9 October 2012 Mapping local optical axis in birefringent samples using polarization-sensitive optical coherence tomography
Author Affiliations +
J. of Biomedical Optics, 17(11), 110501 (2012). doi:10.1117/1.JBO.17.11.110501
An algorithm was developed to obtain depth-resolved local optical axis in birefringent samples by using conventional polarization-sensitive optical coherence tomography (PSOCT) that uses a single circularly polarized incident light. The round-trip sample Jones matrices were first constructed from the cumulative PSOCT results. An iterative method was then applied to construct the depth-resolved local Jones matrix from which the local optical axis was calculated. The proposed algorithm was validated in samples with homogeneous axis and with depth-varying optical axis. Imaging examples were shown to demonstrate the capability of this method for extracting correct local axis and revealing features not evident in other images.
Fan and Yao: Mapping local optical axis in birefringent samples using polarization-sensitive optical coherence tomography



Polarization-sensitive optical coherence tomography (PSOCT)1 enhances conventional optical coherence tomography (OCT)2 by revealing polarization properties, such as retardance and optical axis, in addition to structural information. Such polarization-dependent contrast has potential applications in a variety of areas, such as ophthalmology, dermatology, and dentistry. Due to the “round-trip” signal detection in OCT, conventional PSOCT measurements are integrated results from the sample surface to the measurement depth. Because such cumulative results are not intuitive for interpretation and incorrect for samples with depth-varying optical axis, various algorithms34. have been developed to extract depth-resolved local polarization properties. The majority of these reported methods are Jones matrix-based56.7 OCT systems that require multiple imaging acquisitions with different incident polarization. These algorithms cannot be directly applied to the widely adopted conventional PSOCT systems that use a single circularly polarized incident light. An algorithm8 was recently reported to extract local retardance in birefringent samples using a conventional PSOCT system. However, the local optical axis cannot be obtained by using this algorithm. Here, we extended the previous algorithm8 and developed a new iterative method that can derive a true depth-resolved local optical axis along with local retardance in conventional PSOCT systems.



In a conventional PSOCT system, a circularly polarized light is usually used as the incident light, and the horizontally (H) and vertically (V) polarized backscattered light from the sample are measured. The “single-trip” cumulative retardance δ and “fast” optical axis θ can then be obtained in addition to the intensity I=10log10(H2+V2) as:8


The δ and θ in Eq. (1) are cumulative results that are affected by sample properties along the entire imaging leading to the measurement depth. In birefringent samples with a depth-independent optical axis, the obtained optical axis θ represents the true sample optical axis except for a phase wrapping,10 and the cumulative retardance δ is the true integrated result over the imaging path. However, the same interpretation is not true in samples with depth-varying optical axis.8

To construct the local optical axis and retardance (denoted as ϕ and κ, respectively) at a particular depth in a birefringent sample, the corresponding i’th layer (represented as the i’th image pixel in the A-scan) is modeled as a linear retarder whose Jones matrix Ji is:11


where JL(ϕi,κi) represents the Jones matrix of a linear retarder with retardance of κi and optical axis ϕi. R(ϕ) and Λ(κ) are the standard rotation matrix and phase retardation matrix:11


Since only the round-trip signal can be obtained in OCT measurement, the round-trip cumulative Jones matrix of the top n layers can be represented as:


where Jn is the local Jones matrix of the n’th layer in Eq. (2). JST(n1) is the single-trip Jones matrix of the top (n1) layers:


From Eq. (4), the local Jones matrix of the n’th layer can be calculated as:


The local optical axis ϕ and retardance κ can then be calculated from JnTJn by eigen-decomposition.7

From Eq. (6), both round-trip Jones matrix of the top n layers JRT(n) and the single-trip Jones matrix of the top (n1) layers JST(n1) are needed to calculate the local Jones matrix Jn. The JRT(n) in a birefringent sample is also a linear retarder:8


where δn and θn are the cumulative retardance and optical axis measured in PSOCT from Eq. (1). In other words, the round-trip Jones matrix JRT(n) in Eq. (7) can be constructed completely from the PSOCT measurements.

To calculate the single-trip Jones matrix JST(n1), a recursive algorithm can be applied starting from the first layer where JST(1)=J1=JL(θ1,δ1). Once the local retardance and optical axis are determined for the n’th layer, the next single-trip cumulative Jones matrix for the top n layers is constructed as:


where JL(ϕn,κn) is the local Jones matrix of the n’th layer in the form of a linear retarder [Eq. (2)] and is constructed from calculated ϕn and κn. The procedure from Eqs. (6) to (8) is then repeated for the next depth layer until all pixels are calculated.

Instead of calculating the local retardance κn from JnTJn along with the optical axis ϕn, we obtained κn from the PSOCT measurements of the two neighboring layers as reported previously.8 Briefly, from the PSOCT measurements obtained at depth layer n1 (δn1 and θn1) and depth layer n (δn and θn), the following matrix is constructed:


The local retardance κn was then calculated from the eigen values of the N(n) matrix.7



The above algorithm was tested on PSOCT images acquired using a bulk-optic frequency domain PSOCT system described elsewhere.12 The system has a central wavelength of 844 nm. The depth and lateral resolution of the system are 6.7 and 37 μm in air, respectively. Imaging speed is 50 k Alines/s. The structure image I, cumulative phase retardation δ, relative optical axis θ were calculated by Eq. (1). A 3×3 (pixel) median filter was applied to the measured H and V components to remove the speckle noise prior to applying the proposed algorithm.

An initial test was conducted to verify the algorithm in a tendon sample with uniform fiber orientation. A piece of chicken tendon was slightly stretched and mounted so that the sample surface was roughly perpendicular to the incident light. A series of PSOCT images were obtained and processed using the aforementioned algorithm while the tendon sample was rotated from around 10 to 80 deg with a 10 deg step. The 0 deg was defined as the position when tendon fiber orientation was parallel with the B-scan direction [Fig. 1(a)]. The calculated fast optical axis angle was shifted by 90 deg to display the fibrous orientation, which is along the “slow” optical axis.

Fig. 1

PSOCT imaging of a tendon sample rotated from 10 to 80 deg. (a) An illustration of the sample rotation. (b) Sample images obtained at 10, 30, and 70 deg: intensity I[45dB,85dB], cumulative retardance δ[0,π/2], cumulative axis θ[π/2,π/2], local retardance κ[0,0.5], and local axis ϕ[π/2,π/2]. (c) The mean and standard deviation of the local retardance κ and optical axis ϕ calculated from the region of interest (ROI) marked in the box in (b). The size bar in the intensity image indicates 0.5 mm.


Banded patterns appeared in the images of cumulative retardance δ and optical axis θ as expected [Fig. 1(b)]. The images of local retardance κ were relative homogeneous, suggesting a uniform distribution of optical retardance in the sample. The calculated images of local optical axis ϕ were also uniform at each rotation angle but had different values at different angles. For quantitative comparison, the mean and standard deviation values of the local retardance and optical axis were calculated for a region of interest (ROI) enclosed by the box in Fig. 1(b). As shown in Fig. 1(c), the mean local retardance remained stable at different rotation angles. However, the mean local axis correctly retrieved the sample orientation. The axis measurements were well-fitted by a linear regression of y=0.995x3.6deg (R2=0.997).

We next tested the algorithm in a sample with depth-varying optical axis. A specimen was constructed by stacking two pieces of chicken tendons at an angle. The upper tendon sample was positioned at 72deg, and the lower piece was positioned at about 13deg. In the intensity image [Fig. 2(a)], the boundary of the two pieces of tendon can be vaguely identified. The boundary disappeared in the cumulative retardance [Fig. 2(b)] and optical axis [Fig. 2(c)] images that show continuous banded patterns across the boundary. The image of local retardance [Fig. 2(d)] has a uniform appearance suggesting similar retardance in both tendon pieces. However, the two pieces show a distinct contrast in the image of local optical axis [Fig. 2(e)], while the axis appears to be relatively homogeneous within each piece. Figure 2(f) shows the histograms of local retardance and axis obtained from the two small ROI shown in Fig. 2(d) and 2(e). The calculated optical axis within the two ROIs in Fig. 2(f) is 69.5±3.7deg in the upper piece and 16.0±6.9deg in the lower piece; both are consistent with the specimen configuration.

Fig. 2

PSOCT images of two pieces of tendon samples stacked at an angle. (a) intensity; (b) cumulative retardance; (c) cumulative axis; (d) local retardance; (e) local axis; (f) distributions of local retardance and axis values obtained from the two regions of interest (ROI) shown in (d) and (e). The size bars in (a) indicate 0.5 mm.


Figure 3 shows the en face images extracted along the two dotted lines in Fig. 2(a). Images were filtered by a 5×5 median filter in the en face plane for display. Although the two tendon pieces can be discriminated in the structure image at depths where there was a clear separation [Fig. 3(a)], it is difficult to discriminate the thin layer of tissue with a different fiber orientation in the structure image at another depth [Fig. 3(f)]. Compared to other images, the local axis images [Fig. 3(e) and 3(j)] are the most effective in revealing fiber orientation changes at different imaging depths.

Fig. 3

The en face PSOCT images of the same tendon sample used in Fig. 2. The size bar in (a) indicates 0.5 mm.


The algorithm was also applied to analyze PSOCT images of a plastic specimen. The intensity image [Fig. 4(a)] appears to be homogeneous with little structural variation. However, the cumulative retardance [Fig. 4(b)] shows the typical birefringent band structures. The inhomogeneous banded patterns in the cumulative optical axis [Fig. 4(c)] suggest a varying optical axis, which is clearly confirmed in the local optical axis [Fig. 4(e)]. Figure 4(f)4(j) are the corresponding en face images extracted along the two dashed lines in Fig. 4(a). The images of local axis reveal some polarization features that are not discernible in the cumulative results or the structural images. For example, the “bow tie” pattern in Fig. 4(j2) suggests a twisted optical axis distribution in the en face plane that is gradually formed from a shallow depth [Fig. 4(j1)] to a deep depth [Fig. 4(j2)].

Fig. 4

PSOCT images of a plastic sample. (a)–(e): cross-sectional images. (f1)–(j1) and (f2)–(j2) en face images extracted along the upper and lower dashed lines in (a), respectively. The size bars shown in the intensity images represent 0.5 mm.




We demonstrated an algorithm that can successfully extract depth-resolved local optical axis from conventional PSOCT measurements. Different from the iterative method used previously in Jones matrix-based OCT systems,45.6.7 we used a reconstructed Jones matrix [Eq. (8)] in the calculation instead of the raw calculated Jones matrix. In addition, the local retardance was calculated using PSOCT results obtained in two consecutive layers.8 We believe these modifications helped to control error propagation in the calculation process. This proposed method can be applied to existing PSOCT systems that use a single circularly polarized incident light to map local optical axis and retardance in birefringent samples with negligible diattenuation. This approach can be also extended to Jones matrix-based systems as well to completely characterize samples with strong optical diattenuation.


This project is supported in part by a National Science Foundation Grant CBET-0643190.



M. Heeet al., “Polarization-sensitive low-coherence reflectometer for birefringence characterization and ranging,” J. Opt. Soc. Am. B 9(6), 903–908 (1992).JOBPDE0740-3224http://dx.doi.org/10.1364/JOSAB.9.000903Google Scholar


D. Huanget al., “Optical coherence tomography,” Science 254(5035), 1178–1181 (1991).SCIEAS0036-8075http://dx.doi.org/10.1126/science.1957169Google Scholar


S. Guoet al., “Depth-resolved birefringence and differential optical axis orientation measurements with fiber-based polarization-sensitive optical coherence tomography,” Opt. Lett. 29(17), 2025–2027 (2004).OPLEDP0146-9592http://dx.doi.org/10.1364/OL.29.002025Google Scholar


M. Todorovicet al., “Determination of local polarization properties of biological samples in the presence of diattenuation by use of Mueller optical coherence tomography,” Opt. Lett. 29(20), 2402–2404 (2004).OPLEDP0146-9592http://dx.doi.org/10.1364/OL.29.002402Google Scholar


N. Kempet al., “Depth-resolved optic axis orientation in multiple layered anisotropic tissues measured with enhanced polarization-sensitive optical coherence tomography (EPS-OCT),” Opt. Express 13(12), 4507–4518 (2005).OPEXFF1094-4087http://dx.doi.org/10.1364/OPEX.13.004507Google Scholar


S. Makitaet al., “Polarization contrast imaging of biological tissues by polarization-sensitive Fourier-domain optical coherence tomography,” Appl. Opt. 45(6), 1142–1147 (2006).APOPAI0003-6935http://dx.doi.org/10.1364/AO.45.001142Google Scholar


S. MakitaM. YamanariY. Yasuno, “Generalized Jones matrix optical coherence tomography: performance and local birefringence imaging,” Opt. Express 18(2), 854–876 (2010).OPEXFF1094-4087http://dx.doi.org/10.1364/OE.18.000854Google Scholar


C. FanG. Yao, “Mapping local retardance in birefringent samples using polarization sensitive optical coherence tomography,” Opt. Lett. 37(9), 1415–1417 (2012).OPLEDP0146-9592http://dx.doi.org/10.1364/OL.37.001415Google Scholar


Z. LuD. K. KasaragodS. J. Matcher, “Optic axis determination by fiber-based polarization-sensitive swept-source optical coherence tomography,” Phys. Med. Biol. 56(4), 1105–1122 (2011).PHMBA70031-9155http://dx.doi.org/10.1088/0031-9155/56/4/014Google Scholar


C. FanG. Yao, “Correcting optical axis calculation in polarization-sensitive optical coherence tomography,” IEEE Trans. Biomed. Eng. 57(10), 2556–2559 (2010).IEBEAX0018-9294http://dx.doi.org/10.1109/TBME.2010.2053927Google Scholar


R. C. Jones, “A new calculus for the treatment of optical systems,” J. Opt. Soc. Am. 31(7), 488–493 (1941).JOSAAH0030-3941http://dx.doi.org/10.1364/JOSA.31.000488Google Scholar


C. FanG. Yao, “Single camera spectral domain polarization-sensitive optical coherence tomography using offset B-scan modulation,” Opt. Express 18(7), 7281–7287 (2010).OPEXFF1094-4087http://dx.doi.org/10.1364/OE.18.007281Google Scholar

Chuanmao Fan, Gang Yao, "Mapping local optical axis in birefringent samples using polarization-sensitive optical coherence tomography," Journal of Biomedical Optics 17(11), 110501 (9 October 2012). http://dx.doi.org/10.1117/1.JBO.17.11.110501

Optical coherence tomography

Algorithm development

Wave plates


Iterative methods

Biomedical optics

Imaging systems

Back to Top