A single input polarization sensitive optical coherence tomography method

Abstract

The application discloses a single-input polarization-sensitive optical coherence tomography (PS-OCT) method and system based on polarization state modulation and tracking (PSMT), and aims to solve the problem that the polarization state vector is highly concentrated on the Poincare sphere due to the no or weak birefringence tissue of the shallow layer in the traditional single-input PS-OCT in complex layered biological tissues such as fundus, so that the optical axis and the phase delay calculation error are large. The method introduces a fixed time delay to one of the signals collected in the PS-OCT system, generates a gradient phase difference which changes linearly with depth, and thus forms a stable polarization modulation on the Poincare sphere, so that the originally chaotic polarization vector of the no / weak birefringence tissue is arranged in order. Then, the DDG (discrete differential geometry) method is used to calculate the subnormal vector under the modulation state, that is, the cumulative optical axis, on the Poincare sphere, the optical axis offset caused by polarization modulation is calculated for the normal birefringence tissue by setting a threshold value and performing space vector operation, and then the real optical axis is corrected. Experimental verification shows that the method can significantly reduce the error caused by the no / weak birefringence tissue, improve the accuracy of the local polarization information of the deep tissue, and provide a reliable imaging technology for the diagnosis of some diseases.

Description

Technical Field

[0001] This invention belongs to the field of polarization-sensitive optical coherence tomography (PS-OCT) technology, specifically relating to an improved single-input PS-OCT system and algorithm, particularly suitable for depth-resolved polarization characteristic imaging analysis of multilayered complex birefringent biological tissues (such as the posterior sclera, dermal collagen fibers, and tooth enamel microstructures) in ophthalmology, dermatology, dentistry, and many other fields. This invention primarily addresses the significant errors that arise in traditional single-input PS-OCT systems when dealing with tissues that are non-birefringent or weakly birefringent. Background Technology

[0002] Polarization-sensitive optical coherence tomography (PS-OCT) is a non-invasive imaging method that combines traditional OCT technology with polarization characteristic analysis techniques. It can non-invasively acquire the polarization characteristics (such as optical axis direction and phase retardation) and structural information of biological tissues. It has significant applications in ophthalmology, dermatology, dentistry, and other fields. For example, in ophthalmology, the posterior sclera, as the mechanical support structure of the eyeball, undergoes pathological remodeling of its collagen fiber arrangement, density, and tissue integrity as the disease progresses. In high myopia, the diameter of posterior scleral collagen fibers decreases and their arrangement becomes disordered, leading to significant changes in their polarization characteristics. Traditional OCT relies solely on the intensity of reflected light to construct structural images and cannot capture changes in polarization characteristics. PS-OCT, however, analyzes the dynamic evolution of the polarization state of polarized light passing through the sample tissue with depth, enabling precise quantification of the polarization characteristics of collagen fibers. This allows for non-invasive assessment of tissue structural integrity and provides crucial quantitative imaging indicators for disease progression.

[0003] Traditional single-input PS-OCT systems are considered to have significant clinical translational potential due to their simple hardware structure and strong resistance to motion artifacts. Polarization State Tracking (PST) is currently the most efficient method for solving local polarization information based on single-input PS-OCT. The core assumption of this method is that the polarization state of polarized light can be represented by Stokes parameters, which, after standardization, are represented as a polarization vector on a Poincaré sphere. When incident light passes through sample layers with the same birefringence properties, the points obtained on the Poincaré sphere form a trajectory. Discrete Differential Geometry (DDG) is used to determine the subnormal vector of the plane determined by the trajectory and the angle of rotation of adjacent points around the subnormal vector, thereby determining the optical axis and phase delay. Since polarized light passes through the sample once during incident and once during reflection, the measured polarization information is cumulative and cannot obtain the local polarization characteristics we want. PST uses the modeling of the Poincaré sphere to perform an inverse transformation on the measured polarization data using the polarization information of the upper layer, and recursively obtains the local polarization information of a certain layer we want, thus achieving depth resolution. While the PST method performs well in detecting ordinary homogeneous birefringent tissues, in complex layered biological tissues (such as the fundus), the shallow, non-birefringent or weakly birefringent tissues (e.g., the retina, choroid) have little influence on the polarization state of the polarized light passing through them. Consequently, the obtained Stokes parameters are highly concentrated on the Poincaré sphere, resulting in negligible differences between adjacent polarization state vectors. This makes it difficult to stably determine the trajectory and second normal vector on the Poincaré sphere. This situation not only causes significant errors in calculating the optical axis direction and phase delay at the current layer, but also, because the recursive algorithm used propagates these errors to deeper birefringent tissues (such as the posterior sclera), ultimately leading to very large errors in our analysis of collagen fiber orientation.

[0004] To overcome this limitation, some multi-input PS-OCT systems acquire multi-polarization state signals by time-division or frequency-division to improve polarization information processing capabilities. However, these systems have complex hardware, halved imaging depth, and are susceptible to motion artifacts, making it difficult to meet actual clinical needs.

[0005] Therefore, when detecting complex samples containing non-birefringent or weakly birefringent tissues, how to eliminate the interference of shallow non-birefringent / weakly birefringent tissues on the calculation of deep polarization information has become a bottleneck problem for the clinical application of single-input PS-OCT technology. Summary of the Invention

[0006] This patent innovatively introduces the polarization modulation and tracking (PSMT) method into the traditional single-input PSOCT imaging algorithm, breaking through the technical limitations of traditional single-input PS-OCT in detecting complex birefringent biological tissues containing weak birefringent layers, such as the eye. This enables accurate calculation of polarization information of tissues containing no or weak birefringence, providing an efficient imaging tool for the non-invasive and accurate diagnosis of certain diseases.

[0007] The system has been improved. The interference signal of the traditional single-input PSOCT system is split into two beams, which are then separated into horizontal and vertical components by a polarization beam splitter. These components are then received by two balanced detectors. The balanced detectors convert the interference optical signal into an electrical signal, which is then transmitted to the data acquisition card for data acquisition. During this process, a delay unit is added to one of the channels to create a specific time delay for that signal, thereby providing polarization modulation functionality.

[0008] By introducing a time delay in the system's detection channel, a gradient phase difference that varies linearly with depth is generated between the two-channel interference signals. This gradient phase difference forms a stable polarization modulation background on the Poincaré sphere, causing the polarization state vectors of non- / weakly birefringent structures to be arranged in an orderly manner along the Q-axis, thereby eliminating its random interference.

[0009] Optical axis offset calculation: Based on the modulated Stokes matrix, the optical axis B under polarization modulation state is calculated using Discrete Differential Geometry (DDG). mod (z), and separate the optical axis offset ΔB(z) introduced by polarization modulation, thereby correcting the true optical axis B. real (z)=B mod (z)-ΔB(z)

[0010] Birefringence structure discrimination: A cosine similarity threshold (angle ≤ 8.1°) is used. 8.1° (corresponding to a cosine similarity of 0.99) is chosen as the threshold because in statistics and signal processing, when the cosine similarity reaches 0.99, the directions of the two vectors are very close, and their differences can be considered negligible. This threshold is used to distinguish birefringence structures from non-birefringence / weak birefringence structures. If B mod If the angle between (z) and the Q axis is less than the threshold, it is determined to be a non- / weak birefringent structure, and its optical axis is preserved; otherwise, offset correction is performed to restore the true polarization information.

[0011] The single-input PS-OCT system of this invention employs a MEMS-VCSEL light source (1060nm, 600kHz) and a delay unit, supporting depth-resolution imaging of birefringent biological tissues while maintaining hardware simplicity. Compared with existing technologies, this invention has the following advantages.

[0012] Improved anti-interference capability: Gradient phase difference modulation orders the polarization state of shallow non- / weakly birefringent structures, reducing the optical axis calculation error by more than 70%, and the minimum detectable phase delay is below 0.3° / μm.

[0013] Depth resolution accuracy optimization: Error propagation effect is effectively suppressed, local optical axis accuracy of deep tissues (such as posterior sclera) is improved by more than 90%, and the accuracy of collagen fiber orientation analysis is significantly enhanced.

[0014] Clinical applicability: The system requires no complex hardware modifications, maintains complete imaging depth (≥2.5mm), and has better resistance to motion artifacts than multi-input PS-OCT systems, making it suitable for clinical testing. Attached Figure Description

[0015] Figure 1 This is a diagram of a single-input PS-OCT system that provides polarization modulation functionality. SS-Laser represents a swept-frequency laser source, Coupler is a fiber coupler, ODL is an optical delay line, PC1-4 are polarization controllers, CIR represents an optical circulator, L1 and L2 are lens 1 and lens 2 respectively, GM is a galvanometer, QWP is a quarter-wave plate, BS is a beam splitter, PBS represents a polarization beam splitter, BPD1 and BPD2 are balanced photodetectors 1 and 2 respectively, and Signal delay is a signal delay module.

[0016] Figure 2 The implementation process of the PSMT method. (a) Schematic diagram of delayed signal generation. (b) Schematic diagram of polarization modulation generation. (c) Schematic diagram of optical axis offset calculation introduced by polarization modulation. (d) Processing flow of the PSMT method. This includes the determination and constraint process for the absence of birefringence under polarization modulation, the optical axis correction process for the presence of birefringence, and the calculation process for the local optical axis.

[0017] Figure 3 Numerical simulation results. (a) Simulation of non-birefringent microstructure. (b) Simulation of normal birefringent microstructure. The method presented in this paper has no effect on normal birefringent microstructure.

[0018] Figure 4 Polarization results and statistical analysis of non-birefringent artificial samples obtained by the PSMT method. (a) Structural diagram of the non-birefringent artificial sample. (b)-(d) Polarization results obtained by the PSMT method. (e) Comparison of phase delay results between the two methods in the red box area of ​​Figure (c).

[0019] Figure 5Human fundus PS-OCT imaging results. (a) Fundus structure image. (b)(c) Stocks matrix diagrams under polarization-free and polarization-modulated conditions. (d)–(f) Results obtained by the PST method, including: subnormal vector, phase delay, and optical axis. (g)–(i) Results obtained by the corresponding PSMT method.

[0020] Figure 6 Presentation of PS-OCT enface results for the posterior sclera. (a)(b) Structure diagrams of the retina and posterior sclera enface. (c) Phase retardation diagram of the posterior sclera enface. (d) Local optical axis diagram obtained by the PST method. (e) Local optical axis diagram obtained by the PSMT method. Detailed Implementation

[0021] The following description, with reference to the accompanying drawings, illustrates several preferred embodiments of the present invention to make its technical content clearer and easier to understand. The present invention can be embodied in many different forms, and the scope of protection of the present invention is not limited to the embodiments mentioned herein.

[0022] This invention proposes a single-input PS-OCT imaging method based on polarization modulation and tracking (PSMT) to solve the problem of optical axis and phase delay calculation errors caused by excessive concentration of polarization state vectors in traditional PST methods for samples with no or weak birefringence. The specific implementation is as follows:

[0023] Introducing a specific time delay into one of the channels of a traditional single-input PS-OCT system, such as Figure 1 As shown. Both signals are acquired using the same clock. A fixed time delay introduces a variable phase difference at different depths in the two OCT images. The phase difference introduced by the two channels. It can be represented as:

[0024]

[0025] Where Δλ is the system bandwidth. sweep λ is the sweep rate of the light source. λ0 is the center wavelength of the light source. z represents the image depth (i.e., optical path difference), in millimeters. In the above formula, all parameters except z are constants given a fixed system parameter. Therefore, it can be seen that... It is linearly related to the depth z. That is, the time delay applied to a signal produces a phase difference that varies linearly with the depth, which we call the gradient phase difference.

[0026] With the introduction of gradient phase difference, the calculation of the Stokes parameters can be expressed as follows:

[0027]

[0028] Among them, A H and A V The amplitude information of the two channels is represented by θ(z), which represents the phase difference between the two channels. z represents the depth position reached from the sample surface along the A-Scan direction, and I... mod Q represents the total intensity of the light after polarization modulation. mod and U mod This describes the linear polarization components of the light in the x and y directions after polarization modulation, V mod It is the circularly polarized component of the light after polarization modulation. The introduction of the gradient phase difference does not affect the value of Q, but only affects V and U. Therefore, the modulation effect of the gradient phase difference on the polarization state is equivalent to rotating the polarization state vector around the Q axis by a certain angle on the Poincaré sphere.

[0029] Figure 2 (b) describes the polarization modulation application process. The blue dots P1, P2, and P3 represent the polarization state vectors received by the system detector without polarization modulation. After polarization modulation is applied, the polarization state vectors received by the system detector are Pm1, Pm2, and Pm3 (red dots). They rotate about the Q-axis respectively. At this angle, when the system probes a non-birefringent sample, the polarization state vectors at different depths will also form an ordered arrangement (the arrangement pattern tends to be consistent with the polarization background). Therefore, the optical axis and phase delay of the non-birefringent tissue can be calculated using traditional algorithms. Finally, by adding constraints, the interference of polarization modulation on other tissues can be limited.

[0030] Polarization modulation introduces optical axis shift. Figure 2 (c) illustrates the calculation process for the optical axis offset. For ease of calculation, we first remove the DC component of the phase difference from the three adjacent points (red dots Pm) (i.e., they rotate together around the Q-axis by a specific angle in the opposite direction). This eliminates the global phase shift, aligning the phases of all vectors around a common reference axis, facilitating subsequent calculations of the optical axis offset. The pink ring represents the polarization state vector Pm' after removing the DC component. The blue arrow is the secant vector formed by adjacent vectors P, denoted by T. The red arrow is the secant vector T' formed by adjacent Pm'. The green arrow is the polarization modulation vector Tb. The binormal vector (pink arrow) formed by the three adjacent Pm' (pink rings) can be represented as:

[0031]

[0032] Among them, T' z and T' z+1 We can use T and Tb to represent it, then

[0033]

[0034] The first term in formula (4) is B. real (z), the second term is a constant that only affects the vector length and not the vector direction. On a Poincaré sphere, the normalized vectors are all of unit length. Therefore, the second term can be ignored. The third term is the optical axis offset caused by polarization modulation, which we denot as ΔB(z). Since all parameters in the third term are known, ΔB(z) can be calculated precisely. Formula (4) can be rewritten as:

[0035] B mod (z)=B real (z)+ΔB(z) (5)

[0036] Thus obtain

[0037] B real (z)=B mod (z)-ΔB(z) (6)

[0038] The optical axis shift caused by polarization modulation can be perfectly recovered using formula (6). Therefore, polarization modulation has no effect on samples with birefringence. For samples without birefringence or with weak birefringence, we reduce its influence on the polarization state of other tissues by introducing a discrimination and constraint process. Figure 2 (d) describes the entire processing flow of the algorithm. With polarization modulation introduced, the Stokes parameters for each pixel are calculated. Furthermore, the subnormal vector (B) under polarization modulation is calculated. mod And optical axis offset (ΔB). If the sample birefringence is weak, B mod The angle between the sample and the Q axis will be very small. If the sample has strong birefringence, then B... mod There exists a specific and stable angle between the axis and the Q-axis. Therefore, we set a threshold as the evaluation criterion.

[0039] In statistics, cosine similarity is commonly used to measure the similarity of the directions of two vectors, which is the cosine of the angle between the two spatial vectors. The closer the cosine similarity is to 1, the more similar the directions of the two vectors are. When the cosine similarity is 0.99, it is generally considered that the directions of the two vectors are very close (the angle between them is 8.1°). Therefore, we set the threshold to 8.1°. If B mod If the angle between the optic axis and the Q axis is less than this threshold, it is determined to be a non-birefringent sample, and its cumulative optical axis remains unchanged (B). real =B mod Conversely, ΔB is used to correct the optical axis (B). real =B mod -ΔB).

[0040] Once a stable subnormal vector (cumulative optical axis) is obtained, the phase difference introduced by polarization modulation is removed, and the phase delay is calculated using the stable subnormal vector. This significantly reduces the phase delay error. After obtaining a stable phase delay, a depth-resolved local optical axis is obtained through three-dimensional rotation.

[0041] Perform numerical simulations, such as Figure 3 When simulating a sample without birefringence, the original Stokes parameters are almost constant, and the polarization state vectors are relatively clustered on the Poincaré sphere. After applying polarization modulation, the filtered polarization state vectors are rearranged and are highly consistent with the Q-axis, which significantly reduces the phase delay error.

[0042] Artificial sample experiments, such as Figure 4 The results on the Stokes vector, phase delay, and local optical axis of non-birefringent artificial samples show that the PSMT method can effectively reduce the error of non-birefringent samples, making the calculation results more accurate.

[0043] Double-layer sample and posterior scleral imaging, such as Figure 5 , 6 Experimental results demonstrate that in bilayer samples (without birefringence in the upper layer and with birefringence in the lower layer) and in actual fundus imaging, the PSMT method can both suppress the interference of shallow weak birefringence on deep polarization information and accurately reproduce the true optical axis of posterior scleral collagen fibers, thereby improving the uniformity and accuracy of local polarization information.

[0044] In summary, this invention improves upon the traditional single-input PS-OCT system to achieve polarization modulation, and combines cosine similarity discrimination and spatial vector operation to correct optical axis offset. This effectively overcomes the limitations of traditional methods in samples with no or weak birefringence, and provides reliable technical support for the accurate acquisition of local polarization information of deep tissues (such as the posterior sclera of the fundus).

Claims

1. A single-input PS-OCT imaging method based on polarization modulation and tracking, characterized in that... To reduce the significant errors in traditional single-input PS-OCT imaging of complex biological tissues with no or weak birefringence layers, including: Step 1: Add a signal delay unit to the output channel of one of the balanced detectors in the single-input PS-OCT system to generate a specific time delay for one signal, thereby achieving polarization modulation; Step 2: Based on the acquired modulated Stokes parameter data, calculate the cumulative optical axis B of the sample under polarization modulation. mod And phase delay, and calculate the optical axis offset ΔB introduced by polarization modulation; Step 3: Set a cosine similarity threshold to identify birefringent structures. If it is a non-birefringent / weakly birefringent structure, retain B. mod Otherwise, through B mod Subtracting ΔB yields the true optical axis, avoiding any impact on normal birefringent structures; Step 4: Re-estimate the local phase delay and the local optical axis of the sample using the true cumulative optical axis. Perform pseudo-color encoding on the optical axis and phase delay to obtain more accurate PSOCT images of complex biological tissues containing no / weak birefringence layers.

2. The method according to claim 1, characterized in that, The specific description of step 1 is as follows: By introducing a specific time delay and using the same clock to acquire both signals, the fixed time delay introduces a variable phase difference at different depths in the two OCT images. The phase difference introduced by the two channels... It can be represented as: Where Δλ is the system bandwidth. sweep λ is the sweep rate of the light source. λ0 is the center wavelength of the light source. z represents the image depth (i.e., optical path difference), in millimeters. In the above formula, all parameters except z are constants given a fixed system parameter. It is linearly related to the depth z. That is, the time delay applied to a signal produces a phase difference that varies linearly with the depth, which we call the gradient phase difference. With the introduction of gradient phase difference, the calculation of the Stokes parameters can be expressed as follows: Among them, A H and A V The amplitude information of the two channels is represented by θ(z), which represents the phase difference between the two channels. z represents the depth position reached from the sample surface along the A-Scan direction, and I... mod Q represents the total intensity of the light after polarization modulation. mod and U mod This describes the linear polarization components of the light in the x and y directions after polarization modulation, V mod It is the circularly polarized component of the light after polarization modulation. The introduction of the gradient phase difference does not affect the value of Q, but only affects V and U. Therefore, the modulation effect of the gradient phase difference on the polarization state is equivalent to rotating the polarization state vector around the Q axis on the Poincaré sphere by a certain angle.

3. The method according to claim 1, characterized in that, The optical axis offset ΔB(z) in step 2 is calculated using the following formula: Among them, T z ′ is the tangent vector of the curve formed by adjacent depth points on the Poincaré sphere after polarization modulation, T z ′=T z +Tb z T z Tb is the original tangent vector. z The tangent vector offset introduced for polarization modulation. The first term in equation (3) is B. real The second term (z) is a constant that only affects the vector length. On a Poincaré sphere, vectors are all unit length after standardization, so the second term can be ignored. The third term is the optical axis offset caused by polarization modulation, which we denote as ΔB(z). Since all parameters in the third term are known, ΔB(z) can be calculated precisely.

4. The method according to claim 1, characterized in that, The discrimination threshold in step 3 is specifically described as follows: When the cosine similarity is 0.99, it is generally believed that the directions of the two vectors are very close (the included angle is 8.1°). Therefore, we set the threshold to 8.1°. If B mod If the angle between the optic axis and the Q axis is less than this threshold, it is determined to be a non-birefringent sample, and its cumulative optical axis remains unchanged (B). real =B mod Conversely, ΔB is used to correct the optical axis (B). real =B mod -ΔB).

Images