Image optimization reconstruction method and system based on photon number ratio after logarithmic transformation

Abstract

The invention relates to the technical field of microscopic imaging reconstruction, and particularly discloses an image optimization reconstruction method and system based on a photon number ratio after logarithmic transformation, and the method comprises the steps: building a photon statistical mapping relation between a simulation super-resolution image and a simulation real image through the real experiment background photon number constraint, and determining an RLP interval; meanwhile, the photon number ratio in the real super-resolution image and the real original image under different fluorescence wavelengths is calculated, an RLP-modeling wavelength response curve is constructed, and the fluorescence wavelength corresponding to the point, located in the RLP interval, on the curve serves as the modeling wavelength; and then, a measurement matrix is constructed based on the modeling wavelength, and super-resolution reconstruction is carried out on the to-be-reconstructed original image based on the measurement matrix. According to the method, the spatial resolution and the structural fidelity of the reconstructed image can be remarkably improved on the premise that the system complexity is not increased.

Description

Technical Field

[0001] This invention relates to the field of microscopic imaging reconstruction technology, and more specifically to an image optimization reconstruction method and system based on the photon number ratio after logarithmic transformation. Background Technology

[0002] With the rapid development of fields such as nanobiomedicine, precision manufacturing, and micro / nano detection, higher demands are being placed on the spatial resolution of optical microscopy imaging systems. Traditional far-field optical microscopes are limited by the diffraction limit, with a lateral resolution typically only 200 nm, which is insufficient to meet the observation needs of subcellular structures, viral particles, and molecular-level devices. Therefore, super-resolution microscopy has emerged, among which stochastic optical reconstruction microscopy (STORM), as an important branch of single-molecule localization microscopy, has become a crucial tool in the life sciences and micro / nano detection fields with its imaging resolution of approximately 20 nm.

[0003] However, the imaging quality of STORM is highly dependent on the focusing accuracy of the microscope system. Although fluorescent microbeads are widely used as reference markers for autofocus, slight defocusing is still difficult to completely avoid during actual imaging due to system drift or sample undulations. This leads to a shift in the point spread function (PSF), resulting in problems such as super-resolution image reconstruction distortion, structural blurring, and decreased positioning accuracy. Under slight defocus conditions, the PSF can still be approximated as a Gaussian function, but its modeling wavelength has deviated from the emission peak wavelength of the fluorescent dye. Traditional methods for selecting the PSF based on the nominal wavelength are difficult to adapt to changes in actual imaging conditions. Therefore, how to achieve algorithm-level optimized reconstruction of defocused images without additional hardware focusing has become a key technical challenge that urgently needs to be solved in the engineering application of current STORM systems. Summary of the Invention

[0004] In view of the above problems, the present invention proposes an image optimization and reconstruction method and system based on the photon number ratio after logarithmic transformation, so as to overcome the above problems or at least partially solve the above problems.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides an image optimization and reconstruction method based on the photon number ratio after logarithmic transformation, comprising the following steps: Acquire simulated super-resolution images, simulated original images, and real original images; With the real original image background as a constraint, the ratio of photons in the simulated super-resolution image and the simulated original image at different magnification steps is iteratively calculated according to a preset magnification step size, and is used as the first photon number ratio. Plot the first curve with the magnification rate on the x-axis and the first photon number ratio on the y-axis, and determine the range of the first photon number ratio. According to the preset fluorescence wavelength step size, the real super-resolution image corresponding to the real original image under different fluorescence wavelengths is reconstructed iteratively, and the photon number ratio in the real super-resolution image and the real original image under different fluorescence wavelengths is calculated as the second photon number ratio. Plot a second curve with fluorescence wavelength on the x-axis and the second photon number ratio on the y-axis. Determine the fluorescence wavelength corresponding to the point on the second curve that is in the range of the first photon number ratio, and use it as the modeling wavelength. Construct a measurement matrix based on the modeling wavelength; Super-resolution reconstruction is performed on the original image to be reconstructed based on the measurement matrix.

[0006] Furthermore, the process of acquiring the simulated super-resolution image, the simulated original image, and the real original image includes: Multiple real original image frames were obtained based on a pre-built real experimental platform. The real original image was obtained by superimposing the real original image frames. Multiple simulated original image frames are obtained based on a pre-built simulation experimental platform. The simulated original image is obtained by superimposing the simulated original image frames. Reconstruct all the original simulated image frames to obtain the corresponding simulated super-resolution image frames. Then, superimpose the simulated super-resolution image frames to obtain the simulated super-resolution image. The experimental parameters of the simulation experimental platform and the real experimental platform are consistent.

[0007] Furthermore, in the process of plotting the first curve, each iteration includes: Determine the multiplier for the current iteration stage according to the preset multiplier step size. ,in, This represents the multiplier used in the j-th iteration. This represents the multiplier used in the (j-1)th iteration. This is the preset multiplier step size; The background of the real original image is added to the simulated original image. After adding the background, the simulated original image is magnified and logarithmically calculated. The specific process is as follows:

[0008] in, This represents the original simulation image after the j-th iteration; This represents the original simulation image after the (j-1)th iteration; Represents the background of the original image; The process of performing magnification and logarithmic calculation on the simulated super-resolution image is as follows:

[0009] in, This represents the simulated super-resolution image after the j-th iteration; This represents the simulated super-resolution image after the (j-1)th iteration; Calculate the photon number ratio between the simulated super-resolution image and the simulated original image in the current iteration:

[0010] in, Represents the simulated super-resolution image after the j-th iteration. The total number of photons in the sample; Represents the original simulation image after the j-th iteration. The total number of photons in the sample; The first photon number ratio represents the ratio of the number of photons in the simulated super-resolution image and the simulated original image after the j-th iteration, when the magnification ratio i is the horizontal axis.

[0011] Furthermore, the average photon count of a blank area at a preset distance from the sample structure in the real original image is used as the background and added to the simulated original image.

[0012] Furthermore, the lower limit of the first photon number ratio range is the minimum photon number ratio in the first curve, and the upper limit is the maximum photon number ratio in the first curve.

[0013] Furthermore, in the process of plotting the second curve, each iteration includes: The fluorescence wavelength for the current iteration stage is determined according to the preset fluorescence wavelength step size. , ,in, This indicates the fluorescence wavelength used in the m-th iteration. This indicates the fluorescence wavelength used in the (m-1)th iteration. The preset fluorescence wavelength step size; Based on the fluorescence wavelength at the current iteration stage The corresponding measurement matrix is ​​used to reconstruct all real original image frames and obtain the corresponding real super-resolution image frames respectively. All real super-resolution image frames are superimposed to obtain a real super-resolution image; The logarithm is calculated for both the original image and the super-resolution image. The specific calculation process is as follows:

[0014]

[0015] in, This represents the true super-resolution image after the m-th iteration; This represents the true super-resolution image after the (m-1)th iteration; This represents the actual original image after the m-th iteration; This represents the true original image after the (m-1)th iteration; Calculate the ratio of photon counts in the real super-resolution image and the real original image in the current iteration:

[0016] in, This represents the photon number ratio after the m-th iteration when the fluorescence wavelength λ is the abscissa, i.e., the second photon number ratio. Represents the true super-resolution image after the m-th iteration. The total number of photons in the sample; Represents the true original image after the m-th iteration. The total number of photons in the array.

[0017] Furthermore, the process of constructing the measurement matrix based on the modeling wavelength includes: The pixels of the original image to be reconstructed are subdivided into a dense grid, and each grid point is regarded as a potential monomolecular location; For any grid point, a point spread function (PSF) is constructed based on the modeling wavelength. The original image corresponding to a single molecule of unit intensity at that grid point is generated based on the PSF. The original images of the grid point are concatenated column by column to form a vector, which is used as a column of the measurement matrix. By arranging the vectors obtained from all grid points side by side, we obtain the measurement matrix.

[0018] Furthermore, super-resolution reconstruction is performed on the original image to be reconstructed based on the improved compressed sensing model. The improved compressed sensing model is represented as follows:

[0019] in, , which is a weight vector used to adjust the contribution of molecules at different positions; For super-resolution image vectors; For measurement matrix; ε is the original image vector to be reconstructed; ε is the error tolerance parameter, used to control the balance between reconstruction accuracy and robustness. for The j-th element in; Let be the i-th element in x.

[0020] Secondly, the present invention provides an image optimization and reconstruction system based on the photon number ratio after logarithmic transformation, comprising: The image acquisition module is used to acquire simulated super-resolution images, simulated original images, and real original images; The first iteration module is used to iteratively calculate the ratio of photons in the simulated super-resolution image and the simulated original image at different magnifications, with the real original image background as a constraint and according to a preset magnification step size, as the first photon ratio. The photon number ratio determination module is used to plot a first curve with the multiplier on the horizontal axis and the first photon number ratio on the vertical axis, and to determine the range of the first photon number ratio. The second iteration module is used to iteratively reconstruct the real super-resolution image corresponding to the real original image under different fluorescence wavelengths according to the preset fluorescence wavelength step size, and calculate the photon number ratio in the real super-resolution image and the real original image under different fluorescence wavelengths as the second photon number ratio. The modeling wavelength determination module is used to plot a second curve with fluorescence wavelength as the abscissa and the second photon number ratio as the ordinate, and to determine the fluorescence wavelength corresponding to the point on the second curve that is located in the range of the first photon number ratio, which is then used as the modeling wavelength. The measurement matrix construction module is used to construct a measurement matrix based on the modeling wavelength; The reconstruction module is used to perform super-resolution reconstruction of the original image to be reconstructed based on the measurement matrix.

[0021] Thirdly, the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor; when the processor executes the computer program, it implements the steps of the image optimization and reconstruction method based on the photon number ratio after logarithmic transformation as described above.

[0022] As can be seen from the above technical solution, compared with the prior art, the present invention has the following beneficial effects: This invention uses the real image background as a constraint to construct a photon statistical mapping relationship between simulated and real images, establishing the Relative Photon Response (RLP) interval. It transforms the background photon count, which can be accurately measured in real experiments, into a unified cross-batch RLP interval decoupled from single-molecule brightness. Furthermore, using the RLP interval as a criterion, it transforms the difficult-to-measure defocus amount into a stably calculable RLP, establishes an RLP-modeling wavelength response curve, and selects the optimal diffusion function (PSF) model to achieve post-processing algorithm-level defocus compensation. This significantly improves the spatial resolution and structural fidelity of reconstructed images without increasing system complexity, providing a new technical path for the stable operation and quality control of STORM systems. Simultaneously, this invention does not introduce additional photobleaching or drift risks and can be embedded in online processing workflows, providing a zero-cost quality control paradigm for STORM engineering applications such as industrial inspection and high-throughput screening. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or 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 embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0024] Figure 1 This is a flowchart of an image optimization and reconstruction method based on the photon number ratio after logarithmic transformation provided in an embodiment of the present invention; Figure 2 This refers to the single-frame simulated original image provided in this embodiment of the invention; Figure 3 This is a schematic diagram of the superimposed 20 frames of simulated original images provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the superimposed 20 frames of simulated super-resolution images provided in an embodiment of the present invention; Figure 5 This refers to a single-frame real original image provided in the embodiments of the present invention; Figure 6 This is a schematic diagram of 20 real original images superimposed in an embodiment of the present invention; Figure 7 This is a schematic diagram of the relationship curve between RLP and multiplier i provided in an embodiment of the present invention; Figure 8 The super-resolution image obtained by reconstruction based on different fluorescence wavelengths is provided in the embodiments of the present invention; Figure 9 Provided for embodiments of the present invention Figure 8 Enlarged view of the area within the Chinese border; Figure 10 Photon number distribution curves at different fluorescence wavelengths provided in embodiments of the present invention; Figure 11 This is a schematic diagram of the relationship between RLP and fluorescence wavelength λ provided in an embodiment of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] like Figure 1As shown, this invention discloses an image optimization and reconstruction method based on the photon number ratio after logarithmic transformation, comprising the following steps: S1. Obtain the simulated super-resolution image, the simulated original image, and the real original image; S2. Using the real original image background as a constraint, and according to the preset magnification step size, iteratively calculate the photon number ratio in the simulated super-resolution image and the simulated original image at different magnifications, and use it as the first photon number ratio. Plot the first curve with the magnification rate on the x-axis and the first photon number ratio on the y-axis, and determine the range of the first photon number ratio. S3. According to the preset fluorescence wavelength step size, iteratively reconstruct the real super-resolution image corresponding to the real original image under different fluorescence wavelengths, and calculate the photon number ratio between the real super-resolution image and the real original image under different fluorescence wavelengths as the second photon number ratio. Plot a second curve with fluorescence wavelength on the x-axis and the second photon number ratio on the y-axis. Determine the fluorescence wavelength corresponding to the point on the second curve that is in the range of the first photon number ratio, and use it as the modeling wavelength. S4. Construct a measurement matrix based on the modeling wavelength; Super-resolution reconstruction is performed on the original image to be reconstructed based on the measurement matrix.

[0027] The following is a further explanation of each of the above steps.

[0028] S1. Obtain the simulated super-resolution image, the simulated original image, and the real original image. The specific process includes: This invention pre-builds both a real experimental platform and a simulation experimental platform. All experimental data comes from the self-built STORM system, while simulation data is generated strictly in accordance with experimental parameters to achieve closed-loop verification of the experiment-simulation model.

[0029] The imaging platform employed an Olympus IX-71 inverted microscope equipped with a 100× / 1.40 NA oil immersion objective lens and a 3.5× repeater array, using an EMCCD (Andor DU-897U-CV0) with a pixel size of 16 µm. Single-frame exposure time was 30 ms, and the acquisition frame rate was 46.4 Hz. Sample preparation used HeLa cells and the fluorescent dye Alexa-647 with a peak wavelength of 670 nm. Throughout the experiment, an active drift correction device was used to control the Z-axis drift within ±10 nm, ensuring that the defocusing amount originated only from the initial state of the system and eliminating subsequent drift interference.

[0030] The simulation model is fully aligned with the experimental parameters to ensure consistency in PSF, noise, and pixel grid. The super-resolution grid is 1 / 4 the resolution of the camera pixels, corresponding to a measurement matrix size of 196×4096. Fluorescent molecules are randomly activated in each frame, with each molecule emitting 3000 photons.

[0031] Multiple real original image frames are acquired using a pre-built real experimental platform, and multiple simulated original image frames are acquired using a pre-built simulation experimental platform. The real original image frames are then superimposed to obtain the real original image; the simulated original image frames are then superimposed to obtain the simulated original image. In this embodiment, both the real and simulation experiments acquire 20 frames of original images.

[0032] Reconstructing all original simulated image frames yields corresponding simulated super-resolution image frames. Superimposing these simulated super-resolution image frames results in a simulated super-resolution image. Each image frame is a two-dimensional matrix; the superposition process involves accumulating these matrices to obtain a new matrix.

[0033] S2. RLP calibration based on real-world context constraints, specifically including: Initialization process: Simulated super-resolution image S after stacking of each frame true The simulated original image after overlaying each frame (background is 0) R true The actual original image R after superimposing the frames. real ;R real The background bkg, that is, Figure 6 The average number of photons in the area within the red box (a blank area at a preset distance from the sample structure in the original image); magnification. i =0.1.

[0034] Next, multiple iterations are performed, each iteration process including: ① Determine the multiplier of the current iteration stage according to the preset multiplier step size. ,in, This represents the multiplier used in the j-th iteration. This represents the multiplier used in the (j-1)th iteration. In this embodiment of the invention, the preset scaling step size is used. The value is 0.1; ② Add the background of the real original image to the simulated original image. After adding the background, perform magnification processing and logarithmic calculation on the simulated original image. The specific process is as follows:

[0035] in, This represents the original simulation image after the j-th iteration; This represents the original simulation image after the (j-1)th iteration; Represents the background of the original image; ③ Perform magnification processing and logarithmic calculation on the simulated super-resolution image. The specific process is as follows:

[0036] in, This represents the simulated super-resolution image after the j-th iteration; This represents the simulated super-resolution image after the (j-1)th iteration; ④ Calculate the photon count ratio between the simulated super-resolution image and the simulated original image in the current iteration:

[0037] in, Represents the simulated super-resolution image after the j-th iteration. The total number of photons in the sample; Represents the original simulation image after the j-th iteration. The total number of photons in the sample; The first photon number ratio represents the ratio of the number of photons in the simulated super-resolution image and the simulated original image after the j-th iteration, when the magnification ratio i is the horizontal axis.

[0038] After completing the preset number of iterations, at a multiplier... Plot the first curve with the photon number ratio (RLP) as the ordinate and the x-axis as the y-axis. Find the maximum and minimum photon number ratios in the first curve and output the RLP interval, which is the first photon number ratio interval. This interval is [the minimum and maximum RLP values ​​in the first curve].

[0039] In super-resolution fluorescence microscopy, the consistency between simulated and real experimental data at the photon statistics level is crucial for algorithms to move from simulation to reality. Traditional methods typically involve manually setting a fixed number of simulated photons, leading to inconsistencies between simulated and real experimental photon counts and causing systematic deviations in subsequent calibration. This invention proposes using the background photon count measured in real experiments (i.e.,...) Figure 6 Using the average photon count in the red box region as an anchor point, the relative plasmon density (RLP) of the simulated super-resolution image and the original simulated image is determined through an iterative self-consistent method. This iterative process establishes a quantitative mapping relationship decoupled from the single-molecule photon count, providing a unified benchmark for subsequent PSF optimization. RLP is the ratio of the total number of photons in the super-resolution image after each frame is stacked to the total number of photons in the original simulated image after each frame is stacked and magnified. Before magnification and before taking the logarithm, a real background (bkg) is added to the original simulated image after each frame is stacked. Explicitly injecting the experimentally measured bkg into the original simulated image after each frame ensures that the background level of the stacked original simulated image is consistent with the real original image. This is achieved through scanning magnification... This allows for the systematic alteration of the total photon count in the simulated image, thereby enabling the observation of the convergence behavior of RLP as magnification changes, such as... Figure 7 As shown.

[0040] Figure 2 It is a single-frame simulated original image (background is 0). Figure 3 It is a simulated original image after 20 frames are superimposed. Figure 4 It is a simulated super-resolution image composed of 20 superimposed frames. Figure 5 It is a single frame of the original image. Figure 6 It is the original image after 20 frames are superimposed. Although Figure 5 The noise is obvious and the image is coarse, but due to Figure 6 As can be seen, when superimposed to 20 frames, the background is stable and smooth. Figure 6 The region within the red box is relatively far from the cellular microtubule structure, so its average photon count is used as bkg. bkg can serve as a hard constraint for subsequent iterations. Figure 7 RLP and leverage i The relationship curve shows that RLP varies between 3.8 and 5.7. Therefore, the RLP interval is [3.8, 5.7], which is... Figure 7 and Figure 11 The area between the two red lines.

[0041] Figure 7 The global fluctuation range of RLP is between 3.8 and 5.7, with a very small fluctuation range. i The curve enters a plateau region at >20. i When = 0.1, it corresponds to a single-molecule photon emission number of 300; when i When the value is 200, it corresponds to 600,000 single-molecule photon emissions. This range covers the emission range of fluorescent molecules from low to high brightness, exhibiting good experimental versatility. Therefore, this range encompasses almost all real-world experimental scenarios. This range can be used directly as a unified benchmark for subsequent PSF optimization, independent of the total number of photons emitted by fluorescent molecules in simulation experiments to real-world experiments. This method effectively avoids the technical bottleneck of accurately measuring the number of single-molecule photons in real-world experiments, requiring only accurate statistics of the background photon count in real-world experiments for calibration. Under fixed optical parameters, the RLP range is determined only by the system background level and is decoupled from the number of single-molecule photons emitted, thus possessing cross-experimental universality.

[0042] The RLP range output in this step can be embedded in the S3 step below. Through super-resolution reconstruction, the landing point within the RLP range is determined, thereby achieving defocus compensation and providing a simple, reliable, and low-cost solution for STORM's algorithm-level quality control.

[0043] S3. PSF wavelength selection process based on RLP range: The RLP interval established in step S2 provides a crucial basis for the calibration of simulation and real experimental data. Even with slight defocus, the system PSF can still approximate a Gaussian function, but its modeling wavelength has deviated from the nominal dye value. Continuing to use the nominal wavelength for modeling will directly cause positioning drift and structural broadening. Therefore, in step S3, using the real background as an anchor point, the system scans a multi-wavelength PSF library and plots the RLP-λ response curve, as shown... Figure 11 As shown. Based on RLP interval selection of suitable modeling wavelengths and reconstruction results, pure algorithm-level defocus compensation is achieved. Specifically, this includes: Initialization process: The original image R is formed by superimposing all real original image frames. real fluorescence wavelength .

[0044] Next, the iterative process is executed, with each iteration including: ① Determine the fluorescence wavelength for the current iteration stage according to the preset fluorescence wavelength step size. , ,in, This indicates the fluorescence wavelength used in the m-th iteration. This indicates the fluorescence wavelength used in the (m-1)th iteration. In this embodiment, the preset fluorescence wavelength step size is used. The value is 235; ②Based on the fluorescence wavelength at the current iteration stage The corresponding measurement matrix is ​​used to reconstruct all real original image frames and obtain the corresponding real super-resolution image frames respectively. All real super-resolution image frames are superimposed to obtain a real super-resolution image; ③ Perform logarithmic calculations on both the original image and the super-resolution image. The specific calculation process is as follows:

[0045]

[0046] in, This represents the true super-resolution image after the m-th iteration; This represents the true super-resolution image after the (m-1)th iteration; This represents the actual original image after the m-th iteration; This represents the true original image after the (m-1)th iteration; ④ Calculate the ratio of photon counts in the real super-resolution image and the real original image in the current iteration:

[0047] in, This represents the photon number ratio after the m-th iteration when the fluorescence wavelength λ is the abscissa, i.e., the second photon number ratio. Represents the true super-resolution image after the m-th iteration. The total number of photons in the sample; Represents the true original image after the m-th iteration. The total number of photons in the array.

[0048] After completing the iteration of the preset argument, at the fluorescence wavelength The x-axis represents the ratio of photon numbers. Plot the second curve with y as the ordinate, and place the points on the second curve that are located at... The points corresponding to the interval As the final modeling wavelength, a more accurate measurement matrix can be constructed based on the accurate modeling wavelength of the final output, and a better image reconstruction can be achieved based on the accurate measurement matrix.

[0049] This step follows the same procedure. Figure 5 and Figure 6 Twenty real raw images were used, and four modeling wavelengths were selected: 435, 670, 905, and 1140 nm. 670 nm is the peak wavelength of the fluorescent dye. The wavelengths were spaced 235 nm apart to avoid crosstalk between adjacent PSFs while ensuring the sampling density of the modeling wavelengths. Different measurement matrices can be generated based on the PSFs of the four wavelengths respectively.

[0050] Figure 8 (a)–(d) are super-resolution images obtained by reconstruction based on different modeling wavelengths. Figure 9 (a)–(d) in the text are respectively Figure 8 Enlarged views of the boxed areas in (a)–(d). Figure 10 (a)–(d) in the text are respectively Figure 8 Photon number distribution curves at the underlined points in (a)–(d). Figure 10 (a) in the middle has multiple peaks. Figure 10 (d) in the middle has no peaks, and the reconstruction effect of both is very poor. Figure 10 (b) has multiple peaks, which are very low and wide. Figure 10 In (c), there is only one peak, and the peak is very high and narrow. Therefore, Figures 8-10 (c) showed the best reconstruction effect, with clear microtube boundaries and good continuity. After using a modeling wavelength of 905nm, the super-resolution reconstruction effect was better than that of 670nm.

[0051] Figure 11 This is the relationship curve between the Relative Perspective Level (RLP) and the modeling wavelength obtained after reconstructing the super-resolution image based on 20 frames of real original images and different measurement matrices corresponding to four wavelengths. Only the RLP value corresponding to 905 nm falls within the RLP range, i.e. Figure 11 The area between the two red lines indicates that this wavelength is the optimal modeling wavelength. When the modeling wavelength is too low, PSF broadening leads to energy diffusion, image blurring, and an abnormally low RLP, such as... Figures 9-10 As shown in (a) and (b) in the figure; when the modeling wavelength is too high, the PSF is too narrow, resulting in excessive energy concentration, poor microtubule structure continuity, and abnormally high RLP, such as Figures 9-10 As shown in (d) in the figure.

[0052] S4, Super-resolution reconstruction, specifically includes: S41. Constructing a measurement matrix based on the modeling wavelength, the specific process includes: The pixels of the original image to be reconstructed are subdivided into a dense grid, and each grid point is regarded as a potential monomolecular location; For any grid point, a point spread function (PSF) is constructed based on the modeling wavelength. The original image corresponding to a single molecule of unit intensity at that grid point is generated based on the PSF. The original images of the grid point are concatenated column by column to form a vector, which is used as a column of the measurement matrix. By arranging the vectors obtained from all grid points side by side, we obtain the measurement matrix.

[0053] S42. Perform super-resolution reconstruction of the original image to be reconstructed based on the improved compressed sensing model.

[0054] STORM overcomes the diffraction limit through a sparse activation-precise localization strategy, but it requires thousands of original images, has a long imaging time, and is sensitive to drift, making it difficult to meet the needs of dynamic observation. Compressed sensing theory utilizes the sparsity of the images themselves to achieve high-fidelity reconstruction at a rate far lower than the Nyquist sampling rate, providing a new path for fast STORM.

[0055] Within the compressed sensing framework, the STORM imaging process can be rigorously modeled as the convolution of the fluorescent molecule distribution and the system PSF. After flattening the two-dimensional image into a one-dimensional vector in lexicographical order, this convolution relationship can be transformed into a linear matrix form as shown in equation (1):

[0056] In the formula, The original image vector, The super-resolution image to be recovered. The measurement matrix is ​​determined by the system's PSF. This model utilizes the sparse prior of the image to achieve high-probability reconstruction at a rate much lower than the Nyquist sampling rate.

[0057] Due to background noise and measurement errors in actual imaging, traditional models are often difficult to solve directly. Therefore, an improved compressed sensing model with background estimation and error tolerance is introduced, expressed as follows:

[0058] in, , which is a weight vector used to adjust the contribution of molecules at different positions; For super-resolution image vectors; For measurement matrix; ε is the original image vector to be reconstructed; ε is the error tolerance parameter, used to control the balance between reconstruction accuracy and robustness. for The j-th element in; Let x be the i-th element. The above formula minimizes the weighted L1 norm, effectively suppressing background noise while maintaining image sparsity, and is suitable for STORM reconstruction under high-density molecular distribution conditions.

[0059] Compared to traditional STORM methods, the compressed sensing-based reconstruction strategy of this invention offers the following advantages: 1) It can achieve high-density reconstruction even when fluorescent molecules partially overlap, improving temporal resolution; 2) It reduces the number of image frames required, mitigating the effects of photobleaching and sample drift. This invention utilizes the prior knowledge of image sparsity, allowing the recovery of high-density molecular positions from the original image by solving a sparse reconstruction problem even when corresponding molecular speckles partially overlap. This results in higher-density molecular localization with fewer frames and in a shorter time, ultimately improving temporal resolution.

[0060] In other embodiments, the present invention also provides an image optimization and reconstruction system based on the photon number ratio after logarithmic transformation, comprising: The image acquisition module is used to acquire simulated super-resolution images, simulated original images, and real original images; The first iteration module is used to iteratively calculate the ratio of photons in the simulated super-resolution image and the simulated original image at different magnifications, with the real original image background as a constraint and according to a preset magnification step size, as the first photon ratio. The photon number ratio determination module is used to plot a first curve with the multiplier on the horizontal axis and the first photon number ratio on the vertical axis, and to determine the range of the first photon number ratio. The second iteration module is used to iteratively reconstruct the real super-resolution image corresponding to the real original image under different fluorescence wavelengths according to the preset fluorescence wavelength step size, and calculate the photon number ratio in the real super-resolution image and the real original image under different fluorescence wavelengths as the second photon number ratio. The modeling wavelength determination module is used to plot a second curve with fluorescence wavelength as the abscissa and the second photon number ratio as the ordinate, and to determine the fluorescence wavelength corresponding to the point on the second curve that is located in the range of the first photon number ratio, which is then used as the modeling wavelength. The measurement matrix construction module is used to construct a measurement matrix based on the modeling wavelength; The reconstruction module is used to perform super-resolution reconstruction of the original image to be reconstructed based on the measurement matrix.

[0061] In another embodiment, the present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor; when the processor executes the computer program, it implements the steps of the image optimization and reconstruction method based on the photon number ratio after logarithmic transformation as described above.

[0062] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0063] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An image optimization and reconstruction method based on the photon number ratio after logarithmic transformation, characterized in that, Includes the following steps: Acquire simulated super-resolution images, simulated original images, and real original images; With the real original image background as a constraint, the ratio of photons in the simulated super-resolution image and the simulated original image at different magnification steps is iteratively calculated according to a preset magnification step size, and is used as the first photon number ratio. Plot the first curve with the magnification rate on the x-axis and the first photon number ratio on the y-axis, and determine the range of the first photon number ratio. According to the preset fluorescence wavelength step size, the real super-resolution image corresponding to the real original image under different fluorescence wavelengths is reconstructed iteratively, and the photon number ratio in the real super-resolution image and the real original image under different fluorescence wavelengths is calculated as the second photon number ratio. Plot a second curve with fluorescence wavelength on the x-axis and the second photon number ratio on the y-axis. Determine the fluorescence wavelength corresponding to the point on the second curve that is in the range of the first photon number ratio, and use it as the modeling wavelength. Construct a measurement matrix based on the modeling wavelength; Super-resolution reconstruction is performed on the original image to be reconstructed based on the measurement matrix.

2. The image optimization and reconstruction method based on the photon number ratio after logarithmic transformation as described in claim 1, characterized in that, The process of acquiring simulated super-resolution images, simulated original images, and real original images includes: Multiple real original image frames were obtained based on a pre-built real experimental platform. The real original image was obtained by superimposing the real original image frames. Multiple simulated original image frames are obtained based on a pre-built simulation experimental platform. The simulated original image is obtained by superimposing the simulated original image frames. Reconstruct all the original simulated image frames to obtain the corresponding simulated super-resolution image frames. Then, superimpose the simulated super-resolution image frames to obtain the simulated super-resolution image. The experimental parameters of the simulation experimental platform and the real experimental platform are consistent.

3. The image optimization and reconstruction method based on the photon number ratio after logarithmic transformation as described in claim 1, characterized in that, In the process of plotting the first curve, each iteration includes: Determine the multiplier for the current iteration stage according to the preset multiplier step size. ,in, This represents the multiplier used in the j-th iteration. This represents the multiplier used in the (j-1)th iteration. This is the preset multiplier step size; The background of the real original image is added to the simulated original image. After adding the background, the simulated original image is magnified and logarithmically calculated. The specific process is as follows: in, This represents the original simulation image after the j-th iteration; This represents the original simulation image after the (j-1)th iteration; Represents the background of the original image; The process of performing magnification and logarithmic calculation on the simulated super-resolution image is as follows: in, This represents the simulated super-resolution image after the j-th iteration; This represents the simulated super-resolution image after the (j-1)th iteration; Calculate the photon number ratio between the simulated super-resolution image and the simulated original image in the current iteration: in, Represents the simulated super-resolution image after the j-th iteration. The total number of photons in the sample; Represents the original simulation image after the j-th iteration. The total number of photons in the sample; The first photon number ratio represents the ratio of the number of photons in the simulated super-resolution image and the simulated original image after the j-th iteration, when the magnification ratio i is the horizontal axis.

4. The image optimization and reconstruction method based on the photon number ratio after logarithmic transformation as described in claim 3, characterized in that, The average photon count of a blank area at a preset distance from the sample structure in the real original image is used as the background and added to the simulated original image.

5. The image optimization and reconstruction method based on the photon number ratio after logarithmic transformation as described in claim 1, characterized in that, The lower limit of the first photon number ratio range is the minimum photon number ratio in the first curve, and the upper limit is the maximum photon number ratio in the first curve.

6. The image optimization and reconstruction method based on the photon number ratio after logarithmic transformation as described in claim 1, characterized in that, In the process of plotting the second curve, each iteration includes: The fluorescence wavelength for the current iteration stage is determined according to the preset fluorescence wavelength step size. , ,in, This indicates the fluorescence wavelength used in the m-th iteration. This indicates the fluorescence wavelength used in the (m-1)th iteration. The preset fluorescence wavelength step size; Based on the fluorescence wavelength at the current iteration stage The corresponding measurement matrix is ​​used to reconstruct all real original image frames and obtain the corresponding real super-resolution image frames respectively. All real super-resolution image frames are superimposed to obtain a real super-resolution image; The logarithm is calculated for both the original image and the super-resolution image. The specific calculation process is as follows: in, This represents the true super-resolution image after the m-th iteration; This represents the true super-resolution image after the (m-1)th iteration; This represents the actual original image after the m-th iteration; This represents the true original image after the (m-1)th iteration; Calculate the ratio of photon counts in the real super-resolution image and the real original image in the current iteration: in, This represents the photon number ratio after the m-th iteration when the fluorescence wavelength λ is the abscissa, i.e., the second photon number ratio. Represents the true super-resolution image after the m-th iteration. The total number of photons in the sample; Represents the true original image after the m-th iteration. The total number of photons in the array.

7. The image optimization and reconstruction method based on the photon number ratio after logarithmic transformation as described in claim 1, characterized in that, The process of constructing a measurement matrix based on the modeling wavelength includes: The pixels of the original image to be reconstructed are subdivided into a dense grid, and each grid point is regarded as a potential monomolecular location; For any grid point, a point spread function (PSF) is constructed based on the modeling wavelength. The original image corresponding to a single molecule of unit intensity at that grid point is generated based on the PSF. The original images of the grid point are concatenated column by column to form a vector, which is used as a column of the measurement matrix. By arranging the vectors obtained from all grid points side by side, we obtain the measurement matrix.

8. The image optimization and reconstruction method based on the photon number ratio after logarithmic transformation as described in claim 1, characterized in that, Super-resolution reconstruction of the original image to be reconstructed is performed based on an improved compressed sensing model. The improved compressed sensing model is represented as follows: in, , which is a weight vector used to adjust the contribution of molecules at different positions; For super-resolution image vectors; For measurement matrix; ε is the original image vector to be reconstructed; ε is the error tolerance parameter, used to control the balance between reconstruction accuracy and robustness. for The j-th element in; Let be the i-th element in x.

9. An image optimization and reconstruction system based on the photon number ratio after logarithmic transformation, characterized in that, include: The image acquisition module is used to acquire simulated super-resolution images, simulated original images, and real original images; The first iteration module is used to iteratively calculate the ratio of photons in the simulated super-resolution image and the simulated original image at different magnifications, with the real original image background as a constraint and according to a preset magnification step size, as the first photon ratio. The photon number ratio determination module is used to plot a first curve with the multiplier on the horizontal axis and the first photon number ratio on the vertical axis, and to determine the range of the first photon number ratio. The second iteration module is used to iteratively reconstruct the real super-resolution image corresponding to the real original image under different fluorescence wavelengths according to the preset fluorescence wavelength step size, and calculate the photon number ratio in the real super-resolution image and the real original image under different fluorescence wavelengths as the second photon number ratio. The modeling wavelength determination module is used to plot a second curve with fluorescence wavelength as the abscissa and the second photon number ratio as the ordinate, and to determine the fluorescence wavelength corresponding to the point on the second curve that is located in the range of the first photon number ratio, which is then used as the modeling wavelength. The measurement matrix construction module is used to construct a measurement matrix based on the modeling wavelength; The reconstruction module is used to perform super-resolution reconstruction of the original image to be reconstructed based on the measurement matrix.

10. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor; characterized in that, when the processor executes the computer program, it implements the steps of the image optimization and reconstruction method based on the photon number ratio after logarithmic transformation as described in any one of claims 1-8.

Images