Dual camera misregistration sampling imaging OCT system and method
By introducing a second beam splitter and a dual-camera architecture, combined with a deep neural network model, the problems of low pixel density and slow imaging speed in the full-field OCT system were solved, achieving efficient high-speed and high-resolution imaging, and improving system stability and image quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING XIGUANG MEDICAL TECHNOLOGY CO LTD
- Filing Date
- 2026-03-24
- Publication Date
- 2026-07-07
AI Technical Summary
In existing full-field OCT systems, the use of large FWC cameras results in low pixel density and insufficient imaging sampling rate. Furthermore, traditional beam shifting techniques increase image acquisition time, reducing system stability and imaging speed.
By introducing a second beam splitter and a dual-camera architecture, and through spatial subpixel misalignment acquisition, combined with deep neural network models to reconstruct images, the synchronous acquisition of twice the spatial sampling information is achieved, eliminating the time lag problem and improving system stability and resolution.
High-resolution imaging was achieved without increasing time overhead, motion artifacts were eliminated, system stability and image sharpness were improved, signal-to-noise ratio was enhanced, and high-frequency detail information was recovered.
Smart Images
Figure CN122345931A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of full-field optical coherence microscopy, and in particular to a dual-camera misaligned sampling imaging OCT system and method. Background Technology
[0002] Optical coherence tomography (OCT) is a tomographic imaging technique based on the principle of low coherence interference. It utilizes the low coherence characteristics of a broadband light source to measure the interference signal between the backscattered or reflected light signal from different depths of biological tissue and the reference light, and then reconstructs the sample image through computer processing.
[0003] Related technologies include full-field optical coherence microscopy (FF-OCM), an advanced three-dimensional imaging technique that combines the depth-penetrating capability of optical coherence tomography (OCT) with the advantages of high-resolution microscopy. Unlike traditional OCT, which acquires longitudinal cross-sectional images, FF-OCM primarily acquires orthographic images, eliminating depth-of-field limitations and allowing the use of high numerical aperture optical systems to achieve higher lateral resolution. It acquires interferometric images using an area array camera and performs phase demodulation, achieving high lateral and axial resolution simultaneously with relatively simple experimental setups. In full-field OCT microscopy, the area array camera is a key factor determining the system's imaging performance, and the camera's trap depth (FWC) is a crucial determinant of the system's signal-to-noise ratio. However, large FWC cameras generally have large individual pixel areas, resulting in insufficient pixel density, which in turn leads to insufficient sampling rate. Currently, beam imaging is used to increase sampling density without changing the camera. The shift technique, which involves adding a beam shifter to the camera sensor and shifting the beam two or four times to improve resolution, has the following drawbacks: it doubles the image acquisition time, reduces the system's imaging speed, and the beam shifting hardware introduced by moving parts will reduce the system's stability. It also has disadvantages such as generating motion noise or image blurring and instability for the full-field OCT system. Based on the above analysis of the development status of this technology field, the existing technology lacks a full-field optical coherence tomography scheme that introduces a second beam splitter and a dual-camera optical path architecture, utilizes spatial sub-pixel misalignment to obtain high-frequency spatial information, and combines phase demodulation and deep learning reconstruction algorithms. Summary of the Invention
[0004] The purpose of this invention is to provide a dual-camera misaligned sampling imaging OCT system and method, which aims to solve the above-mentioned problems in the prior art.
[0005] According to a first aspect of the present invention, a dual-camera misaligned sampling imaging OCT system is provided, comprising: Low-coherence light source module, used to provide light source; The interference microscopy optical path module is used to split the light source into beams by the first beam splitter for information modulation, obtain the interference of reflected light and scattered light, and then converge them through the tube lens to obtain the interference beam. The dual-channel detection module is used to split the interference beam through the second beam splitter and acquire the first and second image signals respectively through the spatial sub-pixel misaligned dual cameras; The synchronization control module is used to control the synchronous acquisition of data by the two cameras. The data processing module is used to acquire a synchronization sequence including the first image signal and the second image signal, demodulate the synchronization sequence and synthesize a sparse image, and use a deep neural network model to reconstruct the physical imaging based on the sparse image.
[0006] According to a second aspect of the present invention, a dual-camera misaligned sampling imaging OCT method is provided, comprising: The light source is provided by a low-coherence light source module; The light source is split into beams by the first beam splitter through the interference microscopy optical path module to modulate the information, and the light after reflection and scattering is obtained to interfere with each other. The interference beam is then converged by the tube lens. The interference beam is split by the second beam splitter through the dual-channel detection module, and then acquired by the two cameras with spatial sub-pixel misalignment to obtain the first and second image signals. The dual cameras are controlled to acquire data synchronously via a synchronization control module. The data processing module acquires a synchronization sequence including the first and second image signals. Based on the synchronization sequence, it demodulates and synthesizes a sparse image. Based on the sparse image, a deep neural network model is used to reconstruct the physical imaging.
[0007] According to a third aspect of the present invention, an electronic device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the dual-camera misaligned sampling imaging OCT method provided in the second aspect of the present disclosure.
[0008] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, on which an information transmission implementation program is stored, which, when executed by a processor, implements the steps of the dual-camera misaligned sampling imaging OCT method provided in the second aspect of the present disclosure.
[0009] The technical solution provided by this invention includes the following beneficial effects: Introducing a second beam splitter and a dual-camera architecture, using spatial multiplexing instead of temporal scanning, achieves fully synchronous acquisition of two sub-pixel misaligned images. While acquiring twice the spatial sampling information, no additional time overhead is added, completely eliminating the time lag problem in traditional pixel displacement technology, effectively suppressing motion artifacts caused by sample movement, and making it suitable for high-speed, high-resolution imaging of biological tissues; the static beam splitter and dual cameras fix the optical path, eliminating mechanical moving parts used for lateral scanning, avoiding the influence of mechanical wear and vibration on the stability of interference fringes, while reducing the system's control complexity and improving the long-term stability and durability of the optical system; after physical quantity demodulation of the two synchronously acquired signals and then reconstruction using a deep neural network model, it can learn prior knowledge of biological tissue texture, more accurately predict grayscale changes between sub-pixels, effectively eliminating the jagged effect and blurring phenomenon that may be caused by traditional interpolation, so that the final synthesized high-pixel image retains OCT tomography capabilities while having higher image sharpness and signal-to-noise ratio.
[0010] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description
[0011] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0012] Figure 1 This is a schematic diagram of the dual-camera misaligned sampling imaging OCT system according to an embodiment of the present invention; Figure 2 This is a schematic diagram of spatial subpixel misalignment according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the system device architecture according to an embodiment of the present invention; Figure 4 This is a flowchart of the dual-camera misaligned sampling imaging OCT method according to an embodiment of the present invention; Figure 5 This is a schematic diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation
[0013] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.
[0014] System Implementation Example 1 According to an embodiment of the present invention, a dual-camera misaligned sampling imaging OCT system is provided. Figure 1 This is a schematic diagram of a dual-camera misaligned sampling imaging OCT system according to an embodiment of the present invention, as shown below. Figure 1 As shown, the dual-camera misaligned sampling imaging OCT system according to an embodiment of the present invention specifically includes: Low-coherence light source module 10 is used to provide a light source, specifically for: LEDs with a center wavelength of 630nm and a bandwidth of 20nm are used to provide low-coherence light source illumination.
[0015] The interference microscopy optical path module 12 is used to split the light source through the first beam splitter to modulate the information, obtain the interference of the reflected light and the scattered light to form a two-dimensional interference field carrying the sample depth information, and then converge the interference beam through the tube lens. Information modulation refers to the change in physical properties of an object according to the object's structural information during the interaction between light and the object. The process of interferometric microscopic optical path module is relative to information loading.
[0016] The interference microscope optical path module 12 specifically includes: a first beam splitter (Beam-splitter 1, BS1), an objective lens (MO), a reference arm including a reference mirror and piezoelectric ceramic (PZT), a sample arm, and tubes. The first beam splitter is connected to the low coherence light source module, the objective lens is placed between the first beam splitter and the sample arm, and between the first beam splitter and the reference arm, and the tube lens is in the interference light output path of the first beam splitter.
[0017] Interference microscopy optical path module 12 is specifically used for: The light source transmitted by the low-coherence light source module is received by the first beam splitter, and the light source is split into reference light and sample light. The reference light is reflected by the reference mirror, and the sample light is backscattered by the sample arm for information modulation. After being reflected and scattered, the reflected and scattered light are obtained at the first beam splitter to interfere with each other to redistribute energy, and then converged by the tube lens to obtain an interference beam.
[0018] The dual-channel detection module 14 is used to split the interference beam through the second beam splitter and acquire the first image signal and the second image signal through the dual cameras with spatial sub-pixel misalignment. The dual-channel detection module is used to convert interference optical signals into electrical signals.
[0019] The dual-channel detection module 14 is specifically used for: The interference beam output from the interference microscope optical path module is received by the second beam splitter (Beam-splitter 2, BS2), and the interference beam is equally split into a transmitted beam and a reflected beam. The second beam splitter is located in the exit optical path of the tube lens, that is, behind the tube lens. Preferably, the second beam splitter BS2 is a 50:50 broadband unpolarized beam splitter that divides the interference beam into a transmitted beam and a reflected beam equally in terms of energy.
[0020] The first camera is set in the transmission optical path of BS2, and its target surface is located on the focal plane of the tube lens; the second camera is set in the reflection optical path of BS2, and its target surface is located on the focal plane of the tube lens. Preferably, to maximize the effect of subpixel sampling, the first and second cameras use CMOS image sensors with an FSI front-illuminated structure. The metal wiring layer of the FSI structure sensor is located above the photodiode, which results in a relatively low fill factor. In subpixel misalignment imaging, if the photosensitive area of the pixel is too large, there will be a large overlap area between adjacent sampling points in physical space, resulting in a severe spatial averaging effect, i.e., crosstalk between pixels, which blurs high-frequency details.
[0021] A first image signal is acquired based on the transmitted beam by a first camera positioned in the transmission optical path of the second beam splitter; a second image signal is acquired based on the reflected beam by a second camera positioned in the reflection optical path of the second beam splitter, and the image signal is the interference image signal; The first camera and the second camera form a dual-camera system, both of which are large-depth, large-pixel area array cameras. The first camera and the second camera are conjugate in the optical axis direction (Z-axis), meaning that the images captured by the dual cameras must be equally clear and the same size. On the plane perpendicular to the optical axis (XY-axis), the second camera has a spatial sub-pixel misalignment relative to the first camera, meaning that the two-dimensional images captured by the two optical paths are misaligned.
[0022] Figure 2 This is a schematic diagram of spatial subpixel misalignment according to an embodiment of the present invention, as shown below. Figure 2As shown, the effect of misalignment of the interference images of the first camera and the second camera is illustrated. The interpolated pixels only show the effect of the traditional interpolation method, not the method used in the embodiments of this invention.
[0023] Preferably, the misalignment is 0.5 pixels in both the horizontal (X) and vertical (Y) directions.
[0024] Preferably, to ensure that the dual cameras meet the strict sub-pixel misalignment relationship, the following adjustment method is adopted: (1) Optical axis alignment and confocal adjustment First, adjust the position of the first camera so that it is located at the focal plane of the tube lens to obtain a clear sample image; then connect the second camera and adjust its six-axis displacement stage so that its optical axis center coincides with the optical path, and finely adjust it along the Z-axis until the second camera also obtains a clear image, ensuring that the optical paths of the two cameras are consistent and eliminating defocus differences.
[0025] (2) Subpixel misalignment precision adjustment Use a standard resolution board such as USAF 1951 or a high-contrast speckle sample as a target; fine-tune the position of the second camera in the XY plane using a precision displacement stage.
[0026] After calibration using the above-mentioned precise adjustment method, the cross-correlation function of the two camera images is calculated in real time, or the phase difference of specific edge features is monitored; when the center of the second camera image is offset by 0.5 pixels in both the X and Y directions relative to the first camera, the mechanical position is locked; at this time, the sampling points of the two sets of cameras are spatially staggered to form a quincunx sampling grid.
[0027] Synchronization control module 16 is used to control the synchronous acquisition of data from both cameras, specifically for: The TTL trigger signal is output by the synchronization signal generator and connected to the external trigger interface of both the first and second cameras to ensure that the two cameras perform exposure acquisition at the same time.
[0028] The data processing module 18 is used to acquire a synchronization sequence including a first image signal and a second image signal, demodulate the synchronization sequence and synthesize a sparse image, and use a deep neural network model to reconstruct the physical imaging based on the sparse image.
[0029] The data processing module differs from traditional methods that directly interpolate the original image; it adopts a strategy of demodulation followed by fusion. Data processing module 18 specifically includes: The synchronous acquisition module is used to control the piezoelectric ceramic to perform step phase shift, such as the four-step phase shift method or the two-step phase shift method. At each phase shift step, the synchronous control module triggers the simultaneous exposure of the two cameras. The first camera acquires the first image signal, and the second camera acquires the second image signal to form a synchronous sequence. Assuming N phase shifts are acquired, the image sequence obtained by the first camera is as follows: The second camera obtained the image sequence as follows ; The demodulation module is used to perform OCT demodulation calculations based on the synchronization sequence to obtain a first image result and a second image result. The image result includes a full-field OCT intensity image or a phase image. use The full-field OCT intensity image of the first camera was calculated. or phase image ,use The full-field OCT intensity image of the second camera was calculated. or phase image ; and They represent physical quantities that are complementary at the same tomographic depth at the same time but at the same spatial sampling location.
[0030] The sparse sampling module is used to establish a virtual network, fill the virtual network with the first image result and the second image result, and synthesize a sparse image; The sparse sampling module is specifically used for: Establish a virtual network with twice the resolution of a single camera; The pixel values of the first image result are filled into the even-numbered row and even-numbered column coordinates of the virtual network, and the pixel values of the second image result are filled into the odd-numbered row and odd-numbered column coordinates of the virtual network to synthesize a sparse image.
[0031] Taking full-field OCT intensity images as an example, The pixel values are filled into the even-numbered row and even-numbered column coordinates of the virtual grid. ;Will The pixel values are filled into the odd-numbered rows and odd-numbered columns of the virtual grid. This yields a sparse, high-resolution image with plum blossom-shaped sampling. Of these, approximately 50% of the pixels are and The location is in a state of missing values.
[0032] The deep learning module is used to reconstruct physical images based on sparse images using a pre-trained deep neural network model that includes encoder and decoder structures.
[0033] In order to Recover the full-pixel high-resolution image The embodiments of the present invention employ a sparse recovery algorithm based on deep neural networks.
[0034] The deep learning module is specifically used for: The model is pre-trained on a large-scale full-field OCT image dataset, which includes both static and dynamic images, to learn the continuity features of biological tissue edges and the high-frequency distribution patterns of speckle texture.
[0035] sparse image A pre-trained deep neural network model is input, and deep feature representations are extracted through the convolutional layers in the encoder, i.e., the output of the last convolutional layer of the encoder. In this embodiment of the invention, the structure of other layers is not limited. Based on the deep feature representations, the decoder predicts the missing pixel values of the sparse image and reconstructs the physical image. This method utilizes prior knowledge to perform non-linear prediction and filling of missing pixels; that is, it retains the data before the network outputs the final result. The system already contains real sampled values from the first and second cameras, and only uses network-predicted values for missing locations, thus ensuring the physical authenticity of the imaging results.
[0036] Other easily conceivable input forms for the model are also within the scope of protection of this invention; In this embodiment of the invention, the model adopts a convolutional model of U-Net variant or ResNet architecture. The first layer of the network adopts a partial convolution or dilated convolution layer to process the sparse input data and avoid the interference of missing pixels, i.e. zero values, on feature extraction.
[0037] The above technical solutions of the embodiments of the present invention will be illustrated with reference to the following accompanying drawings.
[0038] Figure 3 This is a schematic diagram of the system device architecture according to an embodiment of the present invention, such as... Figure 3 As shown, the core hardware architecture is illustrated, which mainly includes an interferometric microscopy optical path module and a dual-channel detection module. Based on the original FF-OCT imaging device, a second beam splitter and a dual-camera optical path architecture are introduced.
[0039] In summary, addressing the existing problems, this invention presents a dual-camera misaligned sampling imaging OCT system. It introduces a second beam splitter and a dual-camera architecture, utilizing spatial multiplexing instead of temporal scanning to achieve fully synchronous acquisition of two sub-pixel misaligned images. This acquires twice the spatial sampling information without adding any extra time overhead, completely eliminating the time lag problem in traditional pixel-shifting techniques and effectively suppressing motion artifacts caused by sample movement, making it suitable for high-speed, high-resolution imaging of biological tissues. The static beam splitter and dual cameras fix the optical path, eliminating mechanical moving parts used for lateral scanning, avoiding the impact of mechanical wear and vibration on the stability of interference fringes, while also reducing system control complexity and improving the long-term stability and durability of the optical system. Applications: The dual cameras spatially misalign pixels to form a plum blossom-shaped sampling array, doubling the system's spatial sampling frequency; after physical demodulation of the two synchronously acquired signals and reconstruction using a deep neural network model, it can learn prior knowledge of biological tissue texture, more accurately predict grayscale changes between subpixels, effectively eliminating the jagged edges and blurring that may be caused by traditional interpolation. This results in a high-resolution image that retains OCT tomography capabilities while possessing higher image sharpness and signal-to-noise ratio, effectively restoring high-frequency details lost due to pixel averaging effects; the overall solution balances a large field of view with a high signal-to-noise ratio, allowing the use of general-purpose industrial cameras with large pixels and deep wells, achieving high resolution at low cost through dual-camera technology.
[0040] System Implementation Example 2 According to an embodiment of the present invention, a dual-camera misaligned sampling imaging OCT system is provided. Based on the first embodiment of the system, a micro-hole array mask is tightly attached or integrated in front of the photosensitive chips of the first camera and the second camera. Mask material: Made of a material that is completely opaque to light for the working wavelength of around 630nm in this embodiment, such as a chromium (Cr) layer deposited on the sensor surface by photolithography, or an ultra-thin black metal sheet. Micro-aperture distribution: The mask has light-transmitting micro-apertures arranged in a matrix; the arrangement period (pitch) of the micro-apertures is strictly consistent with the pixel pitch of the CMOS sensor; Alignment: The geometric center of each light-transmitting micro-hole is precisely aligned with the photosensitive area of the corresponding pixel on the CMOS sensor, i.e., the center of the photodiode, in the optical axis direction.
[0041] To completely eliminate crosstalk and improve subpixel imaging quality, the aperture design of the micro-aperture is crucial; therefore, the geometric parameters are set as follows: Let the pixel size of the CMOS sensor be P, for example, P = 5μm; In this embodiment of the invention, the diameter D of the light-transmitting micro-hole on the mask is designed to be less than 50% of the pixel size, i.e., D < 0.5P; preferably, the proportion of the micro-hole area to the total pixel area, i.e., the effective fill factor, is controlled between 10% and 25%. The remaining areas are shaded, except for the micro-holes, which cover the edge areas of the pixels, the wiring areas, and the gaps between adjacent pixels.
[0042] To ensure micron-level alignment accuracy, this embodiment recommends using on-chip lithography technology in manufacturing and integration processes. After the CMOS sensor is manufactured, photoresist is directly spin-coated onto the wafer surface; The micro-hole pattern is established on the photoresist by exposure and development using a photomask; An opaque metal film is deposited on the sensor, and the excess is then peeled off, thus forming a micro-hole array that corresponds one-to-one with the pixels directly on the sensor surface; the integrated structure avoids the mechanical vibration and displacement that may occur with separate installation.
[0043] Without a mask, if the incident light spot illuminates the boundary between two pixels, it will be converted into electrical signals by both pixels simultaneously, causing crosstalk. With a mask, light illuminating the pixel edges or boundaries is physically blocked, and only light illuminating the pixel center, i.e., the micro-aperture, can be detected. This ensures that the reading of each pixel represents only the light intensity at its center point, achieving point sampling. In dual-camera subpixel misalignment imaging, the smaller the overlapping area, the greater the amount of independent information when solving equations or reconstructing using deep learning. By compressing the effective photosensitive area to a very small extent through a physical mask, the images acquired by the two cameras have almost no spatial overlap, significantly improving the system's response to high-frequency spatial information, i.e., improving the high-frequency part of the MTF curve, thereby obtaining clearer and sharper super-resolution OCT images.
[0044] System Implementation Example 3 According to an embodiment of the present invention, a dual-camera misaligned sampling imaging OCT system is provided. Based on the second embodiment of the system, the mechanical structure of the detection end is further optimized to solve the problem of the traditional interferometric system being sensitive to environmental vibration. To ensure the long-term stability of subpixel misalignment, this embodiment of the invention provides an integrated rigid detection base plate. The base plate is made of Invar alloy or aerospace-grade aluminum alloy, which has a low coefficient of thermal expansion and high rigidity. The fixing method is that the second beam splitter (BS2), the first camera mount, and the second camera mount are all directly and rigidly fixed to the same base plate by high-strength screws or UV-curable adhesive (UV Glue), forming an independent compact detection module.
[0045] During the initial assembly, a micrometer-level differential micrometer is installed under the second camera mounting base for initial 0.5-pixel alignment adjustment. After adjustment, the camera position is completely locked using locking screws, and the micrometer is removed. When the system is working, small external vibrations such as fan vibrations and ground vibrations usually cause relative displacement of the optical components. Since BS2 and the two cameras are integrated on the same rigid base plate, they will undergo common-mode displacement as a whole with vibration. Although the detection module as a whole may move, the relative position between the two cameras remains absolutely stationary.
[0046] This design greatly improves the system's anti-interference capability, ensuring that the 0.5-pixel misalignment accuracy will not drift during long-term experimental or clinical use, eliminating the need for frequent recalibration, significantly improving the engineering level of the equipment, or using epoxy resin to pot the moving parts.
[0047] When the system is working, minor external vibrations such as fan vibrations and ground vibrations usually cause relative displacement of the optical components. Since BS2 and the two cameras are integrated on the same rigid base plate, they will undergo common-mode displacement as a whole with the vibration. Although the detection module as a whole may be moving, the relative position between the two cameras remains absolutely stationary.
[0048] This design greatly improves the system's anti-interference capability, ensuring that the 0.5-pixel misalignment accuracy will not drift during long-term experimental or clinical use, eliminating the need for frequent recalibration and significantly enhancing the engineering level of the device.
[0049] System Implementation Example 4 According to an embodiment of the present invention, a dual-camera misaligned sampling imaging OCT system is provided. Based on the third embodiment of the system, an image processing method is further adopted to perform high-resolution image reconstruction and improve the prediction effect. The system controls two cameras to expose simultaneously, acquiring the first image signal respectively. Second image signal ,in This indicates the camera's pixel size; the sample can be in motion or stationary at this time.
[0050] In this embodiment of the invention, the images are preprocessed and registered, specifically by performing dark noise subtraction and flat field correction on the two images respectively; then, feature point matching algorithms such as SIFT or ORB are used to calculate the actual subpixel displacement of the two images, and to verify whether the displacement is maintained within the preset 0.5 pixel accuracy range. If there is a slight deviation, the deviation vector is recorded; Using phase-shifting algorithms of FF-OCT, such as Hilbert transform or multi-step phase-shifting method, the reflectance amplitude map of the sample is demodulated from the interference fringes. Since the two cameras capture the interference state at the same moment, there is no phase drift in time. If sparse images are not synthesized, two demodulated low-resolution amplitude images can be input into a pre-trained deep neural network model. The network uses a large amount of high-resolution slice data of biological tissues as a training set to learn prior knowledge of tissue texture. It not only performs simple pixel interleaving, but also performs deconvolution and denoising on the image based on the point spread function (PSF) characteristics of the FSI sensor. Finally, it outputs a high-resolution OCT tomographic image with twice the resolution and four times the number of pixels of the original image. The system successfully transforms physical spatial reuse into improved image resolution and effectively suppresses jagged edges caused by traditional interpolation algorithms.
[0051] Method Implementation Examples According to an embodiment of the present invention, a dual-camera misaligned sampling imaging OCT method is provided. Figure 4 This is a flowchart of the dual-camera misaligned sampling imaging OCT method according to an embodiment of the present invention, as follows: Figure 4 As shown, the dual-camera misaligned sampling imaging OCT method according to an embodiment of the present invention specifically includes: In step S410, a light source is provided through a low-coherence light source module; In step S420, the light source is split into beams by the first beam splitter through the interference microscopy optical path module to modulate the information, and the light after reflection and light after scattering are obtained to interfere with each other, and the interference beam is obtained by focusing through the tube lens. In step S430, the interference beam is split by the second beam splitter through the dual-channel detection module, and the first and second image signals are acquired by the dual cameras with spatial sub-pixel misalignment. In step S440, the dual cameras are controlled to acquire data synchronously via the synchronization control module; In step S450, a synchronization sequence including the first image signal and the second image signal is obtained through the data processing module. The synchronization sequence is demodulated and a sparse image is synthesized. The physical imaging is then restored using a deep neural network model based on the sparse image.
[0052] In summary, to address the existing problems, this invention proposes a dual-camera misaligned sampling imaging OCT method. It introduces a second beam splitter and a dual-camera architecture, utilizing spatial multiplexing instead of temporal scanning to achieve fully synchronous acquisition of two sub-pixel misaligned images. This method acquires twice the spatial sampling information without adding any extra time overhead, completely eliminating the time lag problem in traditional pixel-shifting techniques and effectively suppressing motion artifacts caused by sample movement, making it suitable for high-speed, high-resolution imaging of biological tissues. The static beam splitter and dual cameras fix the optical path, eliminating mechanical moving parts used for lateral scanning, avoiding the impact of mechanical wear and vibration on the stability of interference fringes, while also reducing system control complexity and improving the long-term stability and durability of the optical system. Applications: The dual cameras spatially misalign pixels to form a plum blossom-shaped sampling array, doubling the system's spatial sampling frequency; after physical demodulation of the two synchronously acquired signals and reconstruction using a deep neural network model, it can learn prior knowledge of biological tissue texture, more accurately predict grayscale changes between subpixels, effectively eliminating the jagged edges and blurring that may be caused by traditional interpolation. This results in a high-resolution image that retains OCT tomography capabilities while possessing higher image sharpness and signal-to-noise ratio, effectively restoring high-frequency details lost due to pixel averaging effects; the overall solution balances a large field of view with a high signal-to-noise ratio, allowing the use of general-purpose industrial cameras with large pixels and deep wells, achieving high resolution at low cost through dual-camera technology.
[0053] Electronic device examples Figure 5 This is a schematic diagram of an electronic device according to an embodiment of the present invention. The electronic device 500 may include at least one processor 510 and a memory 520. The processor 510 can execute instructions stored in the memory 520. The processor 510 is communicatively connected to the memory 520 via a data bus. In addition to the memory 520, the processor 510 can also be communicatively connected to an input device 530, an output device 540, and a communication device 550 via the data bus.
[0054] Processor 510 can be any conventional processor, such as a commercially available CPU. Processors may also include graphics processing units (GPUs), field-programmable gate arrays (FPGAs), systems-on-chips (SoCs), application-specific integrated circuits (ASICs), or combinations thereof.
[0055] The memory 520 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.
[0056] In this embodiment of the present disclosure, the memory 520 stores executable instructions, and the processor 510 can read the executable instructions from the memory 520 and execute the instructions to implement all or part of the steps of the dual-camera misaligned sampling imaging OCT method in any of the exemplary embodiments described above.
[0057] Computer-readable storage medium embodiments In addition to the methods and systems described above, exemplary embodiments of this disclosure may also be a computer program product or a computer-readable storage medium storing the computer program product, the computer product including computer program instructions that can be executed by a processor to implement all or part of the steps described in any of the dual-camera misaligned sampling imaging OCT methods in the exemplary embodiments described above.
[0058] Computer program products can be written in any combination of one or more programming languages to perform the operations of the embodiments of this application. Programming languages include object-oriented programming languages such as Java and C++, as well as conventional procedural programming languages such as C or similar languages, and scripting languages (e.g., Python). The program code can be executed entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.
[0059] Computer-readable storage media may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example,, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media include: static random access memory (SRAM) having one or more electrically connected wires; electrically erasable programmable read-only memory (EEPROM); erasable programmable read-only memory (EPROM); programmable read-only memory (PROM); read-only memory (ROM); magnetic storage; flash memory; magnetic disk or optical disk; or any suitable combination thereof.
[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A dual-camera misaligned sampling imaging OCT system, characterized in that, include: Low-coherence light source module, used to provide light source; The interference microscopy optical path module is used to split the light source into beams by the first beam splitter and modulate the information to obtain the interference between the reflected light and the scattered light, and then converge the interference beam through the tube lens. The dual-channel detection module is used to split the interference beam through a second beam splitter and acquire the first and second image signals respectively through dual cameras with spatial sub-pixel misalignment. A synchronization control module is used to control the synchronous acquisition of data by the dual cameras; The data processing module is used to acquire a synchronization sequence including the first image signal and the second image signal, demodulate the synchronization sequence and synthesize a sparse image, and use a deep neural network model to reconstruct the physical imaging based on the sparse image.
2. The system according to claim 1, characterized in that, The interference microscopy optical path module specifically includes: a first beam splitter, an objective lens, a reference arm including a reference mirror and piezoelectric ceramic, a sample arm, and a tube lens; The first beam splitter is connected to the low coherence light source module, the objective lens is placed between the first beam splitter and the sample arm, and between the first beam splitter and the reference arm, and the tube lens is in the interference light output path of the first beam splitter.
3. The system according to claim 1, characterized in that, The interference microscopy optical path module is specifically used for: The first beam splitter receives the light source transmitted by the low-coherence light source module and splits the light source into reference light and sample light. The reference light is reflected by a reference mirror, and the sample light is backscattered by a sample arm for information modulation. The light after reflection and the light after scattering interfere with each other and are then converged by a tube lens to obtain an interference beam.
4. The system according to claim 1, characterized in that, The dual-channel detection module is specifically used for: The second beam splitter receives the interference beam output from the interference microscope optical path module and equally splits the interference beam into a transmitted beam and a reflected beam. The second beam splitter is located in the exit optical path of the tube lens. A first image signal is acquired based on the transmitted beam by a first camera positioned in the transmission optical path of the second beam splitter; a second image signal is acquired based on the reflected beam by a second camera positioned in the reflection optical path of the second beam splitter. The first camera and the second camera constitute the dual camera system. The first camera and the second camera are conjugate in the optical axis direction. On a plane perpendicular to the optical axis, the second camera has a spatial subpixel misalignment relative to the first camera.
5. The system according to claim 1, characterized in that, The data processing module specifically includes: The synchronous acquisition module is used to control the piezoelectric ceramic to perform step phase shift. At each phase shift step, the synchronous control module triggers the simultaneous exposure of the two cameras. The first camera acquires the first image signal, and the second camera acquires the second image signal to form a synchronous sequence. The demodulation module is used to perform OCT demodulation calculations based on the synchronization sequence to obtain a first image result and a second image result; A sparse sampling module is used to establish a virtual network, fill the virtual network with the first image result and the second image result, and synthesize a sparse image; A deep learning module is used to reconstruct physical images based on the sparse images using a pre-trained deep neural network model that includes encoder and decoder structures.
6. The system according to claim 5, characterized in that, The sparse sampling module is specifically used for: Establish a virtual network with twice the resolution of a single camera; The pixel values of the first image result are filled into the even-numbered row and even-numbered column coordinates of the virtual network, and the pixel values of the second image result are filled into the odd-numbered row and odd-numbered column coordinates of the virtual network to synthesize the sparse image.
7. The system according to claim 5, characterized in that, The deep learning module is specifically used for: The sparse image is input into a deep neural network model, and deep feature representations are extracted through the convolutional layer in the encoder. Based on the deep feature representations, the missing pixel values of the sparse image are predicted by the decoder, and the physical image is reconstructed.
8. A dual-camera misaligned sampling imaging OCT method, used in the dual-camera misaligned sampling imaging OCT system according to any one of claims 1 to 7, characterized in that, include: The light source is provided by a low-coherence light source module; The light source is split into beams by the first beam splitter through the interference micro optical path module and modulated with information. The light after reflection and the light after scattering are then interfered with and converged by the tube lens to obtain the interference beam. The interference beam is split by a second beam splitter through a dual-channel detection module, and then acquired by two cameras with spatial sub-pixel misalignment to obtain the first and second image signals. The dual cameras are controlled to acquire data synchronously via a synchronization control module. The data processing module acquires a synchronization sequence including the first image signal and the second image signal, demodulates the synchronization sequence, synthesizes a sparse image, and uses a deep neural network model to reconstruct the physical imaging based on the sparse image.
9. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the dual-camera misaligned sampling imaging OCT method as described in claim 8.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an implementation program for information transmission, which, when executed by a processor, implements the steps of the dual-camera misaligned sampling imaging OCT method as described in claim 8.