Near-infrared band multi-view three-dimensional fluorescence imaging device

Abstract

According to the near-infrared band multi-view three-dimensional fluorescence imaging device and method provided by the invention, on one hand, the excitation cavity on the inner wall of the plane mirror is combined with the divergent laser light source, a uniform excitation light field is formed through multiple reflections, and uniform excitation of medium and large samples is realized by using good penetrability of the near-infrared band; on the other hand, a multi-view fluorescence image is obtained through one-time imaging by utilizing the inner wall of the plane mirror of the excitation cavity and is used for three-dimensional fluorescence image reconstruction, and high-speed three-dimensional fluorescence imaging of medium and large samples, especially living animal imaging, is realized by integrating the two aspects. According to the invention, the system has visual angle synchronism: image data of multiple angles of an observed object can be obtained through single imaging, so that the problems of low efficiency and data asynchronization caused by time-sharing and manual body position adjustment are fundamentally avoided, and the system is based on a multi-visual-angle two-dimensional image set which is acquired at a time and consistent in space and time; and a complete data basis is provided for subsequently generating a high-precision three-dimensional model through a stereoscopic vision algorithm or a tomographic reconstruction algorithm.

Description

Technical Field

[0001] This invention belongs to the field of bioimaging technology, and more specifically, relates to a near-infrared multi-view three-dimensional fluorescence imaging device. Background Technology

[0002] Fluorescence imaging is an imaging technique that uses fluorescent probes to specifically label target tissues, cells, or molecules, and then induces fluorescence emission through external excitation light to obtain information about the spatial distribution of the target. Compared to traditional bright-field imaging, which relies on reflected or scattered signals, fluorescence signals have advantages such as high selectivity and good background suppression, and can highlight labeled functional regions. Within a suitable optical window, the scattering and absorption of fluorescent photons in biological tissues are significantly reduced, which is beneficial for obtaining deeper structural and functional information, making it suitable for monitoring and analyzing in vivo processes in living organisms.

[0003] Current fluorescence imaging systems are mostly single-view two-dimensional imaging or microscopic three-dimensional imaging, generally targeting slide samples, ex vivo tissues, or millimeter-scale small model organisms. Traditional fluorescence microscopes, confocal microscopes, and light-sheet microscopes have limited imaging fields of view and working distances, requiring sample fixation, slicing, or transparentization, making it difficult to directly cover live animals measuring several centimeters in size with largely intact overall morphology. Furthermore, existing three-dimensional fluorescence imaging largely relies on mechanical scanning or layer-by-layer tomographic acquisition, with each three-dimensional reconstruction taking a considerable amount of time, failing to meet the real-time, high-speed monitoring needs of rapid movement or physiological processes in live animals.

[0004] In recent years, three-dimensional fluorescence imaging technology based on light field imaging has been proposed. By introducing a microlens array into the imaging channel, angular and spatial information can be acquired simultaneously in a single exposure, enabling three-dimensional fluorescence reconstruction within a certain volume range. This type of light field microscopy technique typically combines high numerical aperture objectives and short working distance optical systems, and is mostly applied to small, relatively transparent samples such as zebrafish embryos and fruit fly larvae. When the sample size expands to centimeter-scale, with highly scattering tissues and overall opacity in live small animals, light field fluorescence imaging struggles to obtain deep tissue information, even in the near-infrared band, and cannot obtain information about the back (bottom) of the sample. Therefore, it is difficult to meet the requirements for high-speed three-dimensional live imaging of large or whole samples.

[0005] Existing technologies also offer several rapid 3D imaging methods for medium to large-sized samples, such as multi-view 3D imaging techniques based on structured light projection, stereo vision, or multi-baseline camera arrays, which are relatively mature in natural scene measurement and industrial inspection. These multi-view 3D imaging techniques are mostly applied to natural imaging scenarios. White light-based multi-view 3D imaging relies on back-reflected light from the sample surface. This technique has stringent requirements on sample surface characteristics; when the sample surface structure is complex, undulating, or rough, it is prone to occlusion and scattering noise, resulting in incomplete optical information or a low signal-to-noise ratio, thus causing the 3D reconstruction algorithm to fail. If these methods are applied to fluorescence imaging systems, the difficulty in uniformly exciting fluorescence detected from each viewpoint further complicates 3D reconstruction.

[0006] Currently, there is a lack of systems capable of rapid three-dimensional fluorescence imaging of medium to large-sized samples. Summary of the Invention

[0007] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a near-infrared multi-view three-dimensional fluorescence imaging device. Its purpose is to utilize an excitation cavity with a plane mirror. On one hand, this, combined with an infrared diverging light source, forms a uniform excitation square within the excitation cavity. On the other hand, by utilizing the multiple reflections of the plane mirror, the camera can obtain real or virtual images from multiple angles in a single imaging operation. For centimeter-sized samples, multi-view reconstruction is used to obtain three-dimensional fluorescence images. This solves the technical problem that current three-dimensional fluorescence imaging systems cannot reconstruct three-dimensional fluorescence images from a single exposure for medium to large-sized samples, thus making them unsuitable for live-body imaging.

[0008] To achieve the above objectives, according to one aspect of the present invention, a near-infrared multi-view three-dimensional fluorescence imaging device is provided, comprising an excitation cavity, an excitation light source, and a camera; The excitation cavity has a wall that is a plane mirror; The excitation source is a near-infrared diverging laser source that emits near-infrared illumination laser light into the excitation cavity, thereby exciting the fluorescent protein of the observed object in the excitation cavity to emit fluorescence. The camera detects the fluorescence band inside the cavity from one side of the excitation cavity, obtaining a real image of the observed object inside the cavity and a virtual image reflected from the cavity wall, which is used to create a multi-view fluorescence image of the observed object.

[0009] Preferably, in the near-infrared multi-view three-dimensional fluorescence imaging device, the excitation cavity is surrounded by multiple plane mirrors, wherein at least two plane mirrors are not parallel.

[0010] Preferably, in the near-infrared multi-view three-dimensional fluorescence imaging device, the excitation cavity is formed by two plane mirrors with an included angle of 20° to 30°, allowing simultaneous acquisition of multi-view fluorescence images covering 180° of the imaging field of view through the virtual image of the plane mirrors; or The excitation cavity is formed by two plane mirrors with an included angle of 70° to 80°. The virtual image of the plane mirror can simultaneously obtain a multi-view fluorescence image covering 360° of the imaging field of view.

[0011] Preferably, in the near-infrared multi-view three-dimensional fluorescence imaging device, the principal optical axis of the excitation light source is not parallel to at least one plane mirror.

[0012] Preferably, the near-infrared multi-view three-dimensional fluorescence imaging device has a pair of non-parallel plane mirrors, wherein at least one of the plane mirrors satisfies the following condition:

[0013] in, The divergence angle of the semi-laser source. Let be the angle between the normal to the plane and the principal optical axis of the laser source.

[0014] Preferably, in the near-infrared multi-view three-dimensional fluorescence imaging device, the excitation light source has a divergence angle of 20° to 26°, preferably 24°.

[0015] Preferably, in the near-infrared multi-view three-dimensional fluorescence imaging device, the laser source is a coherent source.

[0016] Preferably, in the near-infrared multi-view three-dimensional fluorescence imaging device, the laser source has a multimode fiber pigtail, and the illumination laser is emitted from the tail end of the multimode fiber pigtail.

[0017] According to another aspect of the present invention, a method for performing fluorescence three-dimensional imaging using the near-infrared multi-view three-dimensional fluorescence imaging device provided by the present invention is provided, comprising the following steps: (1) The fluorescence signal acquired by the near-infrared multi-view three-dimensional fluorescence imaging device is cropped into a real image of the observed object and virtual images from multiple perspectives; (2) Based on the angle relationship of the plane mirrors in the excitation cavity of the near-infrared multi-view three-dimensional fluorescence imaging device, determine the spatial correspondence between the virtual images and the real images from multiple perspectives, and flip the virtual images that are mirror images of the real images, and convert the real images and the virtual images from multiple perspectives to different projection planes under the same coordinate system as the multi-view fluorescence images of the observed objects. (3) The multi-view fluorescence image of the observed object obtained in step (2) is reconstructed into a three-dimensional fluorescence image of the observed object using a multi-view fusion algorithm.

[0018] Preferably, the fluorescence three-dimensional imaging method involves normalizing the fluorescence intensity of the real image and the virtual image respectively, and then performing three-dimensional reconstruction on the normalized multi-view fluorescence image. Step (3) uses an optical projection tomography algorithm or a three-dimensional reconstruction algorithm based on an implicit neural network to fuse the over-view fluorescence image of the observed object into a three-dimensional fluorescence image of the observed object.

[0019] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects: The near-infrared multi-view three-dimensional fluorescence imaging device provided by this invention utilizes, on the one hand, an excitation cavity within a plane mirror, combined with a diverging laser source, to form a uniform excitation light field through multiple reflections. Leveraging the excellent penetrability of the near-infrared band, uniform excitation of medium to large-sized samples is achieved. On the other hand, the plane mirror within the excitation cavity allows for the acquisition of multi-view fluorescence images in a single imaging operation, which are then used for three-dimensional fluorescence image reconstruction. Combining these two aspects, high-speed three-dimensional fluorescence imaging of medium to large-sized samples is achieved, especially for live animal imaging. This invention features viewpoint synchronization: a single imaging operation can acquire image data from multiple angles of the observed object, fundamentally avoiding the inefficiencies and data asynchrony problems caused by time-sharing and manual adjustments to the body position. Based on a single, spatiotemporally consistent multi-view two-dimensional image set, a complete data foundation is provided for subsequent generation of high-precision three-dimensional models using stereo vision algorithms or tomographic reconstruction algorithms.

[0020] In a preferred embodiment, the present invention also features adjustable viewing angle. By adjusting the angle of the reflector, the imaging coverage angle can be changed from 180 degrees to 360 degrees to adapt to different imaging scenarios: for example, whole-body vascular imaging of mice requires a 360-degree imaging field of view, while observation of vascular vessels in the mouse brain only requires a 180-degree imaging field of view.

[0021] This invention offers enhanced anti-interference and deep imaging capabilities: the application of near-infrared light effectively reduces the effects of specular reflection and internal scattering on the surface of biological tissues. Both the illumination laser and the emitted fluorescence can penetrate tissue samples up to 100 micrometers thick, overcoming the occlusion problem caused by surface undulations. This enables clearer imaging of deep tissues and solves the failure problem of traditional white light imaging on rough surfaces and occluded samples. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the near-infrared multi-view three-dimensional fluorescence imaging device provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the imaging principle of the near-infrared multi-view three-dimensional fluorescence imaging device provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the near-infrared multi-view three-dimensional fluorescence imaging device provided in an embodiment of the present invention; Figure 4 This is an image showing the imaging results of mouse blood vessels using the near-infrared multi-view three-dimensional fluorescence imaging device provided in this embodiment of the invention. Figure 4 'a' represents the original image. Figure 4 b. Processed multi-angle projection image; Figure 5 This is a three-dimensional image of mouse blood vessels reconstructed according to an embodiment of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0024] The near-infrared multi-view three-dimensional fluorescence imaging device provided by the present invention includes an excitation cavity, an excitation light source, and a camera; The excitation cavity has a wall made of plane mirrors; the excitation cavity is surrounded by multiple plane mirrors, of which at least two plane mirrors are not parallel. The non-parallel plane mirrors cleverly act on both the excitation light and the fluorescence light path simultaneously. On the one hand, the non-parallel plane mirrors repeatedly reflect the excitation light, thereby making the intensity of the excitation light in the excitation cavity tend to be uniform. On the other hand, they provide the camera with fluorescence images from multiple perspectives. Combining the above two points, a multi-view fluorescence image with uniform intensity is obtained.

[0025] Typically, the excitation cavity is enclosed by two plane mirrors with an angle between 20° and 30°, which can simultaneously acquire multi-view fluorescence images covering 180° of the imaging field of view through the virtual image of the plane mirrors; the excitation cavity is enclosed by two plane mirrors with an angle between 70° and 80°, which can simultaneously acquire multi-view fluorescence images covering 360° of the imaging field of view through the virtual image of the plane mirrors.

[0026] The excitation source is a near-infrared divergent laser source that emits near-infrared illumination laser light into the excitation cavity, exciting the fluorescent protein of the observed object within the excitation cavity to fluoresce. Generally, fluorescence imaging uses collimated lasers as excitation sources, which are then shaped in the illumination optical path to perform point excitation on the sample. Typical examples include confocal fluorescence microscopy, linear excitation (e.g., linear array confocal fluorescence microscopy), or planar excitation (i.e., light sheet fluorescence microscopy). These are all designed for millimeter-scale applications, requiring concentrated and well-defined excitation light to reduce out-of-focus excitation and improve the signal-to-noise ratio. Collimated laser sources cannot meet the rapid imaging requirements of centimeter-scale three-dimensional fluorescence imaging. This invention uses a near-infrared divergent source as the excitation light. The near-infrared band has excellent biological tissue penetration capabilities, enabling effective deep excitation even for centimeter-scale biological samples, including living samples. The divergent source, the transmitted light from the sample, and the near-infrared light reflected from the sample are repeatedly reflected and superimposed within the excitation cavity, which uses a mirror as its cavity wall, forming a uniform excitation light field to uniformly excite the sample in three dimensions.

[0027] To ensure the excitation light is reflected within the excitation cavity, the principal optical axis of the excitation source must not be parallel to at least one plane mirror. Furthermore, the excitation light undergoes multiple reflections within the excitation cavity wall, satisfying the following condition: among the multiple plane mirrors in the excitation cavity, there exists a pair of non-parallel plane mirrors, with at least one plane mirror satisfying the following condition:

[0028] in, The divergence angle of the semi-laser source. Let be the angle between the normal to the plane and the principal optical axis of the laser source.

[0029] The excitation source is preferably a coherent source, where photon energy is coherently superimposed, thereby effectively exciting the fluorescently stained sample in the excitation cavity.

[0030] Preferably, the excitation source is a multimode fiber pigtail, and the illumination laser is emitted from the tail end of the multimode fiber pigtail. The divergence angle of the laser source is preferably between 20° and 26°, so that the excitation light is concentrated in the excitation cavity and the light energy is uniformly and controllably distributed. A typical value is 24°. The divergence angle refers to the angular difference between the diameter of the laser beam at the beam waist and the diameter of the beam at a certain distance from the beam waist during the propagation process. It describes the degree to which the laser beam spreads outward from the focal point.

[0031] The camera detects the fluorescence band inside the cavity from one side of the excitation cavity, obtaining a real image of the observed object inside the cavity and a virtual image reflected from the cavity wall, which is used to create a multi-view fluorescence image of the observed object.

[0032] Existing multi-view 3D imaging devices based on white light back reflection, when directly applied to fluorescence imaging, present numerous challenges leading to difficulties in reconstructing 3D fluorescence images. Firstly, fluorescence imaging requires the application of a specific wavelength, high energy, and spatially uniform excitation field around the sample to ensure a linear response of the overall fluorescence signal. If traditional white light illumination methods such as point sources, structured light projection, or striped illumination are used, the excitation intensity of different parts of the sample often varies significantly from different viewpoints, easily resulting in local under-excitation or over-excitation, leading to fluorescence signal distortion and errors in intensity quantification. Secondly, multi-view structures typically require multiple cameras and illumination units around the sample, making it extremely difficult to simultaneously achieve high uniformity excitation across multiple paths while meeting safe light dose requirements for biological tissues. Therefore, existing multi-view 3D imaging schemes based on white light illumination and surface reflection are difficult to directly apply to fluorescence imaging technology for rapid 3D fluorescence imaging, failing to meet the need for real-time, deep-penetrating, and high spatiotemporal resolution 3D imaging of the internal structure and functional activities of live small animals.

[0033] This invention combines a near-infrared band, a directional and divergent excitation light source, and an excitation cavity with a planar reflector to form a uniform fluorescence excitation light field, improving the consistency of fluorescence excitation intensity from various viewing angles. It enables the acquisition of multi-view images through a single exposure, achieving Hertz-level and higher speed imaging.

[0034] The method for performing fluorescence three-dimensional imaging using a near-infrared multi-view three-dimensional fluorescence imaging device provided by the present invention includes the following steps: (1) The fluorescence signal acquired by the near-infrared multi-view three-dimensional fluorescence imaging device is cropped into a real image of the observed object and virtual images from multiple perspectives; (2) Based on the angle relationship of the plane mirrors in the excitation cavity of the near-infrared multi-view three-dimensional fluorescence imaging device, determine the spatial correspondence between the virtual images and the real images from multiple perspectives, and flip the virtual images that are mirror images of the real images, and convert the real images and the virtual images from multiple perspectives to different projection planes under the same coordinate system as the multi-view fluorescence images of the observed objects. (3) The multi-view fusion algorithm is used to reconstruct the over-view fluorescence image of the observed object obtained in step (2) into a three-dimensional fluorescence image of the observed object.

[0035] Since the fluorescence intensity of real images and virtual images of different levels is different, in order to eliminate the interference of intensity differences from various viewpoints on 3D reconstruction, the fluorescence intensity of real images and virtual images is normalized separately, and then 3D reconstruction is performed on the normalized multi-view fluorescence images.

[0036] Specifically, step (3) uses an optical projection tomography algorithm or a three-dimensional reconstruction algorithm based on an implicit neural network to fuse the over-view fluorescence image of the observed object into a three-dimensional fluorescence image of the observed object.

[0037] The near-infrared multi-view three-dimensional fluorescence imaging device and method provided by this invention enables rapid three-dimensional fluorescence imaging of biological samples at the centimeter level. It is applicable to live samples, especially small animal samples, such as mice, filling the gap in real-time three-dimensional fluorescence imaging technology for large samples using fluorescence imaging and providing a brand-new means of observing the internal life activities of live animals.

[0038] The following is an example: Example 1 The near-infrared multi-view three-dimensional fluorescence imaging device provided in this embodiment, such as Figure 1 As shown, it includes an excitation cavity, an excitation light source, and a camera; In this embodiment, the excitation cavity is formed by two plane mirrors with an included angle of 30°, and the included angle θ1 between the plane mirrors and the platform is 75°.

[0039] The excitation source is a near-infrared 808nm divergent coherent laser source, emitting light from a multimode fiber pigtail. The illumination laser is emitted from the tail end of the multimode fiber pigtail, with a divergence angle (NA) of 0.2. The relationship between the divergence angle and NA is as follows: The corresponding divergence angle is approximately 24°, and the half-divergence angle in the figure is... Approximately 12° To ensure the excitation light is reflected within the excitation cavity, the principal optical axis of the excitation source must not be parallel to at least one plane mirror. Furthermore, the excitation light undergoes multiple reflections within the excitation cavity wall, satisfying the following condition: among the multiple plane mirrors in the excitation cavity, there exists a pair of non-parallel plane mirrors, with at least one plane mirror satisfying the following condition:

[0040] in, The divergence angle of the semi-laser source. The angle between the normal to the plane and the principal optical axis of the laser source is given in this embodiment. A value of 85° is chosen to ensure that both mirrors are illuminated by the laser. The purpose is to utilize the reflective function of the mirrors to uniformly illuminate the fluorescent sample without blind spots. The distance from the laser source to the focal plane... Between 50cm and 70cm.

[0041] The camera detects the fluorescence band inside the cavity from one side of the excitation cavity, obtaining a real image of the observed object inside the cavity and a virtual image reflected from the cavity wall, which is used to create a multi-view fluorescence image of the observed object.

[0042] In this example, the camera height This refers to the focal length of the camera lens; the camera uses a focal length adjustable lens. After the fluorescence is emitted, it is reflected by two plane mirrors to form multiple virtual images. The angle θ1 between the plane mirrors and the platform is 75°, ensuring that the camera captures one real image and four virtual images: two first-order virtual images that are mirror images of the real image, and two second-order virtual images that are aligned with the real image.

[0043] The camera lens is preferably configured with a magnification of 0.5x to ensure that all four virtual images can be captured by the camera's photosensitive surface, while utilizing as much of the photosensitive surface as possible. The relationship between the camera's photosensitive surface size and the reflector is as follows: Figure 2 As shown. Taking the Hamamatsu camera as an example, the pixel size is 6.5μm, the number of pixels is 2048*2048, and the photosensitive surface size l1 is 13312μm. The cross-sectional dimension of the plane mirror and the sample is approximately l2, which is 26000μm. The base magnification of the camera lens is 0.5x, and the image will not exceed the camera's photosensitive surface, thus meeting the imaging requirements.

[0044] Example 2 The method for performing fluorescence three-dimensional imaging using the near-infrared multi-view three-dimensional fluorescence imaging device provided in Example 1 includes the following steps: (1) The fluorescence signal acquired by the near-infrared multi-view three-dimensional fluorescence imaging device is cropped into a real image of the observed object and virtual images from multiple perspectives. The imaging principle is as follows: Figure 2 As shown; In this embodiment, the multi-view images captured by the camera in a single exposure are mouse images at 0 degrees, 72 degrees, 144 degrees, 216 degrees, and 288 degrees.

[0045] By employing appropriate filtering algorithms, such as Gaussian filtering and median filtering, noise in the image can be removed and image quality improved.

[0046] (2) Based on the angle relationship of the plane mirrors in the excitation cavity of the near-infrared multi-view three-dimensional fluorescence imaging device, determine the spatial correspondence between the virtual images and the real images from multiple perspectives, and flip the virtual images that are mirror images of the real images, and convert the real images and the virtual images from multiple perspectives to different projection planes under the same coordinate system as the multi-view fluorescence images of the observed objects. In this embodiment, as Figure 3 The virtual images 1 and 2 shown are first-order virtual images, which are mirror images of the real images and are horizontally flipped according to the requirements of geometric optics.

[0047] Image registration involves transforming the real image and virtual images from multiple viewpoints to different projection planes within the same coordinate system. Since the images are from different angles, image registration is necessary to determine their spatial correspondence. The SIFT (Scale Invariant Feature Transform) algorithm is used to find feature points in the image, and by matching these feature points, the transformation matrix between the images is calculated, thereby aligning them to the same coordinate system.

[0048] (3) The multi-view fusion algorithm is used to reconstruct the over-view fluorescence image of the observed object obtained in step (2) into a three-dimensional fluorescence image of the observed object.

[0049] Since the fluorescence intensity of real images and virtual images of different levels is different, in order to eliminate the interference of intensity differences from various viewpoints on 3D reconstruction, the fluorescence intensity of real images and virtual images is normalized separately, and then 3D reconstruction is performed on the normalized multi-view fluorescence images.

[0050] This embodiment uses a 3D reconstruction algorithm based on implicit neural networks: Implicit Neural Network (INR) technology reconstructs 3D images. The basic principle is to construct a virtual 3D volume and then project it forward from five angles. The loss function is then calculated by comparing this virtual volume with five 2D images taken from different angles, iteratively approximating the true 3D volume. Specifically: Constructing a virtual 3D volume: A virtual INR 3D space is constructed using a Fourier coding scheme, and the density of the lattice in the 3D space is determined by the 3D resolution of the system; The INR reconstruction network is constructed using a multilayer perceptron (MLP) structure with a rectified linear unit (ReLU) activation function. First, the 3D coordinates are flattened into one-dimensional vectors, transforming the spatial structure into a format suitable for the MLP architecture. Next, these 3D coordinates are Fourier-encoded, and then the encoded coordinates are input into the INR reconstruction network. The INR reconstruction network randomly predicts the initial intensity value at each coordinate.

[0051] Constructing a Constraint Network: To constrain the 3D estimated volume obtained by the INR reconstruction network using 2D images, we generate a PSF that matches the optical system parameters. This PSF is then used to project the 3D estimated volume, resulting in multi-angle 2D images. Subsequently, the estimated images are compared with images captured by the camera, and a loss function is calculated. This loss function is used to constrain the output of the INR reconstruction network.

[0052] The reconstructed 3D data is typically a volumetric data field, requiring volume rendering techniques for visualization. Appropriate visualization software is then used to display the reconstructed 3D image. The quality of the reconstructed 3D image is assessed, including calculating metrics such as Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM), and comparing it with the original image or a known real structure to evaluate the accuracy and quality of the reconstruction results. Appropriate object registration optimization is then performed to further improve the quality of the reconstructed image.

[0053] In this embodiment, mice were used as experimental subjects. The device of the present invention was constructed to conduct whole-mouse in vivo imaging experiments, and the three-dimensional imaging results were analyzed.

[0054] Ten-day-old mice were anesthetized with isoflurane during surgery and then fixed to a targeting device after complete sedation. Near-infrared probes were injected into the mice to label blood vessels throughout the body. The exposure time was set to 200 ms. The sample was excited by illumination light, and after reflection by a plane mirror, four virtual images were formed. These images, along with the real image of the object, were captured by the camera. (See details...) Figure 4 a. Align them to the same coordinate system and normalize them to obtain multi-angle projections, such as Figure 4 As shown in b.

[0055] Image reconstruction is achieved using 3D reconstruction algorithms. The resulting 3D image, obtained through multiple iterations, is shown below. Figure 5 As shown, where Figure 5 'a' represents the 3D image obtained using Algorithm 1. Figure 5 b is the 3D image obtained using Algorithm 2. Figure 5 c represents a 3D image reconstructed using multi-angle photography (OPT technology). Due to angle alignment and delay issues in traditional multi-angle photography, the 3D reconstruction is difficult to converge and the results are unsatisfactory.

[0056] Using a near-infrared light field microscope, we recovered structures that were out of focus and blurred under visible light wide-field or light field microscopy. High spatial resolution allows us to observe micron-scale brain blood vessels, while the large depth of field enables simultaneous observation of three-dimensional structures up to 150 µm in size.

[0057] Compared to the camera's perspective, the reconstructed 3D volume can achieve an average SSIM of 96.3% and an average PSNR of 18 in the corresponding perspective, effectively displaying fine structures such as blood vessels.

[0058] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A near-infrared multi-view three-dimensional fluorescence imaging device, characterized in that, This includes the excitation chamber, the excitation source, and the camera; The excitation cavity has a wall that is a plane mirror; The excitation source is a near-infrared diverging laser source that emits near-infrared illumination laser light into the excitation cavity, thereby exciting the fluorescent protein of the observed object in the excitation cavity to emit fluorescence. The camera detects the fluorescence band inside the cavity from one side of the excitation cavity, obtaining a real image of the observed object inside the cavity and a virtual image reflected from the cavity wall, which is used to create a multi-view fluorescence image of the observed object.

2. The near-infrared multi-view three-dimensional fluorescence imaging device as described in claim 1, characterized in that, The excitation cavity is surrounded by multiple plane mirrors, of which at least two plane mirrors are not parallel.

3. The near-infrared multi-view three-dimensional fluorescence imaging device as described in claim 2, characterized in that, The excitation cavity is formed by two plane mirrors with an included angle of 20° to 30°, allowing for the simultaneous acquisition of multi-view fluorescence images covering a 180° imaging field through the virtual images of the plane mirrors; or The excitation cavity is formed by two plane mirrors with an included angle of 70° to 80°. The virtual image of the plane mirror can simultaneously obtain a multi-view fluorescence image covering 360° of the imaging field of view.

4. The near-infrared multi-view three-dimensional fluorescence imaging device as described in claim 2, characterized in that, The principal optical axis of the excitation light source is not parallel to at least one plane mirror.

5. The near-infrared multi-view three-dimensional fluorescence imaging device as described in claim 4, characterized in that, There exists a pair of non-parallel plane mirrors, where at least one plane mirror satisfies the following condition: in, The divergence angle of the semi-laser source. Let be the angle between the normal to the plane and the principal optical axis of the laser source.

6. The near-infrared multi-view three-dimensional fluorescence imaging device as described in claim 2, characterized in that, The divergence angle of the excitation light source is 20° to 26°, preferably 24°.

7. The near-infrared multi-view three-dimensional fluorescence imaging device as described in claim 6, characterized in that, The laser source is a coherent source.

8. The near-infrared multi-view three-dimensional fluorescence imaging device as described in claim 6, characterized in that, The laser source has a multimode fiber pigtail, and the illumination laser is emitted from the tail end of the multimode fiber pigtail.

9. A method for performing fluorescence three-dimensional imaging using the near-infrared multi-view three-dimensional fluorescence imaging device as described in any one of claims 1 to 8, characterized in that, Includes the following steps: (1) The fluorescence signal acquired by the near-infrared multi-view three-dimensional fluorescence imaging device is cropped into a real image of the observed object and virtual images from multiple perspectives; (2) Based on the angle relationship of the plane mirrors in the excitation cavity of the near-infrared multi-view three-dimensional fluorescence imaging device, determine the spatial correspondence between the virtual images and the real images from multiple perspectives, and flip the virtual images that are mirror images of the real images, and convert the real images and the virtual images from multiple perspectives to different projection planes under the same coordinate system as the multi-view fluorescence images of the observed objects. (3) The multi-view fluorescence image of the observed object obtained in step (2) is reconstructed into a three-dimensional fluorescence image of the observed object using a multi-view fusion algorithm.

10. The fluorescence three-dimensional imaging method as described in claim 9, characterized in that, After normalizing the fluorescence intensity of the real and virtual images respectively, the normalized multi-view fluorescence images are then reconstructed in three dimensions. Step (3) uses an optical projection tomography algorithm or a three-dimensional reconstruction algorithm based on an implicit neural network to fuse the over-view fluorescence image of the observed object into a three-dimensional fluorescence image of the observed object.

Images