Meta-endoscope microscopic system and super-resolution image reconstruction method therefor

Abstract

The present invention provides a meta-endoscope microscopic system and a super-resolution image reconstruction method therefor. The system comprises a light source module, an endoscopic detector, an image sensor, and a meta-zoom lens. The light source module is configured to generate an incident light. The endoscopic detector receives the incident light and projects the incident light onto an object to be tested. The endoscopic detector receives a detection object light formed by said object. The image sensor is configured to receive the detection object light to form a wide-field-of-view image. A meta-zoom lens assembly is arranged between the endoscopic detector and the image sensor, and is configured to adjust the depth of focus of the incident light on the image sensor. A computing module is configured to execute an artificial intelligence model. The computing module receives the wide-field-of-view image and converts the wide-field-of-view image into a super-resolution image by means of the artificial intelligence module. In one embodiment, the present invention further uses the meta-endoscope microscopic system to train an artificial intelligence model for reconstructing the wide-field-of-view image into a super-resolution image.

Description

Super-axicon endoscope microscopy system and super-resolution image reconstruction method thereof TECHNICAL FIELD

[0001] The present application belongs to an endoscope detection technology, in particular to a super-axicon endoscope microscopy system and super-resolution image reconstruction method thereof for reconstructing super-resolution images by artificial intelligence and reconstructing surface topography of a measured object by super-axicon lens zoom. BACKGROUND

[0002] Endomicroscopy has become an important clinical tool for minimally invasive and real-time observation of tissue and subcellular structures in organs in vivo. How to quickly and cost-effectively achieve multi-depth super-resolution endomicroscopy images has always been a key issue in clinical applications. However, the current endomicroscopy technology is still unable to fully meet this demand due to limitations of the performance of optical elements or systems themselves. To solve this challenge, structured illumination microscopy (SIM) has emerged. SIM can be used in combination with traditional fluorescent dyes and fluorescent proteins to provide a practical and efficient solution for endomicroscopy super-resolution images. Although SIM shows excellent capabilities, it still needs to be axially scanned by moving the endoscope probe or the sample to obtain multi-depth images, which reduces the image taking efficiency. In addition, the axial scanning process may cause the risk of damage to the tissue by external force, which constitutes a potential challenge in clinical applications.

[0003] In endomicroscopy, the traditional adjustment of the axial focal length mainly relies on various types of adjustable zoom lenses. However, compound zoom lenses have challenges in miniaturization, while liquid lenses can adjust the focal length through variable curvature, but are easily affected by gravity and cause image quality degradation due to aberrations. In addition, existing zoom technology also changes the magnification during zooming, which is not conducive to three-dimensional image reconstruction.

[0004] In view of the above, there is a need in the art for a super-axicon endoscope microscopy system and super-resolution image reconstruction method to solve the problems of the prior art. SUMMARY

[0005] The main purpose of the present application is to provide an endoscopic microscopy system and its super-resolution image reconstruction method, which can perform multi-depth super-resolution image acquisition of ex vivo samples. This system has several significant advantages, including fixed magnification, wide field of view (FOV) and structured light super-resolution capability (resolution improved by about two times) under multi-depth imaging. In addition, by introducing a deep learning model, the method provided by the present application greatly reduces the image acquisition time and system complexity required for past structured light illumination microscopy (SIM) endoscopic imaging, enabling super-resolution observation of ex vivo brain in a very short time, improving imaging speed, and showing great potential in optical live sectioning and surgical applications.

[0006] The present application studies and proposes a SI technology based on Moiré metalens, which breaks through the diffraction limit and realizes super-resolution endoscopic imaging of ex vivo fluorescent samples. In an embodiment, the present application uses a Moiré metalens composed of two complementary dielectric super-structure pieces, and uses a telecentric optical configuration to achieve uniform magnification, which is crucial for three-dimensional imaging. In addition, to greatly reduce the image acquisition time, the present application introduces a Residual Channel Attention Network (RCAN) deep learning model, which can complete SIM endoscopic imaging with a single image. This method combines the powerful capabilities of deep learning algorithms to obtain super-resolution images in a fast and efficient manner.

[0007] In an embodiment, the present application provides an endoscopic microscopy system, which includes a light source module, an endoscopic detector, an image sensor, a super-zoom lens group and a computing module. The light source module is used to generate an incident light. The endoscopic detector receives the incident light and projects it onto a test object. The endoscopic detector receives a detection object light formed by the test object. The image sensor is used to receive the detection object light and form a wide field of view image. The super-zoom lens group is arranged between the endoscopic detector and the image sensor, and is used to adjust the focus depth of the incident light focused on the image sensor. The computing module is used to execute an artificial intelligence model. The computing module receives the wide field of view image and converts it into a super-resolution image through the artificial intelligence model.

[0008] In an embodiment, the present application further provides a super-resolution image reconstruction method, comprising the following steps: first, providing an endoscopic microscopy system, comprising a light source module, an endoscopic probe, an image sensor, a super-resolution zoom lens, and a computing module. Next, causing the light source module to generate multiple sets of structured light with different phases, respectively, and project them onto a sample. Then, causing the image sensor to receive the structured light generated from the sample to generate multiple corresponding detection images; causing the light source module to generate a detection light and project it onto the sample. Next, causing the image sensor to receive the detection light generated by the sample to generate a wide-field sample image. After that, reconstructing the multiple detection images to form a super-resolution sample image. Then, causing the computing module to take the wide-field sample image as input and the super-resolution sample image as output to perform calculation to establish an artificial intelligence model.

[0009] In an embodiment, the super-resolution image reconstruction method further comprises causing the endoscopic microscopy system to detect a test object to obtain a wide-field image of the test object. Finally, causing the computing module to use the artificial intelligence model to calculate the wide-field image to obtain a super-resolution image of the test object. BRIEF DESCRIPTION OF DRAWINGS

[0010] FIG. 1 is a schematic diagram of an embodiment of the endoscopic microscopy system of the present application.

[0011] FIG. 2 is a calculation schematic diagram of the computing module.

[0012] FIGS. 3A-3B are schematic diagrams of the flow of an embodiment of the super-resolution image reconstruction method of the present application.

[0013] FIG. 4 is a schematic diagram of another embodiment of the endoscopic microscopy system of the present application.

[0014] FIG. 5 is a schematic diagram of an embodiment of the structured light of the present application.

[0015] FIG. 6A is a wide-field sample image.

[0016] FIG. 6B is a super-resolution image corresponding to FIG. 6A.

[0017] FIG. 7A is a wide-field image of a test object.

[0018] FIG. 7B is a super-resolution image of the test object.

[0019] FIG. 7C is a schematic diagram of taking images at different depths using FIG. 3C.

[0020] FIG. 7D is a detailed feature map of the reconstructed microvessels and glymphatic system.

[0021] FIG. 7E shows the absolute error map of the ex vivo image of the mouse brain in FIG. 7C.

[0022] FIG. 7F shows a spectrum comparison analysis diagram.

[0023] Reference signs: 2, 2a - super-oviscope microscopic system; 20 - light source module; 200 - light source; 201 - spatial light modulation device; 21 - endoscope probe; 210 - micro objective lens; 211 - relay lens group; 22 - image sensor; 23 - super-zoom lens group; 230 - first super-zoom lens; 231 - second super-zoom lens; 2300 - 24 - operation module; 90 - incident light; 90a - structured light; 900 ~ 902 - structured light pattern; 900a ~ 900c, 901a ~ 901c, 902a ~ 902c - structured light pattern; 91 - object light; 91 - structured object light; 92 - object light; 3 - method; 30 ~ 39c - step. DETAILED DESCRIPTION

[0024] The various illustrative embodiments can be described below in connection with the accompanying drawings, some of which show one or more exemplary embodiments. However, the inventive concept can be embodied in many different forms and should not be construed as limited to the exemplary embodiments set forth herein. Rather, these exemplary embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the inventive concept to those skilled in the art. Like numbers refer to like elements throughout. The super-oviscope microscopic system and its super-resolution image reconstruction method will be described below in connection with various embodiments in conjunction with the drawings, however, the following embodiments are not intended to limit the present invention.

[0025] Please refer to FIG. 1, which is a schematic diagram of an embodiment of the endoscopic microscopic system of the present application. In this embodiment, the endoscopic microscopic system 2 comprises a light source module 20, an endoscope probe 21, an image sensor 22, an ultrathin zoom lens 23, and a computing module 24. The light source module 20 is used to generate incident light 90. In this embodiment, the light source module 20 is a laser light source module, which is used to generate laser light as the incident light. The endoscope probe 21 receives the incident light 90 and projects it onto the object to be measured OB1. The object to be measured OB1 is a biological cell. The endoscope probe 21 receives the object light 91 formed by the object to be measured OB1 after receiving the incident light 90. In this embodiment, the object to be measured OB1 has a fluorescent dye inside, which can generate fluorescence to form the aforementioned object light 91 under the excitation of laser light. In this embodiment, the endoscope probe 21 comprises a miniature objective 210 and a set of rod doublet relay 211. In one embodiment, the objective 210 adopts a combination of doublet and meniscus lens, and follows the design principle of Cooke triplet lens. The relay lens group 211 contains a field lens and a Hopkins-type relay rod lens group optimized for surface curvature, effectively improving the luminous flux and minimizing the vignetting effect.

[0026] The object light 91 is projected to the image sensor 22 through the dichroic mirror 25. The image sensor 22 receives the object light 91 and forms a wide field image. The ultrathin zoom lens 23 is arranged between the endoscope probe 21 and the image sensor 22, which is used to adjust the depth of focus of the incident light 90 focused on the image sensor 22. In this embodiment, the ultrathin zoom lens 23 is a Moiré metalens, which effectively ensures the consistency of the image magnification by adopting a telecentric optical configuration, thereby realizing three-dimensional imaging. The ultrathin zoom lens 23 is composed of two pieces of first and second ultrathin lenses 230 and 231 made of gallium nitride (GaN) and having ultrathin microstructures 2300 and 2310, which are overlapped and arranged, having a phase wrapping contour of 0 to 2π, and its function is similar to that of a Fresnel lens.

[0027] In the present embodiment, the first super-lens 230 has a rotating structure 232, so that the first super-lens 230 and the second super-lens 231 can rotate relative to each other, thereby changing the depth of focus of the detection light while maintaining the magnification. It should be noted that the rotating structure 232 can be arranged on the first super-lens 230 or the second super-lens 231 or both, which can be determined according to the needs of those skilled in the art with ordinary skill in the art, and is not necessarily limited. In operation, the effective focal length of the Moire super-lens 23 can be adjusted by rotating the first super-lens 230 relative to the second super-lens 231 by an angle θ. The effective focal length of the Moire super-lens 23 can be determined according to the relative rotation angle θ between the two super-lenses, which is shown in the following formula (1):

[0028] where f Moire represents the effective focal length of the Moire super-lens, λ represents the wavelength of the incident light, c represents the spatial frequency coefficient of the Moire fringe (related to the lens phase profile), and θ represents the relative rotation angle (radian) between the two superimposed super-lenses.

[0029] The operation processing module 24 is used to execute an artificial intelligence model. The operation processing module 24 receives the wide field of view image and converts it into a super-resolution image through the artificial intelligence model. The operation processing module 24 is a desktop computer, a notebook computer, a workstation, a cloud server, or a handheld intelligent device such as a smartphone or a tablet computer, but is not limited thereto. The artificial intelligence model in the present embodiment is a residual attention neural network (RCAN), but is not limited thereto. As shown in FIG. 2, the figure is an operation diagram of the operation module. Among them, IM1 represents the wide field of view image output by the image sensor 22, and IM2 represents the super-resolution image obtained by the operation processing module 24 using the artificial intelligence model to process the wide field of view image IM1.

[0030] Referring to FIG. 3A-3B and FIG. 4, FIG. 3A-3B are flow diagrams of an embodiment of the super-resolution image reconstruction method of the present application, and FIG. 4 is a schematic diagram of another embodiment of the endomicroscopy system of the present application. In this embodiment, the method 3 includes a step 30 of providing an endomicroscopy system 2a as shown in FIG. 4. Then, a step 31 is performed to cause the light source module 20 to generate a plurality of sets of structured light with different phases, respectively, and project the structured light to the sample OB2. It is noted that the architecture of FIG. 4 is basically similar to that of FIG. 1, except that the light source module 20 in the architecture of FIG. 4 further includes a light source 200 and a spatial light modulation device 201. The light source 200 is a laser light source to generate incident light 90 to be projected on the spatial light modulation device 201. The spatial light modulation device 201 is a digital micromirror device (DMD) in this embodiment, but is not limited thereto. For example, a liquid crystal on silicon (LCOS) can also be implemented.

[0031] Since the spatial light modulation device 201 has an array of reflecting elements, the incident laser light can be modulated into structured light 90a with a specific pattern by controlling the switching of each reflecting element. In an embodiment, as shown in FIG. 5, which is a schematic diagram of an embodiment of the structured light of the present application, the structured light 90a has three structured light patterns 900-902, each having three structured light patterns 900a-900c, 901a-901c, and 902a-902c with different phases, respectively, and is projected to the sample OB2 via the endomicroscopy probe 21. In this embodiment, the sample OB2 is a biological cell containing a fluorescent dye. When the structured light patterns 900a-900c, 901a-901c, and 902a-902c are projected to the sample OB2, the fluorescent dye is excited to generate corresponding structured light 91a.

[0032] Referring back to FIGS. 3A-3B, next, step 32 is performed, in which the image sensor 22 receives the structure light 91a generated from the sample OB2 to generate a plurality of corresponding detection images. In this embodiment, the structure light 91a generated from the sample OB2 is sensed by the image sensor 22 via the endoscope probe 21 to generate nine images corresponding to the nine structured lights of FIG. 5. The plurality of detection images are output to the operation module 24 for subsequent calculation. Next, step 33 is performed, in which the light source module 20 generates the incident light 90 to project onto the sample OB2. In this step, the spatial light modulation device 201 does not modulate the incident light 90 into structured light, but instead completely reflects the incident light 90 to the endoscope probe 21 and then to the sample OB2, so that the sample OB2 generates the measurement light 91 corresponding to the incident light 90. Then, step 34 is performed, in which the image sensor 22 receives the measurement light 92 generated by the sample OB2 to generate a wide-field sample image. In an embodiment of this step, the measurement light 92 is transmitted to the image sensor 22 via the endoscope probe 21 to generate a wide-field sample image as shown in FIG. 6A. The wide-field sample image is then transmitted to the operation module 24 for storage.

[0033] Then, step 35 is performed, in which the operation module 24 reconstructs the plurality of detection images to form a super-resolution sample image. In step 35, the operation module 24 can reconstruct a super-resolution image corresponding to FIG. 6A as shown in FIG. 6B using existing super-resolution SIM calculation algorithms. Algorithms for reconstructing super-resolution images using a plurality of detection images corresponding to different structured light phases are well known to those of ordinary skill in the art, such as the review article by Chen, X. et al. Superresolution structured illumination microscopy reconstruction algorithms: a review. Light Sci. Appl. 12, 172 (2023), published on April 12, 2023, which will not be described here.

[0034] After step 35, step 36 is performed to train the operation module 24 to establish an artificial intelligence model by taking the wide field sample image obtained in step 34 as input and the super-resolution sample image obtained in step 35 as output. In an embodiment of the present application, the artificial intelligence model is established by using the training and verification data set of RCAN. In this embodiment, the data set covers 13 different types of samples, including mammalian pyloric stomach, lily flower, fibrous cartilage, mature ovarian follicle, testis, heart, pancreas, nerve, cerebellum, cuboidal epithelium, bryophyte neck egg, plant cell and mucus tissue, etc. Each group of image pairs contains a wide field sample image and a corresponding super-resolution sample image, and the model prediction output is adjusted by supervised learning to obtain a well-trained RCAN artificial intelligence model, i.e. the artificial intelligence model shown in FIG. 2.

[0035] It is to be noted that RCAN belongs to a kind of deep convolutional neural network (CNN). In resolution enhancement applications, CNN usually increases the depth of the model to have more trainable parameters for image-to-image conversion. However, traditional CNN treats low-frequency and high-frequency features in the channel equally, which leads to a waste of a large amount of computing resources when the model processes unnecessary low-frequency information, thereby affecting the performance. To solve this problem, the RCAN proposed in the present application introduces a large number of skip connections (including long skip connections and short skip connections) to form a “residual-in-residual” mechanism, which can directly bypass the low-resolution component; at the same time, combined with the “channel attention residual block (RCAB)”, the weight proportion of the features in each channel can be adaptively adjusted.

[0036] After the artificial intelligence model is established, the artificial intelligence model can be used to replace the above-mentioned method of reconstructing a super-resolution image by using multiple structure light detection images. Because the traditional super-resolution image algorithm needs to take a large number of original images (for example, the above-mentioned 9 structure light images) to generate a super-resolution image. This will significantly increase the time required to reconstruct the super-resolution image, affect the image acquisition speed, and cause the problem of photobleaching, thereby severely limiting its application in clinical practice. Therefore, by using the method of step 36, it is no longer necessary to reconstruct a super-resolution image by using multiple corresponding structure light detection images, which can greatly reduce the image acquisition time and system complexity required for structure light illumination endoscopy imaging in the past.

[0037] Therefore, after step 36, as shown in FIG. 3B, step 37 is performed to cause the in-vivo hyperfield microscope system 2 to detect the object to obtain a wide-field image of the object. In this step, the actual detection is performed, and at this time, the in-vivo hyperfield microscope system 2 can use the in-vivo hyperfield microscope system 2 or 2a as described in FIG. 1 or FIG. 4, that is, can directly take an image of the object OB1 without the spatial light modulation device 201, that is, the incident light 90 generated by the light source module 20 is projected onto the object OB1 through the endoscope probe 21, and then the object light generated is sensed by the image sensor 22 through the endoscope probe 21 to obtain a wide-field image as shown in FIG. 7A. After obtaining the wide-field image, the wide-field image is transmitted to the operation module 24. Then, step 38 is performed, and the operation module 24 uses the wide-field image of step 37 as the input of the artificial intelligence model trained in step 36, and then causes the artificial intelligence model to perform operation on the wide-field image to output a super-resolution image of the object, as shown in FIG. 7B.

[0038] Please refer to FIG. 3C, which is a flowchart of an embodiment of the present application for reconstructing the three-dimensional topography of the object. The flowchart of this embodiment is basically similar to that of FIG. 3B, and the difference is that after step 38, step 39a is further performed to determine whether all the focal length imaging is completed, and if not, step 39b is performed to change the focal length by using the hyper-zoom lens group 23 in FIG. 1 or FIG. 4 to focus on the position of the next depth of the object OB2, and then steps 37-38 are repeatedly performed to take an image of the newly adjusted depth to obtain a wide-field image corresponding to the depth, and then a corresponding super-resolution image is reconstructed, and this is repeated until all the imaging depths are completed, and then step 39c is performed to cause the operation module 24 to perform surface topography reconstruction operation on all the super-resolution images to obtain the surface topography of the object OB1.

[0039] As shown in FIG. 7C, it is a schematic diagram of the aforementioned different depth imaging completed by using FIG. 3C. In this embodiment, the super-resolution endoscopic fluorescence images of the mouse brain ex vivo sample are used. In the figure, the wide-field images I wide , the verification super-resolution image (ground truth) I SIM , and the super-resolution image I DL reconstructed by using the artificial intelligence model. In this embodiment, by adjusting the rotation angles of the Moiré hyper-lenses 230 and 231, the in-vivo hyperfield microscope system 2 or 2a of the present application can focus on images at different depths (Δz = 0, 50, 100 μm), and use the well-trained RCAN model to quickly and accurately convert the wide-field image at each depth into the corresponding SIM super-resolution image I DLThe tracer injected into the mouse brain is mainly distributed in the perivascular space, enabling the endoscopic microscopy system of the present application to observe the glymphatic system. Due to the influence of strong background noise and diffraction limit, the microvascular structure and glymphatic system in I wide are difficult to analyze. Compared with the wide-field image of the traditional endoscope, the present application uses the super-resolution image I DL predicted by the RCAN model to successfully break through the diffraction limit and significantly improve the resolution of the endoscopic system.

[0040] In the enlarged image of FIG. 7D, the model successfully improves the image quality at three different depths and can clearly reconstruct the detailed features of the microvessels and glymphatic system. In addition, the PSNR and SSIM quantitative evaluation results of I DL reach an average of 29 dB and 0.87, respectively, which is about 12 dB and 1.5 times higher than I wide . Although the ex vivo fluorescence images of the mouse brain are not included in the training data set, the model still performs well, showing its high generalization ability.

[0041] FIG. 7E shows the absolute error map of the ex vivo images of the mouse brain in FIG. 7C, which are from the test data set. The results show that even if the brain images are not included in the training and validation data sets, the super-resolution image I DL using the artificial intelligence RCAN model of the present application at different focusing depths still has comparable super-resolution capability to the corresponding validation super-resolution image (ground truth) I SIM . The left column of FIG. 7E shows that the wide-field image I wide of the traditional endoscope is limited by the diffraction limit and has obvious out-of-focus background scattering noise, which seriously reduces the resolution and quality of the image. In contrast, the background scattering noise can be effectively suppressed using the artificial intelligence model, and the system resolution is improved, successfully converting the original wide-field input into a super-resolution predicted image.

[0042] In addition, to verify the effect of RCAN in extending the frequency domain information, FIG. 7F shows the spectral comparison analysis diagram. The comparison results show that the field of view (FOV) in the frequency domain of the endoscopic system can be significantly expanded using the artificial intelligence model of the present application. Compared with the wide-field image I wide , the frequency domain FOV radius (about 100 pm) of the super-resolution image I DL is almost twice that of the wide-field image I wideThe two times magnification of the image (about 50 pm). This analysis proves that the artificial intelligence model of the present application effectively redistributes the frequency domain information, greatly improving the observation range of the endoscope system in the frequency space. This method breaks through the diffraction limit, enabling the system to capture higher frequency image features, thereby achieving super-resolution imaging capability.

[0043] The above merely describes the preferred embodiments or examples of the technical means adopted by the present application to solve the problems, and is not intended to limit the scope of the present application. That is, any equivalent changes and modifications made in accordance with the meaning of the claims of the present application are all covered by the protection scope of the present application.

Claims

1. An ultravision endomicroscopy system, comprising: The system comprises: a light source module for generating an incident light; an endoscope probe for receiving the incident light and projecting it onto a sample, the endoscope probe receiving a detection light formed by the sample; an image sensor for receiving the detection light and forming a wide-field image; a super-apochromatic zoom lens group arranged between the endoscope probe and the image sensor for adjusting a depth of focus of the incident light on the image sensor; and a computing module for executing an artificial intelligence model, the computing module receiving the wide-field image and converting it into a super-resolution image via the artificial intelligence model.

2. The endomicroscopy system of claim 1, wherein the optical fiber is a single mode optical fiber. The super-apochromatic zoom lens group further comprises: a first super-apochromatic lens having a plurality of first super-apochromatic microstructures; a second super-apochromatic lens arranged on one side of the first super-apochromatic lens, the second super-apochromatic lens having a plurality of second super-apochromatic microstructures corresponding to the plurality of first super-apochromatic microstructures; and a rotating mechanism coupled to the first super-apochromatic lens or the second super-apochromatic lens to cause relative rotation between the first super-apochromatic lens and the second super-apochromatic lens, thereby changing the depth of focus of the detection light while maintaining the magnification. The light source module is a laser light source module, and the detection light is laser light used to excite the sample to generate fluorescent light.

3. The endomicroscopy system of claim 1, wherein the optical fiber is a single mode optical fiber. The super-apochromatic zoom lens group changes different depths of focus, causing the image sensor to generate a plurality of wide-field images corresponding to different depths of focus, and the computing module reconstructs a three-dimensional topography of the sample from the plurality of wide-field images corresponding to different depths of focus.

4. The endomicroscopy system of claim 1, wherein the optical fiber is a single mode optical fiber. The system comprises:

5. A super-resolution image reconstruction method, characterized by, providing a super-apochromatic endoscope microscopy system comprising a light source module, an endoscope probe, an image sensor, a super-apochromatic zoom lens group, and a computing module; causing the light source module to generate a plurality of structured lights having different phases and project them onto a sample; causing the image sensor to receive the structured lights generated from the sample to generate a plurality of detection images; causing the light source module to generate a detection light and project it onto the sample; causing the image sensor to receive the detection light generated from the sample to generate a wide-field sample image; reconstructing the plurality of detection images to form a super-resolution sample image; and causing the computing module to execute an artificial intelligence model, with the wide-field sample image as input and the super-resolution sample image as output. The super-apochromatic zoom lens group further comprises:

6. The super-resolution image reconstruction method of claim 5, wherein, a first super-apochromatic lens having a plurality of first super-apochromatic microstructures; a second super-apochromatic lens arranged on one side of the first super-apochromatic lens, the second super-apochromatic lens having a plurality of second super-apochromatic microstructures corresponding to the plurality of first super-apochromatic microstructures; and a rotating mechanism coupled to the first super-apochromatic lens or the second super-apochromatic lens to cause relative rotation between the first super-apochromatic lens and the second super-apochromatic lens, thereby changing the depth of focus of the detection light while maintaining the magnification. The detection light is laser light used to excite the sample to generate fluorescent light.

7. The super-resolution image reconstruction method of claim 5, wherein, The system further comprises:

8. The super-resolution image reconstruction method of claim 5, wherein, causing the super-apochromatic endoscope microscopy system to detect a sample to obtain a wide-field image of the sample; and causing the computing module to execute an artificial intelligence model on the wide-field image to obtain a super-resolution image of the sample. ​ 9. The super-resolution image reconstruction method of claim 8, wherein, The super-telephoto zoom lens changes different focusing depths, so that the image sensor generates multiple wide-view images corresponding to different focusing depths. The operation module receives the multiple wide-view images corresponding to different focusing depths to reconstruct the three-dimensional topography of the object to be measured.

Images