Construction method and application of fluorescence microscopic signal detection model

Abstract

The invention relates to the technical field of signal identification and positioning, and discloses a construction method and application of a fluorescence microscopic signal detection model. The invention relates to a fitting method based on self-supervised deep learning, which comprises the following steps of: representing a fluorescence microscope signal as a pattern of interest (POI) or a combination thereof, and executing parallel POI fitting and global denoising in a large receptive field by utilizing a self-supervised deep learning framework. The ultra-high-density fluorescence microscopic signal detection under the background of serious noise and strong fluorescence is realized in an unsupervised learning mode, and meanwhile, very high universality and usability are kept. The method can simply adapt to various fluorescence microscopic signal detection tasks by simply setting a POI example provided for the fluorescence microscopic signal detection model and adjusting a hyper-parameter related to sensitivity, and remarkable performance improvement is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of signal recognition and positioning, and in particular to a method for constructing a fluorescence microscopy signal detection model and its application. Background Art

[0002] Fluorescence microscopy, with its unique ability to visualize and quantify vivid details of molecular and cellular processes, greatly contributes to the analysis and understanding of biological mechanisms. However, affected by sample conditions, sensor performance and imaging environment, fluorescence microscopy images often suffer from problems such as weak signals, overlap, multi-source complex noise pollution and complex background interference, which hinder accurate and efficient signal detection.

[0003] There are currently a variety of fluorescence microscopy image processing methods to solve the above problems, but they do not meet the needs of different fields and applications. For example, methods based on spatiotemporal redundancy use the unpredictability of noise and the redundancy of signals to achieve denoising and signal enhancement, but this method is not suitable for processing rapidly changing signals in time and space, nor can it determine the existence and location of signals. At the same time, they can only remove noise, but cannot distinguish between signals and backgrounds with similar spatiotemporal redundancy. For another example, methods based on ground truth values ​​achieve signal detection by training models on labeled datasets to learn the mapping between images and their corresponding ground truth signals, but this method requires a large number of high-quality manual annotation labels, which increases the workload and cost of signal processing, and for some high-precision annotation tasks such as single molecule localization microscopy (SMLM), manual annotation is even impractical. In addition, methods based on conventional fitting use algorithms such as maximum likelihood estimation (MLE) to sequentially fit signals from images for signal detection, but its inherent characteristics, such as sequential fitting and rough modeling of interference factors, significantly reduce performance when dealing with weak or overlapping signals, severe noise or complex backgrounds.

[0004] In general, current methods need to face a trade-off between versatility, performance, and practicality, and there is still a lack of a universal, high-performance, and easy-to-use method for fluorescence microscopy signal detection under a variety of complex conditions. Summary of the invention

[0005] The present invention provides a method for constructing a fluorescence microscopy signal detection model and application thereof, so as to solve the defect of lacking a universal, high-performance and easy-to-use fluorescence microscopy signal detection method.

[0006] The present invention provides a method for constructing a fluorescence microscopy signal detection model, comprising:

[0007] Get the original fluorescence microscope image Iori , where the original fluorescence microscope image I ori The fluorescence microscopy signals in the image have been characterized as a pattern of interest (POI) or a combination thereof;

[0008] Initializing the deep learning model

[0009] Leveraging Deep Learning Models The original fluorescence microscope image I ori Convert to Fitting Graph I fit , where the fitting graph I fit Encoded with the original fluorescence microscope image I ori The location information and amplitude of POI in the

[0010] Fitting Figure I fit Perform group convolution with the preset POI examples to obtain the denoised image I dns ;

[0011] Maximize the original fluorescence microscopy image I ori and denoised image I dns The L2 similarity between them, and minimize the fitting graph I fit L1 regularization term to optimize deep learning models

[0012] Initialize the decoding module

[0013] Using the decoding module From the fitting diagram I fit The original fluorescence microscope image I is obtained by decoding ori Location information of POI in the

[0014] From the original fluorescence microscopy image I ori Subtract the denoised image I dns , get the noise image I ns ;

[0015] The noise image I ns With the fitting diagram I fit The sum over the channel dimension Merge to get the noise fitting graph And based on the noise fitting graph The original fluorescence microscope image I is estimated ori The magnitude of the POI;

[0016] Combined with the estimated original fluorescence microscope image I ori The amplitude of POI, the preset POI example and the decoded original fluorescence microscope image Iori The location information of POI in the image is obtained. syn ;

[0017] Maximize the synthetic image I syn And the denoised image I dns The L1 similarity between them is used to optimize the decoding module

[0018] Combined with optimized deep learning model and decoding module The fluorescence microscopy signal detection model is obtained.

[0019] According to a method for constructing a fluorescence microscopy signal detection model provided by the present invention, the expression of the preset POI example is:

[0020]

[0021] In the expression of the preset POI example, represents a set of C pre-acquired POI images in the channel dimension, (x, y) represents the pixel coordinates, represents the minimum value of the c-th POI image, Represents the maximum value of the c-th POI image.

[0022] According to the method for constructing a fluorescence microscopy signal detection model provided by the present invention, a deep learning model It is a fully convolutional neural network with U-net structure, a deep learning model The loss function for:

[0023]

[0024] Deep Learning Models The loss function middle, I ori represents the original fluorescence microscope image, I dns represents the denoised image, I fit represents the fitting graph, and λ represents the regularization strength.

[0025] According to a method for constructing a fluorescence microscopy signal detection model provided by the present invention, the fitting diagram I fit Perform group convolution with the preset POI examples to obtain the denoised image I dns ,include:

[0026] According to the first expression, the fitting graph I fit Perform group convolution with the preset POI examples to obtain the denoised image I dns , where the first expression is:

[0027]

[0028] In the first expression, I dns represents the denoised image, I fit Indicates a fitted graph, and the colon ":" indicates all x- or y-axis coordinate values ​​in that dimension. represents a two-dimensional convolution, Preset POI example showing mean sum normalization.

[0029] According to a method for constructing a fluorescence microscopy signal detection model provided by the present invention, the fitting diagram I fit The sum over the channel dimension The expression is:

[0030]

[0031] According to a method for constructing a fluorescence microscopy signal detection model provided by the present invention, the decoding module From the fitting diagram I fit The original fluorescence microscope image I is obtained by decoding ori The location information of POI in the , including:

[0032] Based on the fitting graph I fit The sum over the channel dimension Through the second expression, the pixel area that may contain POI is segmented A collection of;

[0033] Based on pixel regions that may contain POIs The third expression is used to calculate the centroid of the ith region of interest (ROI). Sub-pixel coordinates of the POI As the original fluorescence microscopy image I ori The location information of POI in the

[0034] According to a method for constructing a fluorescence microscopy signal detection model provided by the present invention, the second expression is:

[0035]

[0036] In the second expression, θ represents the fit The sum over the channel dimension The segmentation threshold is pre-initialized within the pixel value range.

[0037] According to a method for constructing a fluorescence microscopy signal detection model provided by the present invention, the third expression is:

[0038]

[0039] According to a method for constructing a fluorescence microscopy signal detection model provided by the present invention, the method comprises: ori Subtract the denoised image I dns , get the noise image I ns ,include:

[0040] From the original fluorescence microscopy image I ori Subtract the denoised image I dns , get the noise image I ns , I ns =I ori -I dns ;

[0041] Preset POI example using mean sum normalization The maximum value of the noise image I ns Normalize.

[0042] According to a method for constructing a fluorescence microscopy signal detection model provided by the present invention, the noise image I ns With the fitting diagram I fit The sum over the channel dimension Merge to get the noise fitting graph And based on the noise fitting graph The original fluorescence microscope image I is estimated ori The range of POI, including:

[0043] Through the fourth expression, the normalized noise image I ns With the fitting diagram I fit The sum over the channel dimension Merge to get the noise fitting graph

[0044] By using the fifth expression, each ROI is convolved with the preset POI example corresponding to its z-axis position and the maximum value is calculated to estimate the i-th ROI The amplitude A of the POI contained in i .

[0045] According to a method for constructing a fluorescence microscopy signal detection model provided by the present invention, the fourth expression is:

[0046]

[0047] In the fourth expression, Represents the normalized noise image I ns The value at pixel coordinate (x,y), Represents the fitting graph Ifit The sum over the channel dimension The value at pixel coordinate (x,y), Represents the noise fitting plot The value at pixel coordinate (x,y).

[0048] According to a method for constructing a fluorescence microscopy signal detection model provided by the present invention, the fifth expression is:

[0049]

[0050] In the fifth expression, “⊙” represents matrix element-by-element multiplication, and M(·) represents a function that outputs a mask: pixels in the input set are 1, and the rest are 0, at position The preset POI examples obtained at are obtained by interpolation.

[0051] According to a method for constructing a fluorescence microscopy signal detection model provided by the present invention, the original fluorescence microscopy image I obtained by combining the estimation ori The amplitude of POI, the preset POI example and the decoded original fluorescence microscope image I ori The location information of POI in the image is obtained. syn ,include:

[0052] By combining the estimated original fluorescence microscope image I ori The amplitude of POI, the preset POI example and the decoded original fluorescence microscope image I ori The location information of POI in the image is obtained. syn , where the sixth expression is:

[0053]

[0054] In the sixth expression, H and W represent the height and width of the preset POI instance, respectively. and The pixel value of the preset POI example at is obtained by interpolation.

[0055] According to a method for constructing a fluorescence microscopy signal detection model provided by the present invention, a decoding module The loss function is:

[0056]

[0057] According to a method for constructing a fluorescence microscopy signal detection model provided by the present invention, the fluorescence microscopy signal detection model adopts any one of the following or any combination thereof to optimize model performance: center-aligned upsampling, non-uniform background learning, and POI diffusion.

[0058] The present invention also provides a fluorescence microscopy signal detection method, comprising:

[0059] receiving a fluorescence microscope image to be tested;

[0060] According to the fluorescence microscopy image to be tested, a fluorescence microscopy signal detection result of the fluorescence microscopy image to be tested is obtained by using a fluorescence microscopy signal detection model obtained by any of the above-mentioned methods for constructing a fluorescence microscopy signal detection model, wherein the fluorescence microscopy signal detection result includes position information and / or amplitude information of the fluorescence microscopy signal.

[0061] The present invention also provides a fluorescence microscopy signal detection system, comprising:

[0062] The image receiving module is used to: receive the fluorescence microscope image to be tested;

[0063] A signal detection module is used to obtain, based on the fluorescence microscopy image to be tested, a fluorescence microscopy signal detection model obtained by any of the above-mentioned methods for constructing a fluorescence microscopy signal detection model, a fluorescence microscopy signal detection result of the fluorescence microscopy image to be tested, wherein the fluorescence microscopy signal detection result includes position information and / or amplitude information of the fluorescence microscopy signal.

[0064] The present invention also provides an electronic device, comprising a processor and a memory storing a computer program, wherein when the processor executes the computer program, the method for constructing a fluorescence microscopy signal detection model and / or the fluorescence microscopy signal detection method described above is implemented.

[0065] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for constructing a fluorescence microscopy signal detection model and / or fluorescence microscopy signal detection methods.

[0066] The present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute any of the above-mentioned methods for constructing a fluorescence microscopy signal detection model and / or the fluorescence microscopy signal detection method.

[0067] The present invention provides a method for constructing a fluorescence microscopy signal detection model and its application, which can achieve high-throughput and sensitive fluorescence microscopy signal detection under strong interference by simply setting the POI examples provided to the fluorescence microscopy signal detection model (referred to as the DEPAF model in this article) and adjusting a hyperparameter related to sensitivity, without requiring any benchmark truth value or imaging physics model in a specific field. While ensuring performance, the prediction speed of DEPAF for locating low-density POIs can match that of traditional fitting-based methods, but its performance is better than them when using high-density POIs, because the computational complexity of the first branch of DEPAF depends only on the size of the image, not the number of POIs therein; although the second branch involves sequential processing, it does not include any iterative process found in traditional fitting-based methods. The present invention fully demonstrates the high versatility, performance and practicality of DEPAF in various tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0069] Figure 1 This is one of the flow diagrams of a fluorescence microscopy signal detection method provided by the present invention.

[0070] Figure 2 The second schematic diagram of the process flow of a fluorescence microscopy signal detection method provided by the present invention. In the figure, Figure a shows the first branch in the process flow of the method; Figure b shows the second branch in the process flow of the method. Figure c shows a fully convolutional neural network based on U-net. The structure is used to fit all POIs in the input image and denoise the input image. This neural network includes a downsampling stage followed by an upsampling stage. Each stage consists of multiple (de)convolutional layers with 3×3 filters, as well as non-trainable operations such as maximum pooling (stride 2), Dropout (probability 50%), and rectified linear unit (ReLU) activation. The first convolutional layer in the downsampling stage has N filters, where N is the maximum of 64 and the number of POIs. In each downsampling stage, the resolution is halved and the number of filters is doubled; the opposite is true in each upsampling stage. The black arrows indicate jump connections. Figure d shows the decoding module The prior knowledge used in the decoding module is shown in Figure e. The processing process is used to decode the location information of POI from the fitting graph. Conv. represents the convolution layer; MaxPool. represents the maximum pooling layer; Deconv. represents the deconvolution layer.

[0071] Figure 3 This is a schematic structural diagram of a fluorescence microscopy signal detection system provided by the present invention.

[0072] Figure 4 This is a schematic structural diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0073] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments, and they should not be understood as limitations on the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. In the description of the present invention, it should be understood that the terms used are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0074] Combination Figures 1 to 4 A fluorescence microscopy signal detection method provided by the present invention is described. The fluorescence microscopy signal detection method provided by the present invention can be divided into two aspects: construction and application of a fluorescence microscopy signal detection model. In the construction method of the fluorescence microscopy signal detection model, steps S110-150 are the first branch in the method flow, steps S210-260 are the second branch in the method flow, and step S310 is the final step after the first branch and the second branch in the method flow are completed. The execution subject of a fluorescence microscopy signal detection method provided by the present invention can be any applicable terminal-side device or network-side device, such as a fluorescence microscopy signal detection device, etc.

[0075] See also Figure 1 The present invention provides a fluorescence microscopy signal detection method, which may include:

[0076] S110, obtaining original fluorescence microscope image I ori , where the original fluorescence microscope image I ori The fluorescence microscopy signals in have been characterized as a Pattern of Interest (POI) or a combination thereof.

[0077] S120, Initialize the deep learning model

[0078] S130, using deep learning models The original fluorescence microscope image I ori Convert to Fitting Graph I fit , where the fitting graph I fit Encoded with the original fluorescence microscopy image i ori The location information and amplitude of POI in the

[0079] S140, fitting graph i fit Perform group convolution with the preset POI examples to obtain the denoised image i dns .

[0080] S150, maximize the original fluorescence microscopy image i ori and denoised image I dns The L2 similarity between them, and minimize the fitting graph I fit L1 regularization term to optimize deep learning models

[0081] This embodiment pre-creates a set of C pre-acquired POI images in the channel dimension And normalize it to the range of 0-1 to model it as a preset POI example. The expression of the preset POI example is:

[0082]

[0083] In the expression of the preset POI example, represents a set of c pre-acquired POI images in the channel dimension, (x, y) represents the pixel coordinates, represents the minimum value of the c-th POI image, Represents the maximum value of the c-th POI image.

[0084] In this embodiment, according to the first expression, the fitting diagram I fit Perform group convolution with the preset POI examples to obtain the denoised image i dns , where the first expression is:

[0085]

[0086] In the first expression, i dns represents the denoised image, i fit Indicates a fitted graph, and the colon ":" indicates all x- or y-axis coordinate values ​​in that dimension. represents a two-dimensional convolution, Preset POI example showing mean sum normalization.

[0087] The deep learning model in this embodiment It is a fully convolutional neural network with a U-net structure, which is used to simultaneously fit all POIs in the input image and denoise the input image, a deep learning model The loss function for:

[0088]

[0089] Deep Learning Models The loss function middle, I ori represents the original fluorescence microscope image, I dns represents the denoised image, I fit represents the fitting graph, and λ represents the regularization strength.

[0090] Among them, the loss function It includes an L2 term and an L1 regular term. The L2 term promotes I dns and I ori The similarity between them is used to achieve POI reconstruction in the image, while the L1 regularization term promotes I fit The regularization strength λ is an adjustable hyperparameter, so the denoising strength and the sensitivity of POI detection can be controlled. fit To suppress all other image information that is different from the POI spatial distribution, thereby eliminating various composite noises. This method does not require noise modeling and only relies on POI information.

[0091] After the first branch in the method flow is completed, the second branch is initiated.

[0092] The second branch of this embodiment uses an optimizable decoding module From I fit The POI location is decoded from ori The noise in the POI (not just the POI) is also similar. fit may contain fitting values ​​generated by noise rather than POI. These noise fitting values ​​usually do not exceed the noise level. Therefore, this embodiment uses a threshold θ to filter them out. However, since the noise level is difficult to estimate, this embodiment uses an iterative optimization method to robustly solve this problem. Specifically, this embodiment first optimizes I along the channel dimension. fit Sum to get That is, fitting diagram I fit The sum over the channel dimension The expression is:

[0093]

[0094] S210, initializing an optimizable decoding module Used to decode the position of POI from the fitted image.

[0095] S220, using decoding module From the fitting diagram I fit The original fluorescence microscope image I is obtained by decoding ori The location information of POI in the

[0096] Specifically, this embodiment is based on the fitting diagram I fit The sum over the channel dimension Through the second expression, the pixel area that may contain POI is segmented A set of; where the second expression is:

[0097]

[0098] In the second expression, θ represents the fit The sum over the channel dimension The segmentation threshold is initialized within the pixel value range.

[0099] In order to further divide the region of interest (ROI) containing a single POI, this embodiment summarizes the following two priors:

[0100] (a) Shape prior: Due to the discrete nature and intrinsic size of pixels, a point in the continuous space (i.e., a point at sub-pixel coordinates in the image) is represented in the discrete pixel space as 1 to 4 pixels in a 2×2 pixel block, which corresponds to the shape of each ROI.

[0101] (b) Priority prior: For ROIs containing 1 to 4 non-zero pixels, the sub-pixel coordinate ranges they can represent increase in sequence. According to the classical probability model, their probability of occurrence also increases in sequence. Therefore, their division priority also increases in sequence.

[0102] Therefore, based on the above priors, the ROI containing the fitted position and magnitude of a single POI is divided by Algorithm 1.

[0103]

[0104] It is worth noting that the computational efficiency can be significantly improved by processing each 8-separated pixel region separately (ie, all pixels in each region do not have any 8-neighborhood relationship with each other) instead of directly applying the algorithm to the large-size original image.

[0105] Then, this embodiment is based on the pixel area that may contain POI The third expression is used to calculate the centroid of the i-th ROI. Sub-pixel coordinates of the POI As the original fluorescence microscopy image I ori The location information of the POI in , where the third expression is:

[0106]

[0107] S230, from the original fluorescence microscope image I ori Subtract the denoised image I dns , get the noise image I ns , I ns =I ori -I dns , and use the preset POI example of mean sum normalization The maximum value of the noise image I ns Normalize.

[0108] S240, the noise image I ns With the fitting diagram I fit The sum over the channel dimension Merge to get the noise fitting graph And based on the noise fitting graph The original fluorescence microscope image I is estimated ori The magnitude of the POI.

[0109] Specifically, this embodiment uses the fourth expression to convert the normalized noise image I ns With the fitting diagram I fit The sum over the channel dimension Merge to get the noise fitting graph By using the fifth expression, each ROI is convolved with the preset POI example corresponding to its z-axis position and the maximum value is calculated to estimate the i-th ROI The amplitude A of the POI contained in i .

[0110] Among them, the fourth expression is:

[0111]

[0112] In the fourth expression, Represents the normalized noise image I ns The value at pixel coordinate (x,y), Represents the fitting graph I fit The sum over the channel dimension The value at pixel coordinate (x,y), Represents the noise fitting plot The value at pixel coordinate (x,y).

[0113] The fifth expression is:

[0114]

[0115] In the fifth expression, “⊙” represents matrix element-by-element multiplication, and M(·) represents a function that outputs a mask: pixels in the input set are 1, and the rest are 0, at position The preset POI examples obtained at are obtained by interpolation.

[0116] S250, combining the estimated original fluorescence microscope image I through the sixth expression ori The amplitude of POI, the preset POI example and the decoded original fluorescence microscope image I ori The location information of POI in the image is obtained. syn , where the sixth expression is:

[0117]

[0118] In the sixth expression, H and W represent the height and width of the preset POI instance, respectively. and The pixel value of the preset POI example at is obtained by interpolation.

[0119] S260, maximize the composite image I syn and denoised image I dns The L1 similarity between them is used to optimize the decoding module

[0120] Based on the above construction, a key trade-off emerges: choosing a value of θ below the noise level will result in The noise fit values ​​in are retained. This results in syn Generate multiple Noise POIs derived from the noise in the image. Due to the randomness and instability of the noise, these noise POIs cannot be compared with I dns The unreal POIs in , which are obtained by convolving the noise fitting values ​​with the POI examples. syn Will deviate from I dns Conversely, choosing a value of θ that is too large above the noise level will result in erroneous filtering of The non-noise fitted value in I syn The non-noise POI generated in dns The number of syn Deviation I dns Therefore, θ in this embodiment can be obtained by maximizing I dns and I synThe similarity between them is optimized, that is, the loss function is minimized, and the decoding module The loss function is:

[0121]

[0122] because Since θ is not differentiable and the optimization is univariate, this embodiment uses a gradient-free optimization method. Specifically, this embodiment first uses Bayesian optimization to quickly estimate The shape of the curve is calculated and a suboptimal value of θ is found. After that, a grid search is performed around this suboptimal value of θ to find the optimal value of θ. This method aims to balance the speed and accuracy of optimization.

[0123] After the first branch and the second branch in the method flow are completed, S130 may be executed.

[0124] S310, combined with optimized deep learning model and decoding module The fluorescence microscopy signal detection model is obtained.

[0125] In one embodiment, any one of the following or any combination thereof may be used to optimize the performance of the fluorescence microscopy signal detection model: center-aligned upsampling technology to improve the accuracy of POI positioning; non-uniform background learning to enhance anti-interference capabilities; POI diffusion to promote model convergence in multi-POI detection scenarios.

[0126] (a) Center-aligned upsampling: In the second branch of the method, the inherent pixel size may limit The ability to distinguish overlapping POIs and accurately estimate their locations. To address this limitation, the pixel size can be reduced by applying the same upsampling factor to the POI examples and the input image at the beginning of the pipeline. However, an upsampling factor that is too large may cause the method to mistakenly detect a single POI as multiple POIs, thereby increasing the false alarm rate. Therefore, in scenarios that require high positioning accuracy and distinction between overlapping POIs (such as SMLM), this embodiment uses an upsampling factor of 1.5. In general, a factor of 1 (i.e., no upsampling) is sufficient.

[0127] This upsampling should also maintain the shape of the POI in the POI example and the input image, as well as the center position of the POI example. The shape of the POI is affected by the interpolation algorithm used in the upsampling process. Different tasks may require different interpolation algorithms. For example, cubic spline interpolation has been shown to accurately capture the shape of any point spread function (PSF), so it is attempted to be used in PSF-related tasks in this embodiment. However, this embodiment finds that the choice of interpolation algorithm is not critical, and cubic spline interpolation is effective even for non-PSF-related tasks. In addition, the interpolation algorithm used to obtain sub-pixel values ​​in the pipeline is the same as that used here.

[0128] To illustrate how to maintain the center position of a POI example after upsampling, this embodiment considers the following single-channel two-dimensional POI example with a height and width of H and W. Assume that the pixel coordinate vector along the x-axis in the POI example is:

[0129] x ori =(0,1,…,W-1).

[0130] Assume that the initial upsampling factor is k>1. In order to ensure that the side length of the POI example remains an odd number after upsampling, this embodiment first corrects the upsampling factor in the following manner:

[0131]

[0132] Then the interpolation step size and upsampling width x ori Expands to x exp :

[0133]

[0134] Then x ori and x exp Perform center alignment to obtain the x-axis coordinate x of the interpolation query point qry :

[0135]

[0136] in and They are x ori and x exp Similarly, the y-axis coordinate y of the interpolated query point qry The same steps can be used to calculate the coordinates of the interpolated query point. qry and qry The complete grid is constructed as:

[0137]

[0138] To simplify the calculation, the upsampling process of the input image is the same as that of the POI example, except that the side length after upsampling is required to be an odd number. Therefore, after POI positioning on the upsampled image, the POI position can be mapped back to the original coordinate system in the following way:

[0139]

[0140] where x up is the x-axis coordinate of the POI in the upsampled image. Similarly, can also be calculated accordingly.

[0141] (b) Non-uniform background learning. Non-uniform background in an image may affect the performance of POI detection. To solve this problem, this embodiment adds an additional constant image block (i.e., a small image with equal values ​​everywhere) with a size three times the center area of ​​the 95% pixel value distribution of the signal POI example as the background POI. bg and use The output channel generates the background fitting image I fitBG This method enables the model to learn arbitrary non-uniform backgrounds when detecting POIs. When applying this technique, it is used to optimize The loss function becomes:

[0142]

[0143] Note that since the background does not exhibit natural sparsity, I fitBG Not subject to L1 regularization.

[0144] (c) POI diffusion. When this method is applied to the multi-POI detection task, the number of POI examples determines the deep learning model. Output I fit Therefore, as the number of POI examples increases, I fit The search space will also expand, which may affect To solve this problem, this embodiment introduces POI diffusion technology. This technology first merges all POI examples into the central channel through mean pooling. In the last convolutional layer of the network, only the kernels corresponding to the central channel are activated for training. During training, if the validation loss does not decrease after a certain number of iterations, POI diffusion is triggered. At this time, new activation channels are determined by bisection, and POI examples are merged back into their nearest activation channels by mean pooling. The kernels corresponding to these newly activated channels inherit the initial weights of the previously activated kernels and are activated for training. This process is repeated until all channels are activated. This technique aims to make Progressive learning is achieved by finding the optimal solution in a smaller search space and then gradually expanding to a larger search space. In this process, by initially merging POI examples, their frequency is minimized and then gradually increased through POI diffusion during training. This is consistent with the natural property of the model to gradually fit from low-frequency patterns to high-frequency patterns. These factors make this technique effective in promoting the convergence of the method in multi-POI detection tasks.

[0145] The fluorescence microscopy signal detection model obtained by the method for constructing a fluorescence microscopy signal detection model provided by the present invention can be applied to the identification of various fluorescence signals (such as dynamic fluorescence signals) to form a fluorescence microscopy signal detection method, which can include the following steps: obtaining a fluorescence microscopy image to be tested; according to the fluorescence microscopy image to be tested, obtaining a fluorescence microscopy signal detection result of the fluorescence microscopy image to be tested by using the fluorescence microscopy signal detection model obtained by any of the methods for constructing a fluorescence microscopy signal detection model described above, wherein the fluorescence microscopy signal detection result includes position information and / or amplitude information of the fluorescence microscopy signal.

[0146] Among them, dynamic fluorescence signals include but are not limited to FISH signals, CRISPR signals and nano signals.

[0147] Among them, FISH signals include but are not limited to the following types: (a) Conventional DNA / RNA FISH: used to detect and locate specific DNA or RNA sequences. (b) Multiplex DNA / RNA FISH: used to simultaneously detect multiple different DNA or RNA sequences, including but not limited to MERFISH, seqFISH, seqFISH+, etc. (c) Quantitative DNA / RNA FISH: used to quantitatively determine the amount of target DNA or RNA. (d) Alternating DNA / RNA FISH: used to detect complex genomic rearrangements. (e) Color-coded DNA / RNA FISH: used to distinguish multiple target sequences by different colors. (f) High-resolution DNA / RNA FISH: used to achieve fine positioning at the subcellular level. (g) Super-resolution DNA / RNA FISH: combined with super-resolution microscopy technology to improve detection accuracy. (h) Reversible DNA / RNA FISH: allows the same sample to be reused for the detection of different targets. (i) Rapid DNA / RNA FISH: used to speed up the detection process. (j) Automated DNA / RNA FISH: used for high-throughput automated analysis. (k) All other fluorescent FISH methods that have been published to date, including but not limited to: genomic in situ hybridization (GISH), multicolor fluorescent in situ hybridization (M-FISH), primer in situ DNA synthesis (PRINS), in situ PCR hybridization (IS-PCR), DNA fiber fluorescent in situ hybridization (Fiber-FISH), bacterial artificial chromosome fluorescent in situ hybridization (BAC-FISH), tyramide signal amplification fluorescent in situ hybridization (TSA-FISH), molecular beacon fluorescent in situ hybridization (MB-FISH), peptide nucleic acid fluorescent in situ hybridization (PNA-FISH), flow fluorescent in situ hybridization (FlowFISH), Oligopaint and its series of techniques, HD-FISH, Cas-FISH, CRISPRLiveFISH, GOLD FISH, etc.

[0148] CRISPR signals include, but are not limited to, the following types: (a) CRISPR-Cas9: a conventional CRISPR system for gene editing. (b) CRISPR-Cas12: a CRISPR system with an expanded target recognition range. (c) CRISPR-Cas13: a CRISPR system for RNA editing. (d) CRISPRa (CRISPR activation): used for gene activation. (e) CRISPRi (CRISPR interference): used for gene suppression. (f) Base Editing CRISPR: used for precise single-base modification. (g) Pulsed CRISPR: used for temporally controlled gene editing.

[0149] Nanofluorescent signals include but are not limited to the following types: (a) Nanoparticle Signals: used to label and track specific molecules. (b) Quantum Dot Signals: used for high brightness and high stability fluorescent labels. (c) Nanotube Signals: used to enhance signal transmission and detection. (d) Gold Nanoparticle Signals: used to enhance optical contrast. (e) Nanoshell Signals: used for multifunctional diagnosis and treatment. (f) Nanoprobe Signals: used for highly sensitive molecular detection.

[0150] The fluorescence microscopy signal detection model obtained by the construction method of a fluorescence microscopy signal detection model provided by the present invention can also realize a fluorescence microscopy signal detection method based on fluorescence lifetime (Fluorescence Lifetime), and the fluorescence microscopy signals include but are not limited to the following types: (a) Fluorescence Lifetime Imaging Microscope Signal (FLIM): used to measure molecular environment and interactions. (b) Fluorescence Lifetime Sensor Signal (Lifetime Fluorescence Sensors): used to monitor intracellular processes in real time. (c) Time-Resolved Fluorescence Signals: used to improve detection sensitivity. (d) Two-Photon Fluorescence Lifetime Signals: used for deep tissue imaging. (e) Multiphoton Fluorescence Lifetime Signals: used for imaging of complex biological samples.

[0151] The fluorescence microscopy signal detection method provided by the present invention can be used to perform the following operations, including but not limited to: (a) super-resolution imaging. (b) FISH fluorescence signal point detection. (c) denoising, spike signal estimation and neuron segmentation based on neuronal calcium imaging images and videos, two-photon neuronal calcium imaging images and videos, and three-photon and above neuronal calcium imaging images and videos. (d) accurate estimation of any non-uniform fluorescence background.

[0152] The following are specific application examples:

[0153] (1) Using the calibrated PSF image as a POI example, DEPAF can achieve millisecond-level dynamic super-resolution imaging. By optimizing the direct stochastic optical reconstruction microscopy (dSTORM) technology, SMLM imaging is performed on dynamic living cell samples. The turn-on time of the fluorescent molecules is ultra-short, only about 10 milliseconds, and the capture frame rate is as high as about 781 frames per second (FPS). Using DEPAF, super-resolution videos are rendered with a temporal resolution of about 76.8 milliseconds (about 13FPS) and a spatial resolution of about 30 nanometers. These results show that DEPAF achieves millisecond-level dynamic 2D SMLM, making it possible to observe rapid and subtle changes in dynamic subcellular structures.

[0154] (2) Using the fitted Gaussian PSF image as a POI example, DEPAF can achieve high-throughput and high-sensitivity FISH spot signal detection. The application of DEPAF on MERFISH data collected from samples with strong background and high-density labeling is demonstrated. The results show that DEPAF significantly improves the throughput and sensitivity of the MERFISH technology because it is highly robust to highly overlapping spot signals and has a high detection rate for overly bright or dark spot signals. In addition, DEPAF reduces the error rate of each bit, indicating that it has the potential to enable MERFISH to use longer error-resistant codes, thereby enabling the simultaneous detection of more RNA species. Moreover, since DEPAF does not use an explicit threshold to determine the presence of a signal, it can globally detect spot signals in these images with consistent sensitivity after a simple bleaching correction of hybridization images from different rounds. Overall, DEPAF effectively solves the inherent challenges faced by the MERFISH technology in data analysis, greatly promotes the development of MERFISH towards higher throughput and sensitivity, and makes more detailed and comprehensive RNA abundance measurements and spatial organization analysis possible.

[0155] (3) Using the fitted spike signal and standard Gaussian PSF images as POI examples, DEPAF can achieve accurate denoising, highly robust spike signal estimation, and cost-effective segmentation of activated neurons in two-photon neuronal calcium imaging videos in an integrated manner. Considering the data of each pixel along the time axis in the two-photon calcium imaging video as a one-dimensional image, DEPAF is able to denoise the two-photon calcium imaging video and locate overlapping pulse signals in significant noise. At the same time, through the image synthesis module of DEPAF, noise-free signals separated from the background are generated. After reshaping these noise-free signals into a three-dimensional video, accurately denoised two-photon calcium imaging videos that only retain active neuronal signals can be obtained. In addition, DEPAF is able to robustly handle pulse amplitude differences between different neurons and cortical regions by simple normalization based on the median and the first quartile of each one-dimensional image. This advantage is also attributed to DEPAF's avoidance of using explicit thresholds to determine the presence of signals. Based on the denoised video, DEPAF can detect and locate high-density PSFs that constitute active neuronal areas in the spatial dimension. Subsequently, active neurons are extracted and segmented by performing simple DBSCAN clustering and boundary generation on the localized PSF. Comparative experimental results show that in this task, DEPAF outperforms other methods that rely on a large amount of real data as a method that significantly reduces the need for real data. DEPAF unifies the denoising, pulse estimation, and active neuron segmentation tasks in two-photon calcium imaging through a single framework, demonstrating excellent performance while maintaining high practicality. This further demonstrates the universal advantages of DEPAF in a wider range of data modalities and tasks, and also provides potential for two-photon calcium imaging and analysis of finer structures using lower photon budgets in more challenging imaging environments.

[0156] (4) Using constant-value image patches as POI examples, DEPAF is able to achieve accurate estimation of arbitrary heterogeneous backgrounds. For DEPAF, the key to achieving accurate estimation of arbitrary backgrounds lies in the design of background POI examples. First, previous methods generally reveal that the background distribution is concentrated in low frequencies, which suggests that background POI examples should represent some form of low-frequency information. Second, since the background does not contain image contributions of signals or noise, it should be as independent of these as possible to prevent DEPAF from fitting background POIs to signals or noise. Therefore, background POI examples should be orthogonal to various signals and noises. Based on these considerations, the background POI examples are designed as uniform patches of a certain size, which is essentially a small image of a constant value. The reasons are as follows: (1) The inherent size of this uniform patch makes it impossible to fit high-frequency information in the image, so it tends to be used to fit low-frequency information; (2) Due to the extensive periodic fluctuations of the signal and the randomness of the noise, this uniform patch can achieve extensive orthogonality with various signals and noises. These two factors meet the requirements well. In addition, benefiting from DEPAF's ability to fit multiple POI examples simultaneously, by using this background POI example as an additional example besides the signal POI example, DEPAF can simultaneously optimize background estimation and signal detection. This adaptive distinction further improves the accuracy of background estimation. Experimental results show that DEPAF achieves high-precision background estimation in simulated data with different background frequencies and exhibits excellent performance on experimental data. This method provides new perspectives and methods for a wider range of fluorescence background estimation.

[0157] The present invention proposes a universal fluorescence microscopy signal detection method by treating the signal as a POI, which has high versatility, performance and practicality. In terms of millisecond-level dynamic two-dimensional SMLM imaging, the method demonstrates its ability to locate single-shape signals under high signal density and noise conditions. In terms of 3D SMLM imaging, the method demonstrates its ability to simultaneously locate multi-shape signals under the same harsh imaging conditions. It is applied to FISH point signal detection to achieve high-throughput and sensitive MERFISH analysis. To demonstrate its ability to process data across dimensions and other different potential applications, such as one-dimensional signal detection, no prior denoising, and geometric segmentation based on high-density point positioning, this embodiment applies it to two-photon calcium imaging video processing to achieve accurate denoising, highly robust spike signal estimation, and cost-effective segmentation of activated neurons. To prove that it can embed a more generalized prior that is independent of signal shape, this embodiment applies it to fluorescence background estimation to achieve accurate separation of arbitrary non-uniform backgrounds.

[0158] The signal detection step of the above results is achieved by simply modifying the POI examples provided to the DEPAF model and adjusting the sensitivity-related hyperparameters, without the need for any ground truth or domain-specific model of imaging physics. This is possible because the structure and regular features of fluorescence microscopy signals, such as PSFs in SMLM, periodic fluctuations in spike estimation, and specific frequency distributions in fluorescence background estimation, can all be extensively represented using POI examples. By further integrating the powerful representation and utility of self-supervised neural network architectures, the model can be easily adjusted to extract signals based on variable signal features without compromising performance. These cases repeatedly validate the new perspective of signals as patterns and fully demonstrate the high versatility, performance, and practicality of DEPAF in a variety of tasks.

[0159] While maintaining performance, DEPAF can match the prediction speed of traditional fitting-based methods for locating low-density POIs, but outperforms them when using high-density POIs. This is because the computational complexity of the first branch of DEPAF depends only on the size of the image, not the number of POIs in it. Although the second branch involves sequential processing, it does not include any iterative process used in traditional fitting-based methods during the prediction stage.

[0160] The fluorescence microscopy signal detection system provided by the present invention is described below. The fluorescence microscopy signal detection system described below and the fluorescence microscopy signal detection method described above can be referenced to each other.

[0161] See also Figure 3 The present invention also provides a fluorescence microscopy signal detection system, comprising:

[0162] The image receiving module is used to: receive the fluorescence microscope image to be tested;

[0163] A signal detection module is used to obtain, based on the fluorescence microscopy image to be tested, a fluorescence microscopy signal detection model obtained by any of the above-mentioned methods for constructing a fluorescence microscopy signal detection model, a fluorescence microscopy signal detection result of the fluorescence microscopy image to be tested, wherein the fluorescence microscopy signal detection result includes position information and / or amplitude information of the fluorescence microscopy signal.

[0164] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830 and a communication bus 840, wherein the processor 810, the communication interface 820 and the memory 830 communicate with each other via the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to execute any of the above-mentioned methods for constructing a fluorescence microscopy signal detection model and / or the fluorescence microscopy signal detection method.

[0165] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0166] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute any of the above-mentioned methods for constructing a fluorescence microscopy signal detection model and / or the fluorescence microscopy signal detection method.

[0167] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute any of the above-mentioned methods for constructing a fluorescence microscopy signal detection model and / or the fluorescence microscopy signal detection method.

[0168] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0169] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a fluorescence microscopy signal detection model, characterized in that: include: Get the original fluorescence microscope image I ori , where the original fluorescence microscope image I ori The fluorescence microscopy signals in the image have been characterized as a pattern of interest (POI) or a combination thereof; Initializing the deep learning model Leveraging Deep Learning Models The original fluorescence microscope image I ori Convert to Fitting Graph I fit , where the fitting graph I fit Encoded with the original fluorescence microscope image I ori The location information and amplitude of POI in the Fitting Figure I fit Perform group convolution with the preset POI examples to obtain the denoised image I dns ; Maximize the original fluorescence microscopy image I ori And the denoised image I dns The L2 similarity between them, and minimize the fitting graph I fit L1 regularization term to optimize deep learning models Initialize the decoding module Using the decoding module From the fitting diagram I fit The original fluorescence microscope image I is obtained by decoding ori Location information of POI in the From the original fluorescence microscopy image I ori Subtract the denoised image I dns , get the noise image I ns ; The noise image I ns With fitting diagram I fit The sum over the channel dimension Merge to get the noise fitting graph And based on the noise fitting graph The original fluorescence microscope image I is estimated ori The magnitude of the POI; Combined with the estimated original fluorescence microscope image I ori The amplitude of POI, the preset POI example and the decoded original fluorescence microscope image I ori The location information of POI in the image is obtained. syn ; Maximize the synthetic image I syn And the denoised image I dns The L1 similarity between them is used to optimize the decoding module Combined with optimized deep learning model and decoding module The fluorescence microscopy signal detection model is obtained.

2. The method for constructing a fluorescence microscopy signal detection model according to claim 1, characterized in that: The expression of the preset POI example is: In the expression of the preset POI example, represents a set of C pre-acquired POI images in the channel dimension, (x, y) represents the pixel coordinates, represents the minimum value of the c-th POI image, represents the maximum value of the c-th POI image; and / or, Deep Learning Models It is a fully convolutional neural network with U-net structure, a deep learning model The loss function for: Deep Learning Models The loss function middle, I ori represents the original fluorescence microscope image, I dns represents the denoised image, I fit represents the fitting graph, λ represents the regularization strength; and / or, The fitting diagram I fit Perform group convolution with the preset POI examples to obtain the denoised image I dns ,include: According to the first expression, the fitting graph I fit Perform group convolution with the preset POI examples to obtain the denoised image I dns , where the first expression is: In the first expression, I dns represents the denoised image, I fit Indicates a fitted graph, and a colon ":" indicates all x- or y-axis coordinate values ​​in that dimension. represents a two-dimensional convolution, Preset POI examples representing mean sum normalization; and / or, Fitting Figure I fit The sum over the channel dimension The expression is:

3. The method for constructing a fluorescence microscopy signal detection model according to claim 2, characterized in that: The use of decoding module From the fitting diagram I fit The original fluorescence microscope image I is obtained by decoding ori The location information of POI in the , including: Based on the fitting graph I fit The sum over the channel dimension Through the second expression, the pixel area that may contain POI is segmented A collection of; Based on pixel regions that may contain POIs The third expression is used to calculate the centroid of the i-th ROI. Sub-pixel coordinates of the POI As the original fluorescence microscopy image I ori Location information of POI in the The second expression is: In the second expression, θ represents the fit The sum over the channel dimension The segmentation threshold is pre-initialized within the pixel value range of ; The third expression is:

4. The method for constructing a fluorescence microscopy signal detection model according to claim 3, characterized in that: The original fluorescence microscope image I ori Subtract the denoised image I dns , get the noise image I ns ,include: From the original fluorescence microscopy image I ori Subtract the denoised image I dns , get the noise image I ns , I ns =I ori -I dns ; Preset POI example using mean sum normalization The maximum value of the noise image I ns normalize; and / or, The noise image I ns With fitting diagram I fit The sum over the channel dimension Merge to get the noise fitting graph And based on the noise fitting graph The original fluorescence microscope image I is estimated ori The range of POI, including: Through the fourth expression, the normalized noise image I ns With fitting diagram I fit The sum over the channel dimension Merge to get the noise fitting graph By using the fifth expression, each region of interest (ROI) is convolved with the preset POI example corresponding to its z-axis position and the maximum value is calculated to estimate the i-th ROI The amplitude A of the POI contained in i ; Among them, the fourth expression is: In the fourth expression, Represents the normalized noise image I ns The value at pixel coordinate (x,y), Represents the fitting graph I fit The sum over the channel dimension The value at pixel coordinate (x,y), Represents the noise fitting plot The value at pixel coordinate (x,y); The fifth expression is: In the fifth expression, "⊙" represents matrix element-by-element multiplication, and M(·) represents a function that outputs a mask: pixels in the input set are 1, and the rest are 0, at position The preset POI examples obtained at are obtained by interpolation.

5. The method for constructing a fluorescence microscopy signal detection model according to claim 4, characterized in that: The original fluorescence microscope image I obtained by combining the estimation ori The amplitude of POI, the preset POI example and the decoded original fluorescence microscope image I ori The location information of POI in the image is obtained. syn ,include: By using the sixth expression, the estimated original fluorescence microscope image I ori The amplitude of POI, the preset POI example and the decoded original fluorescence microscope image I ori The location information of POI in the image is obtained. syn , where the sixth expression is: In the sixth expression, H and W represent the height and width of the preset POI instance, respectively. and The pixel value of the preset POI example at is obtained by interpolation; and / or, Decoding module The loss function is:

6. The method for constructing a fluorescence microscopy signal detection model according to any one of claims 1 to 5, characterized in that: The fluorescence microscopy signal detection model uses any of the following or any combination of them to optimize model performance: center-aligned upsampling, non-uniform background learning, and POI diffusion.

7. A fluorescence microscopy signal detection method, characterized in that: include: receiving a fluorescence microscope image to be tested; According to the fluorescence microscopy image to be tested, a fluorescence microscopy signal detection result of the fluorescence microscopy image to be tested is obtained by using the fluorescence microscopy signal detection model obtained by the method for constructing a fluorescence microscopy signal detection model described in any one of claims 1 to 6, wherein the fluorescence microscopy signal detection result includes position information and / or amplitude information of the fluorescence microscopy signal.

8. A fluorescence microscopy signal detection system, characterized in that: include: The image receiving module is used to: receive the fluorescence microscope image to be tested; A signal detection module, used to: obtain a fluorescence microscopy signal detection result of the fluorescence microscopy image to be tested based on the fluorescence microscopy signal detection model obtained by the construction method of the fluorescence microscopy signal detection model described in any one of claims 1 to 6, wherein the fluorescence microscopy signal detection result includes position information and / or amplitude information of the fluorescence microscopy signal.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for constructing a fluorescence microscopy signal detection model as described in any one of claims 1 to 6 and / or the fluorescence microscopy signal detection method as described in claim 7 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for constructing a fluorescence microscopy signal detection model as described in any one of claims 1 to 6 and / or the method for detecting fluorescence microscopy signals as described in claim 7 are implemented.

Images