Base cluster positioning and spatial crosstalk correction method, electronic device and storage medium

Abstract

The application provides a base cluster positioning and spatial crosstalk correction method, an electronic device and a storage medium, and applies to the technical field of gene sequencing. The method comprises the following steps: obtaining a plurality of cycle original gray images of base clusters to be processed in the same field of view; performing image filtering processing on the original gray images to obtain target gray images; positioning the positions of the base clusters in the plurality of cycle target gray images according to a preset algorithm to obtain an initial base cluster set; efficiently removing the base clusters through a re-projection method to obtain a target base cluster set; processing the obtained base clusters to form a brightness projection matrix; and processing the brightness projection matrix by using a spatial crosstalk removal convolution kernel, so that the brightness projection matrix after convolution is the brightness of the base cluster after spatial crosstalk correction. The method provided in the application can improve the number of recognized base clusters and the accuracy of base interpretation.

Description

Technical Field

[0001] This invention belongs to the field of gene sequencing technology, specifically relating to a method for base cluster localization and spatial crosstalk correction, electronic equipment, and storage medium. Background Technology

[0002] Gene sequencing technology has undergone three generations of development. The first generation, primarily referring to Sanger sequencing (also known as end-termination sequencing), was invented by Frederick Sanger in 1977. This method, with its high accuracy and reliability, has long been widely used and is considered a pioneering technology in gene sequencing. The second generation, often called high-throughput sequencing, opened a new chapter in genomics research. High-throughput sequencing significantly improved sequencing speed and throughput while substantially reducing sequencing costs, making large-scale genome sequencing economically feasible. The Illumina platform is the most representative example. The third generation of sequencing is characterized by long reads and single-molecule real-time sequencing, providing researchers with unprecedented high-quality data and the ability to analyze complex genomes. Representative platforms for third-generation sequencing include nanopore sequencing and single-molecule real-time sequencing (SMRT) technologies.

[0003] Currently, second-generation sequencing remains the mainstream in the market due to its advantages in data volume and cost brought by high throughput. In second-generation sequencing, the main processing flow involves using image sensors to record base fluorescence signals to obtain sequencing images, and then detecting the base sequence from these images. The core of second-generation gene sequencing algorithms is locating base clusters from the image, extracting and correcting signals, and finally interpreting and assembling the bases into a sequence.

[0004] Currently, numerous methods have been described for the extraction and correction of base signals. Regarding base cluster localization, traditional methods involve segmentation based on thresholds followed by local growth or watershed algorithms. However, these methods are time-consuming, and high-density base clusters are prone to being lost, failing to meet the requirements of high throughput and high accuracy. While many methods exist for base signal crosstalk correction and phase correction, another type of noise signal originates from spatial crosstalk between adjacent base clusters and needs to be removed. This is a significant cause of sequencing errors. Summary of the Invention

[0005] The main objective of this application is to provide a method, electronic device, and storage medium for base cluster localization and spatial crosstalk correction, aiming to solve the technical problems of incomplete base cluster localization and spatial crosstalk of base cluster signals.

[0006] A first aspect of this application provides a method for base cluster localization and spatial crosstalk correction, comprising:

[0007] Acquire original grayscale images with multiple cycles and multiple channels under the same field of view;

[0008] The original grayscale image is subjected to image filtering processing to obtain the target grayscale image;

[0009] Histogram statistics are performed on the grayscale values ​​of the target grayscale image to calculate the mean and standard deviation of the background.

[0010] Using the mean and standard deviation of the background, a threshold for identifying base clusters is calculated;

[0011] The target grayscale image is used to identify base clusters using the threshold to obtain an initial set of base clusters;

[0012] The initial base clusters obtained from the same field of view are deduplicated to obtain the target base cluster set;

[0013] Multiple independent base clusters in the target grayscale image are processed to generate spatially de-crosstalk convolution kernels;

[0014] Extract the grayscale values ​​corresponding to all base cluster positions of the target base cluster as the original brightness of the base cluster, and form a brightness projection matrix;

[0015] The brightness projection matrix is ​​processed using the spatial crosstalk-free convolution kernel, and the resulting brightness projection matrix is ​​the base cluster brightness after spatial crosstalk correction.

[0016] Optionally, the step of performing histogram statistics on the grayscale data of the target grayscale image and calculating the mean and standard deviation of the background includes:

[0017] The grayscale of the target grayscale image is processed to obtain a target grayscale image with continuous grayscale, which at least includes a background;

[0018] Perform statistical analysis of gray-level histograms on gray-level images of targets with continuous gray levels;

[0019] The grayscale histogram is smoothed to obtain a smoothed grayscale histogram;

[0020] The gray level corresponding to the maximum frequency in the smoothed gray-level histogram is used as the estimated mean of the background.

[0021] The left and right boundaries are obtained by multiplying the maximum frequency corresponding to the background by a preset coefficient, and half of the difference between the left and right boundaries is used as the estimated standard deviation of the background.

[0022] A Gaussian distribution function is fitted to the smoothed grayscale histogram to obtain the fitting mean and fitting standard deviation;

[0023] Based on the magnitudes of the estimated mean and the fitted mean, as well as the magnitudes of the estimated standard deviation and the fitted standard deviation, the mean and standard deviation of the background are determined. If the fitted standard deviation is less than the estimated standard deviation, the fitted mean and the fitted standard deviation are taken as the mean and standard deviation of the background; otherwise, the estimated mean and the estimated standard deviation are taken as the mean and standard deviation of the background.

[0024] Optionally, the step of calculating the threshold for identifying base clusters using the mean and standard deviation of the background includes:

[0025] Calculate the threshold used to identify base clusters using the following formula:

[0026] threshold = μ + 3 × δ;

[0027] Where threshold is the threshold used to identify base clusters, μ is the mean of the background, and δ is the standard deviation of the background.

[0028] Optionally, the step of using the threshold to perform base cluster identification on the target grayscale image to obtain an initial base cluster set includes:

[0029] Traverse the target grayscale image and find the local maximum point in the 8-neighborhood as candidate points to obtain the candidate point set;

[0030] Candidate points with a gray level greater than a threshold in the candidate point set are taken as initial base clusters, and the initial base clusters are added to the initial base cluster set.

[0031] Optionally, the step of removing duplicates from the initial base clusters obtained from the same field of view to obtain the target base cluster set includes:

[0032] Obtain the initial base cluster set corresponding to the target grayscale image of four channels in one loop within the same field of view;

[0033] The initial base cluster is deduplicated to obtain a cyclic intermediate base cluster;

[0034] The intermediate base clusters of at least three cycles are gradually deduplicated to obtain the target base cluster.

[0035] Optionally, the step of removing duplicates from the initial base cluster to obtain a cyclic intermediate base cluster includes:

[0036] The base clusters corresponding to the four channels in the initial base cluster set are projected sequentially onto a zero matrix with the same grayscale size as the original. During each projection, it is determined whether the position of the base cluster on the zero matrix is ​​0.

[0037] If so, project the base cluster onto the zero matrix and set the position of the base cluster and its 8-neighborhood to 1;

[0038] By projecting the initial base clusters corresponding to the four channels in a cycle sequentially according to the above steps, an intermediate base cluster of a cycle is obtained.

[0039] Optionally, the step of processing multiple independent base clusters in the target grayscale image to generate a spatially de-crosstalked convolution kernel includes:

[0040] Obtain the positions of multiple independent base clusters in the target grayscale image;

[0041] Calculate the standard deviation of the point spread function of the base clusters based on the gray-scale distribution of multiple base cluster positions within a 5*5 area;

[0042] Generate a Gaussian smoothed convolution kernel based on the standard deviation of the point spread function;

[0043] Normalize the convolution kernel parameters;

[0044] The normalized convolution kernels, except for the center pixel, are negativeed to generate spatially de-crosstalk convolution kernels.

[0045] This application also provides an electronic device, the electronic device comprising: a processor, a memory, a data bus for communication between the processor and the memory, and a program stored in the memory that can be transmitted via the data bus and executed by the processor, wherein when the program is executed by the processor, it implements the steps of the base cluster localization and spatial crosstalk correction method according to any one of claims 1 to 7.

[0046] This application also provides a storage medium, which is a computer-readable storage medium, storing one or more programs that can be executed by one or more processors to implement the steps of the base cluster localization and spatial crosstalk correction method according to any one of claims 1 to 7.

[0047] As can be seen from the above technical solutions, the present invention provides a method for base cluster localization and spatial crosstalk correction, an electronic device, and a storage medium. The method includes: obtaining multiple loops of original grayscale images of base clusters to be processed within the same field of view; performing image filtering on the original grayscale images to obtain a target grayscale image; locating the positions of base clusters in each loop of images according to a preset algorithm to obtain an initial base cluster set; efficiently deduplicating the base clusters using a reprojection method to obtain a target base cluster set; processing the obtained base clusters to form a brightness projection matrix; and processing the brightness projection matrix using a spatial crosstalk-free convolution kernel, the resulting brightness projection matrix being the brightness of the base clusters after spatial crosstalk correction. The method provided in this application can improve the number of identified base clusters and the accuracy of base interpretation. Attached Figure Description

[0048] Figure 1This is a flowchart of the base cluster localization and spatial crosstalk correction method provided in the embodiments of this application.

[0049] Figure 2 is a schematic diagram of the original grayscale image and the processed target grayscale image obtained in an embodiment of this application.

[0050] Figure 3 is an embodiment of this application.

[0051] Figure 4 This is a flowchart illustrating the integration and deduplication of base clusters from multiple channels and multiple cycles in this embodiment of the application.

[0052] Figure 5 is a schematic diagram of the Gaussian smoothing convolution kernel and the spatial crosstalk removal convolution kernel provided in the embodiments of this application.

[0053] Figure 6 is a schematic diagram of the base cluster brightness projection matrix before and after correction provided in the embodiments of this application.

[0054] Figure 7 This is a schematic diagram of the base cluster positioning and spatial crosstalk correction device provided in the embodiments of this application. Detailed Implementation

[0055] The technical solutions in the embodiments of the present invention will be clearly and completely described below. 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.

[0056] Unless otherwise defined, all technical and scientific terms used in this specification shall be interpreted in accordance with their common understanding in the art to which they pertain. The terminology used herein is for the purpose of describing particular embodiments of this application only and is not intended to limit the scope of this application.

[0057] First, we will analyze and explain several key terms used in this application:

[0058] The field of view (FOV) refers to the area on the chip surface corresponding to a single acquisition of base cluster signals by an optical system. The size of the FOV is limited by the objective lens's field of view. Based on the principle of sequence by synthesis (SBS), the process of synthesizing one base at a time is called a cycle; in each cycle, the signal acquisition system collects the signal of the synthesized base to identify the base category. Therefore, during sequencing, base clusters on the chip surface within one FOV will synthesize fluorescent dNTPs multiple times before their fluorescence signals are acquired, resulting in a raw grayscale image containing the fluorescence signals of the base clusters across multiple cycles within one FOV.

[0059] A multi-channel image from a field of view (FOV) corresponds to four wavelengths that can pass through filters in an optical system. These wavelengths correspond to the main frequency bands of the fluorescent groups carried by the four bases and are recorded. Therefore, an FOV needs to acquire four images simultaneously. Each of the four bases—A (adenine), C (cytosine), G (guanine), and T (thymine)—carries a specific fluorescent group. When excited by a laser, they emit fluorescence in a specific wavelength band, which is separated and recorded by the filters. The recorded signals help distinguish the types of bases. Multi-channel imaging helps to locate all base clusters and extract the signals of the base clusters from the multi-channel images for base interpretation.

[0060] Due to light diffraction and the resolution limits of optical systems, point sources naturally diffuse. Gaussian diffusion functions are the most commonly used, and point diffusion of base clusters is similar. Spatial crosstalk between adjacent base clusters refers to the interference between signals of adjacent base clusters in each channel when they are too close together in a high-density environment. The degree of interference is close to that of point diffusion, and in severe cases, fusion and overlap can occur, making them difficult to distinguish and resulting in incomplete detection of base clusters.

[0061] Due to the urgent need for high throughput in next-generation sequencing, improving the accuracy and completeness of base cluster localization has become a pressing issue. Therefore, the main objective of this application is to provide a method, electronic device, and storage medium for base cluster localization and spatial crosstalk correction, aiming to solve the technical problems of incomplete base cluster localization and spatial crosstalk in base cluster signals.

[0062] The main objective of this application is to provide a method for base cluster localization and spatial crosstalk correction, referencing... Figure 1 The embodiments of this application include the following steps:

[0063] S101, acquire the original grayscale image with multiple cycles and multiple channels under the same field of view.

[0064] In this step, a base cluster of a field of view (FOV) is excited by a laser after binding with fluorescent dNTPs. Its emitted light is separated by a specific optical path and passes through a color filter, and is finally recorded by a monochrome camera to obtain a multi-channel raw grayscale image to be processed.

[0065] S102, perform image filtering processing on the original grayscale image to obtain the target grayscale image.

[0066] In this step, a suitable standard deviation is first set, and the convolution kernel is calculated based on the Gaussian-Laplace function; then the convolution kernel is applied to the original grayscale image to obtain the target grayscale image.

[0067] The Laplacian of Gaussian (LoG) operator described above is a commonly used edge detection operator in image processing, combining Gaussian blur and the Laplacian operator. This method is used to enhance the edge features of an image by first applying a Gaussian blur and then applying the Laplacian operator to detect edges. Another application of this method is blob detection, and the standard deviation of this operator has a selective effect on blobs of different scales, making it particularly suitable for fluorescence images of base clusters. Choosing an appropriate standard deviation helps to accommodate the point diffusion of base clusters and enhance the base cluster signal while removing background. The Gaussian-Laplacian function is shown below:

[0068]

[0069] Where (x, y) are the x and y coordinates in the two-dimensional space. Let σ be the Gaussian-Lapsian function, and σ be the standard deviation.

[0070] In some instances, the standard deviation σ of the LoG operator is set to 0.8, and a 7*7 convolution kernel is generated. After convolving with the original grayscale image, an enhanced target grayscale image with the background removed is obtained. Figure 2 is an example, where Figure 2(a) is the original grayscale image and Figure 2(b) is the target grayscale image after filtering.

[0071] S103, perform histogram statistics on the grayscale of the target grayscale image, and calculate the mean and standard deviation of the background.

[0072] In one feasible implementation, S103 includes the following steps:

[0073] S1031, the grayscale of the target grayscale image is scaled to obtain a target grayscale image with continuous grayscale, which at least includes a background;

[0074] S1032, Perform statistical analysis of the gray-level histogram for a target gray-level image with continuous gray levels;

[0075] S1033, Smooth the grayscale histogram to obtain a smoothed grayscale histogram;

[0076] S1034, obtain the gray level corresponding to the maximum frequency in the smoothed gray-level histogram as the estimated mean of the background;

[0077] S1035, multiply the maximum frequency corresponding to the background by a preset coefficient to obtain the left and right boundaries, and use half of the difference between the left and right boundaries as the estimated standard deviation of the background.

[0078] S1036, Fit a Gaussian distribution function to the smoothed gray-level histogram to obtain the mean and standard deviation of the fit;

[0079] S1037, take the set with the smaller standard deviation between the estimated standard deviation of the background and the fitted standard deviation as the mean and standard deviation of the background.

[0080] In this embodiment, the target grayscale image typically includes a background and light spots, with the background area making up the majority. The purpose of step S1031 is to ensure the grayscale of the image is continuous. This is because some cameras amplify the grayscale when saving grayscale images; if this is not considered, the resulting histogram will be discontinuous, making feature calculation and fitting impossible. The interval width of the statistical histogram is set to 1 to construct the grayscale histogram. The horizontal axis of the grayscale histogram represents the grayscale level, and the vertical axis represents the frequency or number of occurrences of each grayscale level.

[0081] In step S1033 of this embodiment, the smoothing method can be selected as a Savitzky-Golay filter. It is a one-dimensional data filter designed to smooth data while maintaining a good fit, and is particularly suitable for reducing noise without significantly changing the shape of the signal.

[0082] In steps S1034 and S1035 of this embodiment, it is necessary to calculate the initial estimates of the background mean and standard deviation based on the features. As an example, Figure 3(a) shows a schematic diagram of the LoG filtered image, which can be divided into three parts: the background area, the DNC area, and the DNC edge area; Figure 3(b) shows the smoothed grayscale histogram and the fitted Gaussian distribution curve obtained from the LoG filtered image shown in Figure 3(a); since the background area accounts for the vast majority of the image, the interval with the highest frequency in the smoothed grayscale histogram is taken as the background mean, and the grayscale value corresponding to the highest frequency in this interval is the initial estimate of the background mean; and the maximum frequency is multiplied by the coefficient e^-0.5 as the threshold to obtain the left and right boundaries, and half of the difference between the left and right boundaries is taken as the initial estimate of the background standard deviation.

[0083] S104, using the mean and standard deviation of the background, calculate the threshold for identifying base clusters.

[0084] In this embodiment, the purpose of setting this threshold is to eliminate the interference of background noise. Step S104 specifically includes:

[0085] Calculate the threshold used to identify base clusters using the following formula:

[0086] threshold = μ + 3 × δ;

[0087] Where threshold is the threshold used to identify base clusters, μ is the mean of the background, and δ is the standard deviation of the background.

[0088] S105, use the threshold to identify base clusters in the target grayscale image to obtain an initial base cluster set.

[0089] In this step, only candidate points with grayscale values ​​greater than a threshold are considered base clusters and added to the initial base cluster set for that channel within the field of view (FOV). The specific steps are as follows:

[0090] S1051, Traverse the target grayscale image and find the local maximum point in the 8-neighborhood as candidate points to obtain a candidate point set;

[0091] S1052, select candidate points with gray values ​​greater than a threshold from the candidate point set as initial base clusters, and add the initial base clusters to the initial base cluster set.

[0092] In this embodiment, the purpose of threshold filtering is to assume that the grayscale of the background pixels follows a normal distribution. By using threshold as the threshold, more than 99.73% of noise points can be filtered out, ensuring that the obtained local maximum points are reliable bright points.

[0093] S106, deduplication is performed on the initial base clusters obtained from the same field of view to obtain the target base cluster set.

[0094] In this embodiment, S106 removes duplicates from the initial base clusters obtained from the multi-channel sample, using a projection method to avoid base clusters that are too close together, including the following steps:

[0095] S1061, Obtain the initial base cluster set corresponding to the target grayscale image of four channels in one loop of the same field of view;

[0096] S1062, The initial base cluster is deduplicated to obtain a cyclic intermediate base cluster;

[0097] S1063, gradually remove duplicates from the intermediate base clusters of at least three cycles to obtain the target base cluster.

[0098] In this embodiment, the method used in S1062 can be called the projection deduplication method, and the detailed steps are as follows:

[0099] S10621, the base clusters corresponding to the four channels in the initial base cluster set are projected sequentially onto a zero matrix with the same grayscale size as the original. During each projection, it is determined whether the position of the base cluster on the zero matrix is ​​0.

[0100] S10622, if yes, then project the base cluster onto the zero matrix and set the position of the base cluster and its 8-neighborhood to 1; if not 0, it means that the position has already been projected, then the base cluster will not be projected. This step can set the 8-neighborhood of the base cluster to 1 to avoid the repeated acquisition of subsequent base clusters with a distance < 2.

[0101] S10623, the initial base clusters corresponding to the four channels in a cycle are projected sequentially according to the above steps to obtain the intermediate base clusters of a cycle.

[0102] Reference Figure 4 The schematic diagram shows that, in step S1061, an initial base cluster is obtained from the target grayscale image of channel A in the first loop, while the other three are obtained from the target grayscale image of channel CGT; the initial base cluster is obtained from the target grayscale image of channel C in the second loop, while the other three are obtained from the target grayscale image of channel AGT. Step S1062 obtains intermediate base cluster 1 from the first loop and intermediate base cluster 2 from the second loop. Step S1063 removes duplicates from intermediate base clusters 1, 2, ... to obtain the target base cluster.

[0103] Optionally, the step of deduplicating at least three cycle intermediate base clusters to obtain the target base cluster is consistent with the step of obtaining one cycle intermediate base cluster, referring to the steps above.

[0104] S107, process multiple independent base clusters in the target grayscale image to generate a spatially de-crosstalk convolution kernel.

[0105] In this embodiment, step S107 includes:

[0106] S1071, Obtain the positions of multiple independent base clusters in the target grayscale image;

[0107] S1072, calculate the standard deviation of the point diffusion function of the base clusters based on the gray-scale distribution of multiple base cluster positions within a 5*5 range;

[0108] S1073, Generate a Gaussian smoothed convolution kernel based on the standard deviation of the point spread function;

[0109] S1074, normalize the convolution kernel parameters;

[0110] S1075 performs negative processing on the normalized convolution kernel except for the center pixel to generate a spatially de-crosstalk convolution kernel.

[0111] In step S1071 of this embodiment, independent base clusters are obtained from the target grayscale images of different channels. The purpose is to eliminate the influence of base clusters that are too close together, so that the brightness distribution of the base clusters does not meet the Gaussian point diffusion model.

[0112] In step S1072 of this embodiment, a Gaussian distribution is fitted to the brightness distribution of individual base clusters, and the standard deviation reflects the degree of influence. It is worth noting that in some embodiments, to accommodate different channel wavelengths, different point spread functions are fitted to different channels separately. The mean of the obtained standard deviations of the base clusters is used as the standard deviation of the point spread function for one channel.

[0113] In step S1073 of this embodiment, a despatialization convolution kernel is generated based on the point spread standard deviation. First, a Gaussian smoothing convolution kernel is generated based on the standard deviation (std), with a size of size*size, where size is an odd number and satisfies size>=6*std, and std is the standard deviation.

[0114] In step S1074 of this embodiment, the convolution kernel parameters are normalized. The normalization formula is as follows: Where max(Kernel) is the largest element of the convolution kernel, and Kernel is the convolution kernel.

[0115] In step S1075 of this embodiment, all elements of the convolution kernel except the center pixel are negative. The negative convolution kernel parameters are used to remove the influence of adjacent base clusters. The value of the parameter reflects the degree of influence of adjacent base clusters caused by distance. The generated convolution kernel is shown in Figure 5. Figure 5(a) shows the Gaussian smoothing convolution kernel, and Figure 5(b) shows the spatial crosstalk removal convolution kernel.

[0116] S108, extract the grayscale values ​​corresponding to all base cluster positions of the target base cluster as the original brightness of the base cluster, and form a brightness projection matrix.

[0117] In this step, the brightness of the base clusters is reprojected onto a 0 matrix of the same size as the target grayscale image. The position of the base clusters is the pixel-level precision of the identified position. The resulting base cluster brightness projection matrix is ​​shown in Figure 6(a).

[0118] S109, the brightness projection matrix is ​​processed using the spatial crosstalk removal convolution kernel, and the convolved brightness projection matrix is ​​the base cluster brightness after spatial crosstalk correction.

[0119] In this step, the base cluster brightness projection matrix is ​​processed using the aforementioned despatialized convolution kernel to obtain the convolved base cluster brightness projection matrix, as shown in Figure 6(b). As can be seen from the figure, the pixels surrounding the base clusters exhibit a black negative value state, which is the effect of removing the influence of the base clusters. Closely located base clusters will mutually remove a certain proportion of their influence, thereby improving signal purity and increasing the probability of correct interpretation.

[0120] This application provides a method for base cluster localization and spatial crosstalk correction. The method includes: obtaining multiple loops of original grayscale images of base clusters to be processed within the same field of view; performing image filtering on the original grayscale images to obtain an enhanced image with background removed; locating the positions of base clusters in each loop of images according to a preset algorithm; efficiently deduplicating base clusters using a reprojection method to obtain sufficient base cluster positions; extracting brightness signals from the obtained base clusters and correcting the spatial crosstalk signals of adjacent base clusters. This application can improve the number of identified base clusters and the accuracy of base interpretation.

[0121] Please see Figure 7 This application also provides a base cluster localization and spatial crosstalk correction device, which can realize the above-mentioned base cluster localization and spatial crosstalk removal method. The device includes:

[0122] Image acquisition module 100 is used to acquire original grayscale images with multiple cycles and multiple channels under the same field of view;

[0123] Image processing module 200 is used to perform image filtering processing on the original grayscale image to obtain a target grayscale image;

[0124] The image background estimation module 300 is used to perform histogram statistics on the gray levels of the target grayscale image and calculate the mean and standard deviation of the background.

[0125] The calculation module 300 is used to calculate a threshold for identifying base clusters using the mean and standard deviation of the background.

[0126] The base cluster localization module 400 is used to identify base clusters in the target grayscale image using the threshold to obtain an initial base cluster set.

[0127] The base cluster deduplication module 500 is used to deduplicat initial base clusters obtained from the same field of view to obtain the target base cluster set;

[0128] The kernel generation module 600 is used to process multiple independent base clusters in the target grayscale image to generate spatially de-crosstalked convolution kernels;

[0129] The brightness extraction module 700 is used to extract the grayscale corresponding to the position of all base clusters in the target base cluster set as the original brightness of the base clusters, and form a brightness projection matrix.

[0130] The brightness correction module 800 is used to process the brightness projection matrix using the spatial crosstalk removal convolution kernel. The convolved brightness projection matrix is ​​the base cluster brightness after spatial crosstalk correction.

[0131] The specific implementation of this base cluster positioning and spatial crosstalk correction device is basically the same as the specific implementation of the base cluster positioning and spatial crosstalk correction method described above, and will not be repeated here.

[0132] This application also proposes an electronic device, which includes a processor, a memory, a data bus for communication between the processor and the memory, and an executable program. The executable program implements the aforementioned base cluster localization and spatial crosstalk correction functions on the electronic device. This electronic device can include any terminal such as a personal computer, server, industrial control unit, mobile phone, tablet computer, or vehicle-mounted device.

[0133] This application also provides a storage medium, which is a computer-readable storage medium, characterized in that the storage medium stores one or more programs, which can be executed by one or more processors to implement the steps of the base cluster localization and spatial crosstalk correction method.

[0134] The electronic device in this embodiment may include the following external devices: input devices such as touch screens, keyboards, mice, industrial cameras, microphones, and various sensors; output devices such as displays, speakers, buzzers, and vibrators; external storage devices such as solid-state drives, hard disk drives, and optical discs; and communication devices. The communication devices enable the electronic device to exchange data with other devices via wireless or wired connections. Although the figures show electronic devices with various systems, it should be noted that actual implementations are not required to include all the systems shown. Furthermore, these devices may selectively include more or fewer systems.

[0135] In addition to the most widely used Central Processing Unit (CPU), processors also include other computing-capable devices such as Graphics Processing Units (GPUs), microprocessors, Application-Specific Integrated Circuits (ASICs), and Field-Programmable Gate Arrays (FPGAs). Processors can execute program instructions and are the foundation for implementing the methods of this embodiment.

[0136] The memory may include read-only memory (ROM), random access memory (RAM), and external memory. ROM is mainly used for permanent data storage, while RAM is used for temporary data storage requiring fast read and write access. ROM, RAM, and the processor all communicate via a data bus. When executing the method of the embodiment, various programs and data required are read into RAM for temporary storage and processed by the processor line by line via the data bus. In addition, input / output interfaces are also connected to the data bus.

[0137] Furthermore, the program code contained on this storage medium can be transmitted via various transmission methods, including but not limited to wires, optical cables, and radio frequency signals. The program instructions stored on this medium enable the electronic device to extract base clusters from the target grayscale image during execution, extract the base cluster signals, and perform spatial crosstalk correction to obtain the desired detection results.

[0138] The main objective of this application is to provide a method for base cluster localization and spatial crosstalk correction, an electronic device, and a storage medium. This application achieves this by obtaining a single field of view. The system obtains multiple loops of original grayscale images from a Field of View (FOV), typically 3-5 loops. Image filtering is applied to these initial grayscale images to obtain a target grayscale image. Histogram statistics are then performed on the grayscale values ​​of the target grayscale image to obtain a grayscale histogram. Next, the background portion of the grayscale histogram is fitted to obtain the mean and standard deviation of the background, which are used as thresholds for identifying base clusters. A preset base cluster localization algorithm is applied to the multi-loop, multi-channel target grayscale image of a FOV, and base cluster identification is performed using the threshold to obtain an initial set of base clusters. The base clusters obtained from the multi-channel target grayscale image of a FOV are deduplicated to obtain an intermediate set of base clusters for one loop. The intermediate set of base clusters from the 3-5 loops is deduplicated to obtain the final target set of base clusters. The target grayscale image is then processed using a preset method to extract the multi-channel grayscale signals of all base cluster positions in the target set of base clusters as the original brightness of the base clusters. A preset algorithm is then applied to the obtained original brightness of the base clusters to remove spatial crosstalk. This process can reduce the loss of base clusters, and the reprojection convolution method can quickly and effectively remove the noise signals of adjacent base clusters, improve the purity of bases, and reduce the probability of base misclassification.

[0139] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A method for base cluster localization and spatial crosstalk correction, characterized in that, include: Acquire original grayscale images with multiple cycles and multiple channels under the same field of view; The original grayscale image is subjected to image filtering processing to obtain the target grayscale image; Histogram statistics are performed on the grayscale values ​​of the target grayscale image to calculate the mean and standard deviation of the background. Using the mean and standard deviation of the background, a threshold for identifying base clusters is calculated; The target grayscale image is used to identify base clusters using the threshold to obtain an initial set of base clusters; The initial base clusters obtained from the same field of view are deduplicated to obtain the target base cluster set; Multiple independent base clusters in the target grayscale image are processed to generate spatially de-crosstalk convolution kernels; Extract the grayscale values ​​corresponding to all base cluster positions of the target base cluster as the original brightness of the base cluster, and form a brightness projection matrix; The brightness projection matrix is ​​processed using the spatial crosstalk-free convolution kernel, and the resulting brightness projection matrix is ​​the base cluster brightness after spatial crosstalk correction.

2. The method for base cluster localization and spatial crosstalk correction according to claim 1, characterized in that, The steps of performing histogram statistics on the grayscale data of the target grayscale image and calculating the mean and standard deviation of the background include: The grayscale of the target grayscale image is processed to obtain a target grayscale image with continuous grayscale, which at least includes a background; Perform statistical analysis of gray-level histograms on gray-level images of targets with continuous gray levels; The grayscale histogram is smoothed to obtain a smoothed grayscale histogram; The gray level corresponding to the maximum frequency in the smoothed gray-level histogram is used as the estimated mean of the background. Multiply the maximum frequency by a preset coefficient to obtain the left and right boundaries, and use half of the difference between the left and right boundaries as the estimated standard deviation of the background. A Gaussian distribution function is fitted to the smoothed grayscale histogram to obtain the fitting mean and fitting standard deviation; Based on the magnitudes of the estimated mean and the fitted mean, as well as the magnitudes of the estimated standard deviation and the fitted standard deviation, the mean and standard deviation of the background are determined. If the fitted standard deviation is less than the estimated standard deviation, the fitted mean and the fitted standard deviation are taken as the mean and standard deviation of the background; otherwise, the estimated mean and the estimated standard deviation are taken as the mean and standard deviation of the background.

3. The method for base cluster localization and spatial crosstalk correction according to claim 1, characterized in that, The step of calculating the threshold for identifying base clusters using the mean and standard deviation of the background includes: Calculate the threshold used to identify base clusters using the following formula: threshold = μ + 3 × δ; Where threshold is the threshold used to identify base clusters, μ is the mean of the background, and δ is the standard deviation of the background.

4. The method for base cluster localization and spatial crosstalk correction according to claim 1, characterized in that, The step of using the threshold to identify base clusters in the target grayscale image to obtain an initial base cluster set includes: Traverse the target grayscale image and find the local maximum point in the 8-neighborhood as candidate points to obtain the candidate point set; Candidate points with a gray level greater than a threshold in the candidate point set are taken as initial base clusters, and the initial base clusters are added to the initial base cluster set.

5. The method for base cluster localization and spatial crosstalk correction according to claim 1, characterized in that, The steps for removing duplicates from initial base clusters obtained from the same field of view to obtain the target base cluster set include: Obtain the initial base cluster set corresponding to the target grayscale image of four channels in one loop within the same field of view; The initial base cluster is deduplicated to obtain a cyclic intermediate base cluster; The intermediate base clusters of at least three cycles are gradually deduplicated to obtain the target base cluster.

6. The method for base cluster localization and spatial crosstalk correction according to claim 5, characterized in that, The step of removing duplicates from the initial base cluster to obtain a cyclic intermediate base cluster includes: The base clusters corresponding to the four channels in the initial base cluster set are projected sequentially onto a zero matrix with the same grayscale size as the original. During each projection, it is determined whether the position of the base cluster on the zero matrix is ​​0. If so, project the base cluster onto the zero matrix and set the position of the base cluster and its 8-neighborhood to 1; By projecting the initial base clusters corresponding to the four channels in a cycle sequentially according to the above steps, an intermediate base cluster of a cycle is obtained.

7. The method for base cluster localization and spatial crosstalk correction according to claim 1, characterized in that, The steps of processing multiple independent base clusters in the target grayscale image to generate spatially de-crosstalk convolution kernels include: Obtain the positions of multiple independent base clusters in the target grayscale image; Calculate the standard deviation of the point spread function of the base clusters based on the gray-scale distribution of multiple base cluster positions within a 5*5 area; Generate a Gaussian smoothed convolution kernel based on the standard deviation of the point spread function; Normalize the convolution kernel parameters; The normalized convolution kernels, except for the center pixel, are negativeed to generate spatially de-crosstalk convolution kernels.

8. An electronic device, characterized in that, The electronic device includes: a processor, a memory, a data bus for communication between the processor and the memory, and a program stored in the memory that can be transmitted via the data bus and executed by the processor, wherein when the program is executed by the processor, it implements the steps of the base cluster localization and spatial crosstalk correction method according to any one of claims 1 to 7.

9. A storage medium, said storage medium being a computer-readable storage medium, characterized in that, The storage medium stores one or more programs, which can be executed by one or more processors to implement the steps of the base cluster localization and spatial crosstalk correction method according to any one of claims 1 to 7.

Images