemes employing one detection channel for each co- and cross-polarized signals [8]. The objective of the research presented in this Letter is to provide a simple approach providing simultaneous OCT imaging of co- and cross-polarized light. By introducing a depolarizing element into the source arm of the OCT interferometer, our novel approach renders virtually any OCT layout sensitive to backscattered signals in any polarization state and therefore effectively removes frequent polarization artifacts in OCT data.

The effect of the input polarization state on the polarization trajectory along an A-scan for OCT imaging of a birefringent sample is illustrated in Fig. 1. In the simulation in Fig. 1(a), a linearly polarized input state ($-$S1) is launched and orbits around the Poincaré sphere [10] while passing through a birefringent specimen oriented at 45°. When the state reaches +S1, it is cross-polarized with respect to the incident state—it cannot interfere with the reference state and thus does not produce a detectable OCT signal. In contrast to the polarized input state (degree of polarization DOP = 1) in Fig. 1(a), unpolarized light consists of two or more simultaneous polarization states and typically has a low DOP. Completely unpolarized light, as symbolized in Fig. 1(b), does not have a single polarization state but rather covers a continuum of polarization states covering the entire Poincaré sphere (DOP = 0; note that all indicated Stokes vectors were scaled to unity length for better visibility). The hardware depolarizer used in this work (Thorlabs DPP25-B [9]) was a liquid crystal polymer based patterned microretarder array which transforms a polarized input light beam into a continuum of polarization states covering an 8-shaped trace on the Poincaré sphere [represented by the white line in Fig. 1(c)]. When propagating through birefringent material, each state on this 8-shaped trace follows a particular circular trajectory. Consequently, some elements of the partly depolarized beam will always contain a co-polarized component, and the signal will never become completely cross-polarized. This is evident in Fig. 1(d) which shows the S1 component of the trajectories of the polarized and depolarized beams from Figs. 1(a) and 1(c), respectively. Figure 1(e) shows the retardation and axis orientation pattern of the depolarizing element as well as the beam profile used in our experimental setup. These data served as the input for the simulations in Figs. 1(c) and 1(d).

 

Fig. 1. OCT imaging with polarized and depolarized light. (a) Trace of a linearly polarized input state (red arrow) as light propagates through a birefringent sample oriented at 45°. After circling around half the Poincaré sphere, the initially vertically polarized light has transformed into a horizontal state, i.e., it is cross-polarized. (b) Completely depolarized light can cover polarization states all around the Poincaré sphere. (c) Partly depolarized light produced out of the linear state in panel (a) by means of a hardware depolarizer comprises a continuum of states represented by the 8-shaped trace (white). Every bundle of the light beam fed into the interferometer has a distinct state on this trace which takes a specific circular trace (shown in blue) as the beam travels into the sample. (d) Simulations of the amplitude evolution along the S1 axis is shown for a single vertical input state (red) and several input states emerging from the partly depolarized input light (blue). (e) Effective phase retardation for 840-nm wavelength light and the axis orientation of the depolarizer are shown in the left and center panel (derived from [9]). A Gaussian beam profile is shown in the right panel.

Download Full Size | PDF

A simple spectral domain OCT setup based on a free-space Michelson interferometer layout was used to prove the principle and demonstrate its performance (Fig. 2). An SLD centered at 840 nm with a full width at half maximum of 50 nm was used as a light source. The light source emitted highly polarized light with a power of 3.1 mW after passing an isolator. The liquid crystal polymer depolarizer was mounted in the source arm of the interferometer with a 50/50 splitter. The sample arm featured a pair of galvanometer scanners and an achromatic lens ($f = 30$ mm). The reference arm included a variable neutral density filter and, for some experiments, was equipped with an achromatic quarter-wave plate (QWP, Thorlabs). The spectral interferograms were acquired by a homemade spectrometer including an 1800-lines/mm transmission grating (Wasatch Photonics), a 100-mm f-theta lens (Cloudray), and a line scan camera (Vieworks) with 4k-pixels read out at 50 kHz. The system provided a depth range of 5.72 mm (in air), an axial resolution of 7.6 $\mathrm{\mu}$m (in air), a sensitivity of 95 dB, and a signal-to-noise ratio (SNR) roll-off of 6.5 dB over 3 mm.

 

Fig. 2. OCT interferometer with spectral domain detection and depolarizer in the source arm. Light source, LS; isolator, ISO; collimator, C; depolarizer, DP; beam splitter, BS; neutral density filter, ND; position for quarter-wave retarder, R; reference mirror, RM; galvanometer scanners, GS; lens, L; sample, S; spectrometer, SP.

Download Full Size | PDF

We tested the dependence of our setup without and with a depolarizer in the input path on different retarder orientations. A mirror served as the sample and an achromatic QWP was inserted into the reference arm to produce elliptical reference states from co-polarized to cross-polarized and back. Interference signals were recorded in 15° intervals for QWP orientations from 0° (co-polarized) to 180°. Each set of measurements was performed without and with the depolarizer in the source arm. As expected, Fig. 3 shows a strong dependence of the interference signal on the orientation of the QWP. Conversely, the interference signal intensity stays almost constant for all QWP orientations when the depolarizer is employed. Note, however, that the overall signal intensity is halved for partly depolarized input light.

 

Fig. 3. Influence of retarder orientation on OCT imaging with polarized and partly depolarized light. The signal intensity was measured for the same sample and reference arm intensity while the orientation of a QWP in one arm of the Michelson interferometer arm was rotated.

Download Full Size | PDF

In Fig. 4, we compare the imaging performance in birefringent tissue. We imaged a formalin-fixed, ex vivo dura mater specimen from the neurobiobank of the Medical University of Vienna (IRB approval number: 396–2011). Dura mater is a highly collageneous tissue [11] and exhibits strong birefringence. Figure 4(a) shows a B-scan image of the specimen in the standard OCT configuration, i.e., without a depolarizer. The typical stripe pattern observed in OCT images of birefringent tissues can be observed. When using the depolarizing element in the source path, these artifacts vanish and the morphology of the tissue can be interpreted more easily [Fig. 4(b)]. Next, we inserted the QWP into the reference arm again that would enable us to dial in co- and cross-polarized reference states. For the standard OCT configuration, the striped artifact and its complementary artifact can be observed for QWP orientations at 0° and 45°, respectively [Figs. 5(a) and 5(b)]. In contrast, the OCT images look virtually identical for co- and cross-polarized reference light when the depolarizer is inserted into the source arm of the interferometer [Figs. 5(d) and 5(e)]. By incoherently adding the intensity of the B-scans acquired with polarized input light at QWP orientations of 0° and 45°, an image similar to the images obtained with depolarized light can be generated. Such composite images are typically referred to as reflectivity images in polarization-sensitive (PS) OCT [12]. Note that the co-polarized, cross-polarized, and reflectivity images obtained with the depolarizer in place are virtually identical.

 

Fig. 4. Conventional OCT imaging of a dura mater specimen with polarized light and polarization-insensitive OCT imaging using partially depolarized light. (a) Conventional OCT B-scan acquired with polarized light shows the typical striped pattern in birefringent tissue. Hypointense polarization artifacts at locations with predominantly cross-polarized signals are indicated by yellow arrows. (b) Using the depolarizer in the source arm of the interferometer, these artifacts are diminished and only hypointense signals owing to the tissue morphology remain (green arrows). The scale bars in panel (b) apply for both panels.

Download Full Size | PDF

 

Fig. 5. Co- and cross-polarized signals for conventional OCT with polarized light and polarization-insensitive OCT imaging. A QWP was used in the reference arm to rotate the detected polarization state. (a) Co-polarized image with stripe patterns similar to Fig. 4(a). (b) Cross-polarized image with complementary stripe pattern. (c) Reflectivity image devoid of polarization artifacts. (d)–(f) Polarization-insensitive OCT images acquired with a depolarizer in the interferometer source arm. The OCT images at QWP orientations of (d) 0° and (e) 45° as well as (f) their incoherent sum show virtually identical image structures without polarization artifacts. The scale bars in panel (f) apply for all panels.

Download Full Size | PDF

To explore the performance of our approach in fiberized OCT setups, we implemented an OCT interferometer based on a wideband 50/50 single mode fiber coupler (see Section 1 and Supplementary Fig. S1 in the Supplement 1). Analogous to the measurements shown in Fig. 3, we investigated the dependence of the interference signal intensity for both polarized and partly depolarized input light. The results are shown in Supplementary Fig. S2, Supplement 1 and agreed well with the performance of the free-space layout, as did the data from an imaging experiment in the dura mater sample (see Fig. S3, Supplement 1).

Our results showcase the impact of the polarization of the input light on the obtained OCT images. We demonstrate that polarization artifacts such as the typical stripe pattern observed in birefringent tissue can be avoided by using unpolarized light. While we present an efficient and cost-effective approach to convert a polarized light beam into partly depolarized light, it is important to recall that inherently unpolarized light sources have already been used for OCT imaging. For instance, white light sources based on thermal light, such as halogen lamps used for full-field OCT [13], emit unpolarized light albeit with low spatial coherence. Also light-emitting diodes and some types of superluminescent diodes emit unpolarized light [14]. Depending on their implementation, Fourier domain mode-locked (FDML) lasers provide light with a low DOP [15]. Nonwithstanding these (at least partly) unpolarized examples, a large number of light sources used for state-of-the-art OCT imaging emit polarized light and may—depending on the use case—benefit from a depolarization stage before hitting the interferometer.

In this work, we used a patterned microretarder array to depolarize the input beam. Depolarizers have previously found applications in OCT to condition the input polarization in depth-encoded PS-OCT layouts [16,17]. There, the depolarizer was used to generate an unpolarized beam that, upon being split at an polarizing beam splitter, provided equal co- and cross-polarized reference signal intensities. In the approach presented in this Letter, the depolarized beam is used for both sample and reference arm, and the resulting reflectivity signal is acquired without splitting between orthogonal polarization states, thereby providing a polarization-insensitive OCT signal in a single shot and maintaining the full depth range.

The patterned microretarder used in this work affects different parts of the input beam cross section with a distinct retardance and optic axis [see Fig. 1(e)]. As the observed retardance depends on wavelength, chromatic retardation effects are expected. While this effect will be minor for a rather narrowband spectrum as used in our setup, where $\Delta \lambda /\lambda = 80/840 \sim 0.1$, any drift of the polarization state across the wavelength spectrum may result in a reduced axial resolution in some imaging configurations in birefringent tissues, especially when extremely broadband sources are used. Likewise, the resolution may be affected by dispersion-like effects owing to the input beam containing a continuum of differently retarded beam components. At the same time, it is important to note that also in many standard OCT layouts—particularly when light is delivered via optical fibers—the polarization state launched into the system and to the sample is unknown. Even when a linearly polarized source is used, fiber optics will chromatically influence the polarization of the guided light beams and may lead to a degraded point spread function and reduced resolution [6].

The approach presented here enables the visualization of tissue morphology free from polarization artifacts such as signal voids in locations where birefringent tissue renders co- to cross-polarized light. However, since there will always be a component which is cross-polarized with respect to the reference light bundle, the fringe visibility will halve and so will the signal amplitude (see Fig. 3). The average intensity in birefringent tissue will be similar to the average intensity measured with a co-polarized detection in standard OCT. However, instead of meandering between 100$\%$ and 0$\%$ for cross-polarized backscatter light, the signal amplitude in birefringent tissue will be at 50$\%$ of the reflectivity signal everywhere when an unpolarized illumination is in use. In other words, for the configuration presented here, the obviation of the polarization artifacts comes at the cost of a 3-dB SNR hit. Also, we expect that the design of a more sophisticated microretarder pattern tailored to the respective illumination beam specifications (e.g., a retarder pattern with circular symmetry or a retardance profile customized for the used wavelength range) may improve the performance of polarization-insensitive OCT imaging.

Polarization-insensitive OCT images, typically called reflectivity images, can also be obtained by PS-OCT. PS-OCT can provide a wealth of information on a sample’s polarization characteristics including birefringence (local retardation), birefringent axis orientation, diattenuation, and depolarization [18]. Still, PS-OCT typically requires more sophisticated optical layouts, increased data throughput (or, alternatively, reduced ranging depth), additional data processing, and thus comes at higher cost and complexity. The polarization-insensitive approach presented here is somewhat analogous to the role of speckle in OCT imaging, which may provide important information about the microstructural sample composition but whose grainy pattern is sometimes removed to achieve a clearer morphological visualization [19]. Our approach does not provide any of the additional contrast PS-OCT can deliver, it does however provide polarization-insensitive OCT images akin PS-OCT reflectivity images. Its novelty and appeal is that only an inexpensive modification of the standard OCT layout is required to obtain this reflectivity contrast and avoid polarization artifacts. We believe that OCT imaging with depolarized light—or, in general, unpolarized light—may be particularly useful for catheter-based OCT imaging of tubular organs. Fiber motion in catheters and endoscopic systems alters the polarization state and thus may reduce the signal when interfering sample and reference signals. Moreover, vascular and gastrointestinal tissues exhibit birefringence and thus are prone to produce polarization-induced image artifacts.

In conclusion, we have demonstrated polarization-insensitive OCT imaging. Using a simple and inexpensive modification by introducing a depolarizer in the source arm of a standard OCT layout, we achieved OCT imaging free from polarization artifacts. Our image data demonstrate the performance of our approach in birefringent biological tissue and indicate its potential for imaging applications in birefringent tissue and/or using flexible probe designs otherwise vulnerable to polarization artifacts.