A method and system for recognizing and positioning colloidal gold particles in cryo-em images

Abstract

The disclosure provides a kind of recognition positioning method and system of colloidal gold particles in cryo-EM image, the scheme includes: obtaining the electron microscope image to be identified, and carries out corresponding pretreatment;Wavelet transform is handled to the image after pretreatment based on Atrous algorithm, obtains corresponding wavelet image;For the wavelet image obtained, in turn, binaryzation processing, noise point removal and clustering operation are carried out, obtain the corresponding several clusters of colloidal gold particles in binary image;From the several clusters, the cluster satisfying the preset constraint is screened out, and the average template is obtained based on the obtained cluster satisfying the preset constraint;Template matching is carried out using the average template and the several clusters respectively, and the cluster whose matching result is lower than the preset threshold is filtered out, and the remaining cluster is used as the candidate colloidal gold particle target;The position and size estimation of the candidate colloidal gold particle target obtained are obtained by corresponding the original image.

Description

Technical Field

[0001] This disclosure belongs to the field of image recognition technology, and in particular relates to a method and system for identifying and locating colloidal gold particles in cryo-electron microscopy images. Background Technology

[0002] The statements in this section are merely background information relating to this disclosure and do not necessarily constitute prior art.

[0003] Cryo-electron tomography (Cryo-ET) is a technique for visualizing submicroscopic structures within biological cells and is also an application of cryo-transmission electron microscopy (Cryo-TEM). Cryo-ET places biomolecules on a rotating sample stage, and through the rotating stage, obtains a projection sequence of the sample from multiple angles. When the stage rotation angle is too large, the electron beam's penetration path in the sample becomes longer, resulting in almost no useful information being obtained in the projected image. Therefore, in Cryo-ET, the stage rotation range is mostly between -60° and 60°. However, this leads to the loss of projection information for samples with large rotation angles, resulting in incomplete information in the imaging results, commonly known as missing wedges. Secondly, because electron beams can easily damage biological samples, low-dose electron beams are used for projection. Due to instrument and biological sample limitations, the signal-to-noise ratio (SNR) in the projected images from all directions is very low, typically less than 0.1. Therefore, direct reconstruction using the projected image sequence is difficult, and images of more similar biomolecules must be calibrated and averaged to achieve an acceptable reconstructed structure. A crucial step in this process is to pick up as many colloidal gold particles as possible from the image.

[0004] Colloidal gold particles are added to biological samples as reference points for image alignment operations, and often appear as dark spots in electron microscope images. By identifying and locating colloidal gold particles on each projection image, alignment between the projection images is achieved, thereby obtaining the reconstruction of Cryo-ET. In Cryo-ET, the identification and localization of colloidal gold particles is crucial and may very well become a bottleneck limiting the development of Cryo-ET.

[0005] The inventors discovered that in the early stages, researchers mostly selected colloidal gold particles manually, which resulted in enormous manpower and time costs in the particle selection phase. With the continuous advancement of algorithm research, some semi-automated and automated colloidal gold particle picking methods have emerged. Traditional particle picking methods can be mainly divided into template-based methods and feature-based methods. Template matching algorithms are based on linear image formation models and have significant theoretical application value. The main drawbacks of this algorithm are: due to template limitations, it can only identify biomolecules with known structures; the identification result depends on the template selection; and it is prone to misidentifying the background. Secondly, this algorithm is easily affected by the state of the dataset (lighting, noise, etc.). Feature-based recognition methods do not require pre-prepared template images; they only need a small portion of local or globally significant features of the particle image, such as edge features. The main drawback of this algorithm is that it may be difficult to extract highly identifiable features when processing low-contrast, low-signal-to-noise ratio images.

[0006] With the advent of deep learning methods, neural networks have also been used for the identification and localization of colloidal gold particles. Using neural network methods for colloidal gold particle identification requires inputting a large dataset labeled with colloidal gold particles into the model to ensure that the trained model can accurately identify and locate colloidal gold particles in images. Compared to traditional methods, deep learning methods can achieve stable and fast identification and localization of colloidal gold particles. Some neural networks exhibit excellent properties in target recognition and localization, such as recognition algorithms based on convolutional neural networks (CNN), U-Net, and YOLO. However, although deep learning methods can achieve stable and fast identification after model training, they require a long training time and a large amount of labeled dataset during training. Summary of the Invention

[0007] To address the aforementioned problems, this disclosure provides a method and system for identifying and locating colloidal gold particles in cryo-electron microscopy images. The scheme utilizes wavelet transform to initially filter unnecessary information, narrow the scope of template matching, and generate an average colloidal gold particle template. Then, template matching is used to select colloidal gold particles, thereby solving the problems of template selection and background misidentification in traditional template matching algorithms, as well as the problem of poor recognition performance of feature-based recognition methods on low-contrast, low-signal-noise-ratio datasets.

[0008] According to a first aspect of the present disclosure, a method for identifying and locating colloidal gold particles in cryo-electron microscopy images is provided, comprising:

[0009] Acquire the electron microscope image to be identified and perform appropriate preprocessing;

[0010] The wavelet transform based on the Atrous algorithm is used to process the preprocessed image to obtain the corresponding wavelet image;

[0011] The obtained wavelet image is sequentially binarized and noise points are removed. Then, each target to be identified in the binarized image after noise removal is clustered to obtain several clusters corresponding to the colloidal gold particles in the binarized image.

[0012] Clusters that satisfy preset constraints are selected from the aforementioned clusters, and an average template is obtained based on the obtained clusters that satisfy the preset constraints.

[0013] The average template is used to perform template matching with the several clusters respectively, and clusters with matching results lower than a preset threshold are filtered out. The remaining clusters are used as candidate colloidal gold particle targets.

[0014] The obtained candidate colloidal gold particles are matched with the original image to obtain the estimated position and size of the candidate colloidal gold particles.

[0015] Furthermore, the wavelet transform based on the Atrous algorithm is specifically as follows:

[0016] Step 1: Process the input image With transformation kernel Perform convolution processing to obtain the convolved image. ;

[0017] Step 2: Transform kernel Perform zero-padding operation between elements;

[0018] Step 3: Put and Perform the difference operation to obtain the wavelet image at this scale;

[0019] Step 4: Determine if the preset scale has been reached. If not, then... Assign to If yes, return to step 1; otherwise, proceed to step 5.

[0020] Step 5: Output the wavelet image at the last scale and the wavelet image at the first scale, wherein the wavelet image at the first scale is used for threshold calculation of wavelet image binarization.

[0021] Furthermore, the preset constraint specifically means that the pixel value of the center point of the cluster is lower than a preset first threshold, and the number of pixels in the cluster is less than twice the area corresponding to the initial radius value.

[0022] Furthermore, the step of matching the obtained candidate colloidal gold particle targets with the original image specifically involves: performing Gaussian fitting on the original image to obtain the center coordinates of each candidate colloidal gold particle; calculating the diameter of the cluster corresponding to each candidate colloidal gold particle; and obtaining a radius estimate of the candidate colloidal gold particle based on the diameter.

[0023] Furthermore, the step of performing template matching with the average template and the plurality of clusters respectively specifically involves calculating the normalized cross-correlation value between the average template and the candidate particles.

[0024] Furthermore, the clustering operation specifically involves taking the set of pixels corresponding to each target to be identified as a cluster, and performing clustering based on the MeanShift clustering method to obtain the coordinate set of each cluster and the center coordinates of the cluster.

[0025] Furthermore, the corresponding preprocessing specifically involves: performing an average filtering process on the electron microscope image to be identified to obtain a smooth electron microscope image.

[0026] According to a second aspect of the present disclosure, a system for identifying and locating colloidal gold particles in cryo-electron microscopy images is provided, comprising:

[0027] The data acquisition unit is used to acquire electron microscope images to be identified and perform corresponding preprocessing.

[0028] The wavelet transform unit is used to process the preprocessed image using wavelet transform based on the Atrous algorithm to obtain the corresponding wavelet image;

[0029] The wavelet image processing unit is used to perform binarization and noise point removal on the obtained wavelet image in sequence, and to perform clustering operation on each target to be identified in the binarized image after noise point removal to obtain several clusters corresponding to colloidal gold particles in the binarized image.

[0030] An average template acquisition unit is used to filter out clusters that satisfy preset constraints from the plurality of clusters, and obtain an average template based on the obtained clusters that satisfy the preset constraints.

[0031] The colloidal gold particle target acquisition unit is used to perform template matching between the average template and the plurality of clusters respectively, filter out clusters whose matching results are lower than a preset threshold, and use the remaining clusters as candidate colloidal gold particle targets.

[0032] The localization unit is used to match the obtained candidate colloidal gold particle targets with the original image to obtain the position and size estimate of the candidate colloidal gold particles.

[0033] According to a third aspect of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and running on the memory. When the processor executes the program, it implements the method for identifying and locating colloidal gold particles in a cryo-electron microscopy image.

[0034] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the method for identifying and locating colloidal gold particles in a cryo-electron microscopy image.

[0035] Compared with the prior art, the beneficial effects of this disclosure are:

[0036] (1) This disclosure provides a method and system for identifying and locating colloidal gold particles in cryo-electron microscopy images. The scheme uses wavelet transform to initially filter unnecessary information, narrow the scope of template matching, and generate an average colloidal gold particle template. Then, template matching is used to select colloidal gold particles to solve the problems of template selection and background misidentification in traditional template matching algorithms, as well as the problem of poor recognition effect of feature-based recognition methods on low-contrast and low-signal-noise ratio datasets. It effectively improves the shortcomings of single template matching algorithms and feature-based recognition algorithms, reduces the misidentification of background information by template matching, and can automatically generate the required average template.

[0037] (2) Compared with the colloidal gold particle identification and localization model based on deep learning, the scheme described in this disclosure does not require a large number of labeled electron microscope images for training, which greatly saves the manpower and material resources required for the colloidal gold identification and localization model.

[0038] (3) Compared with the general wavelet method, the scheme described in this disclosure adopts the wavelet method based on the Atrous algorithm, so that the image maintains the same size before and after the wavelet transform, thus solving the bias problem caused by image scaling.

[0039] Advantages of this disclosure in additional aspects will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this disclosure. Attached Figure Description

[0040] The accompanying drawings, which form part of this disclosure, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.

[0041] Figure 1 This is a flowchart of the identification stage in a method for identifying and locating colloidal gold particles in a cryo-electron microscopy image as described in an embodiment of this disclosure.

[0042] Figure 2This is a flowchart of the localization stage in a method for identifying and locating colloidal gold particles in a cryo-electron microscopy image as described in an embodiment of this disclosure.

[0043] Figure 3 This is a flowchart of the wavelet transform based on the Atrous algorithm described in the embodiments of this disclosure. Detailed Implementation

[0044] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0045] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.

[0046] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this disclosure. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0047] Where there is no conflict, the embodiments and features described herein can be combined with each other.

[0048] Example 1:

[0049] The purpose of this embodiment is to provide a method for identifying and locating colloidal gold particles in cryo-electron microscopy images.

[0050] A method for identifying and locating colloidal gold particles in cryo-electron microscopy images, comprising:

[0051] Acquire the electron microscope image to be identified and perform appropriate preprocessing;

[0052] The wavelet transform based on the Atrous algorithm is used to process the preprocessed image to obtain the corresponding wavelet image;

[0053] The obtained wavelet image is sequentially binarized and noise points are removed. Then, each target to be identified in the binarized image after noise removal is clustered to obtain several clusters corresponding to the colloidal gold particles in the binarized image.

[0054] Clusters that satisfy preset constraints are selected from the aforementioned clusters, and an average template is obtained based on the obtained clusters that satisfy the preset constraints.

[0055] The average template is used to perform template matching with the several clusters respectively, and clusters with matching results lower than a preset threshold are filtered out. The remaining clusters are used as candidate colloidal gold particle targets.

[0056] The obtained candidate colloidal gold particles are matched with the original image to obtain the estimated position and size of the candidate colloidal gold particles.

[0057] Furthermore, the wavelet transform based on the Atrous algorithm is specifically as follows:

[0058] Step 1: Process the input image With transformation kernel Perform convolution processing to obtain the convolved image. ;

[0059] Step 2: Transform kernel Perform zero-padding operation between elements;

[0060] Step 3: Put and Perform the difference operation to obtain the wavelet image at this scale;

[0061] Step 4: Determine if the preset scale has been reached. If not, then... Assign to If yes, return to step 1; otherwise, proceed to step 5.

[0062] Step 5: Output the wavelet image at the last scale and the wavelet image at the first scale, wherein the wavelet image at the first scale is used for threshold calculation of wavelet image binarization.

[0063] Furthermore, the preset constraint specifically means that the pixel value of the center point of the cluster is lower than a preset first threshold, and the number of pixels in the cluster is less than twice the area corresponding to the initial radius value.

[0064] Furthermore, the step of matching the obtained candidate colloidal gold particle targets with the original image specifically involves: performing Gaussian fitting on the original image to obtain the center coordinates of each candidate colloidal gold particle; calculating the diameter of the cluster corresponding to each candidate colloidal gold particle; and obtaining a radius estimate of the candidate colloidal gold particle based on the diameter.

[0065] Furthermore, the step of performing template matching using the average template and the plurality of clusters specifically involves: calculating the normalized cross-correlation value between the average template and the candidate particles.

[0066] Furthermore, the clustering operation specifically involves taking the set of pixels corresponding to each target to be identified as a cluster, and performing clustering based on the MeanShift clustering method to obtain the coordinate set of each cluster and the center coordinates of the cluster.

[0067] Furthermore, the corresponding preprocessing specifically involves: performing an average filtering process on the electron microscope image to be identified to obtain a smooth electron microscope image.

[0068] Specifically, for ease of understanding, the following detailed description of the solution in this embodiment is provided in conjunction with the accompanying drawings:

[0069] To address the problems existing in the prior art, this embodiment provides a method for identifying and locating colloidal gold particles in cryo-electron microscopy images. This method mainly includes two stages: colloidal gold particle identification and colloidal gold particle localization. The specific process is as follows:

[0070] (a) Colloidal gold particle identification stage

[0071] (1) Image preprocessing: The image is processed by moving average to obtain a smoothed image. Moving average kernel as follows,

[0072]

[0073] Average kernel Move across the entire image range and with the kernel Convolution is performed on the covered region to obtain the final smoothed image. This process can be represented by the following formula:

[0074]

[0075] in, Indicates the core. Represents an image. The parameters are related to the core size:

[0076]

[0077]

[0078] To ensure that the image size remains unchanged before and after image preprocessing, boundary processing is required before convolution. Since the moving average uses a kernel size of... Therefore, a 1-pixel wide mirror overlay is required on the input image.

[0079] (2) Wavelet Transform

[0080] Perform wavelet transform based on the Atrous algorithm on the preprocessed image, such as Figure 3 The diagram illustrates the process of wavelet transform based on the Atrous algorithm, where the one-dimensional transform kernel of the wavelet transform is:

[0081]

[0082] After applying the transform kernel to each row of the image, and then to each column of the image, an approximate image after wavelet transform is obtained. Due to a certain scale wavelet transform output image It is the input image The result after removing high-frequency information is that the high-frequency information of the input image can be obtained by subtracting adjacent approximate images.

[0083]

[0084] The scheme described in this embodiment employs a multi-scale approach to acquire high-frequency information: at lower scales, the high-frequency information removed by wavelet transform is often noise or other non-interesting content; at higher scales, wavelet transform removes useful information from the high frequencies of the input image, which is the information of interest. Therefore, it is only necessary to calculate the difference between adjacent approximate images at higher scales.

[0085]

[0086] The difference between adjacent approximate images at each scale is the wavelet image at that scale, i.e.

[0087]

[0088] Furthermore, the wavelet image is processed to obtain candidate colloidal gold particles: Since the wavelet image is obtained by subtracting adjacent approximate images, the element values ​​of the wavelet image are floating-point data, and there will only be significant values ​​at the coordinates where the pixel values ​​change drastically in the original image. Therefore, the wavelet image is processed in the following order: binarization, denoising and clustering.

[0089] 1) Binarization:

[0090] In the binarization operation of wavelet images, the binarization threshold is set to a constant multiple of the variance of the initial wavelet image, i.e. , where the parameters The value is usually set to 2, but can be set according to actual needs; the wavelet image after binarization becomes a binary image with a value of 0 or 1.

[0091] 2) Noise reduction:

[0092] In the denoising of binary wavelet images, foreground elements with fewer than 2 pixels (i.e., each target to be identified) are treated as noise and removed.

[0093] 3) Clustering:

[0094] The remaining set of pixels corresponding to each foreground element is used as an initial cluster. Each cluster corresponds to a bright spot on the wavelet image, i.e., a candidate colloidal gold particle. The set of pixels corresponding to the cluster is the set of pixels that constitute that bright spot. The MeanShift clustering method is then used to cluster the wavelet image. The MeanShift algorithm requires an initial radius parameter. The value is usually determined by the relative size of the colloidal gold particles in the electron microscope image, with commonly used values ​​being 4 and 8. The specific value can be determined according to actual needs, resulting in the coordinate set of each cluster and the center coordinates of the cluster. Since the original image sub-image corresponding to each cluster obtained here is a candidate colloidal gold particle, for ease of description, it will be referred to as a candidate particle below.

[0095] Furthermore, a template is generated based on the candidate particle set, and candidate particles are selected based on a template matching algorithm, specifically:

[0096] Based on the characteristics of colloidal gold particles, candidate particles that satisfy the following constraints are more likely to be colloidal gold particles:

[0097] 1) The pixel value at the center point of the candidate particle is small enough (i.e., it meets the preset threshold);

[0098] 2) The area of ​​the candidate particles does not exceed .

[0099] Select a number of candidate particles that meet the above requirements (in this embodiment, the number is set to a maximum of 4) and take the average as the template for template matching.

[0100] The average template is matched with all candidate particles, i.e., the normalized cross-correlation (NCC) value between the average template and candidate particles is calculated.

[0101] The NCC value can be used to describe the correlation between two windows of the same dimension, and its value range is [value range missing]. Its value means the following: close to This indicates a negative correlation between the two windows; values ​​closer to 0 indicate no correlation between the two windows; values ​​closer to 0 indicate no correlation between the two windows. This indicates a positive correlation between two windows. Therefore, this indicator can be used to describe the similarity between two windows. The average template is compared with all candidate particles to calculate the NCC value. Each candidate particle corresponds to an NCC value. Candidate particles are filtered by setting a threshold. In the scheme described in this embodiment, the threshold is set to 0.75, retaining candidate particles with NCC values ​​greater than the threshold.

[0102] (II) Positioning of colloidal gold particles

[0103] The scheme described in this embodiment achieves particle localization by performing Gaussian fitting on the final candidate particle set, specifically as follows:

[0104] Since the pixel value distribution of colloidal gold particles is characterized by "low values ​​in the center and high values ​​around the edges," the image of colloidal gold particles can be inverted and approximated using a Gaussian function. The two-dimensional Gaussian function is shown below:

[0105]

[0106] in, It is the amplitude. These are the coordinates of the center point. It is the variance of the distribution, where the fitted parameters are... These are the coordinates of the center point corresponding to the candidate particle.

[0107] The radius of a candidate particle can be approximated as half the diameter of the cluster, and the diameter of the coordinate set is defined as...

[0108]

[0109] in, and These are the coordinates of the pixels in the cluster.

[0110] By calculating the diameter of the cluster coordinate set for each candidate particle Then the radius of the candidate particle is approximately... .

[0111] Example 2:

[0112] The purpose of this embodiment is to provide a system for identifying and locating colloidal gold particles in cryo-electron microscopy images.

[0113] A system for identifying and locating colloidal gold particles in cryo-electron microscopy images, comprising:

[0114] The data acquisition unit is used to acquire electron microscope images to be identified and perform corresponding preprocessing.

[0115] The wavelet transform unit is used to process the preprocessed image using wavelet transform based on the Atrous algorithm to obtain the corresponding wavelet image;

[0116] The wavelet image processing unit is used to perform binarization and noise point removal on the obtained wavelet image in sequence, and to perform clustering operation on each target to be identified in the binarized image after noise point removal to obtain several clusters corresponding to colloidal gold particles in the binarized image.

[0117] An average template acquisition unit is used to filter out clusters that satisfy preset constraints from the plurality of clusters, and obtain an average template based on the obtained clusters that satisfy the preset constraints.

[0118] The colloidal gold particle target acquisition unit is used to perform template matching between the average template and the plurality of clusters respectively, filter out clusters whose matching results are lower than a preset threshold, and use the remaining clusters as candidate colloidal gold particle targets.

[0119] The localization unit is used to match the obtained candidate colloidal gold particle targets with the original image to obtain the position and size estimate of the candidate colloidal gold particles.

[0120] Furthermore, the system described in this embodiment corresponds to the method described in Embodiment 1, and its technical details have been described in detail in Embodiment 1, so they will not be repeated here.

[0121] In further embodiments, the following is also provided:

[0122] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When executed by the processor, the computer instructions perform the method described in Embodiment 1. For brevity, further details are omitted here.

[0123] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0124] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0125] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in Embodiment 1.

[0126] The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0127] Those skilled in the art will recognize that the units, i.e., algorithm steps, of the various examples described in connection with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.

[0128] The above embodiments provide a method and system for identifying and locating colloidal gold particles in cryo-electron microscopy images, which can be implemented and has broad application prospects.

[0129] The above description is merely a preferred embodiment of this disclosure and is not intended to limit this disclosure. Various modifications and variations can be made to this disclosure by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for identifying and locating colloidal gold particles in cryo-electron microscopy images, characterized in that, include: Acquire the electron microscope image to be identified and perform appropriate preprocessing; The wavelet transform based on the Atrous algorithm is used to process the preprocessed image to obtain the corresponding wavelet image; The obtained wavelet image is sequentially binarized and noise points are removed. Then, each target to be identified in the binarized image after noise removal is clustered to obtain several clusters corresponding to the colloidal gold particles in the binarized image. Clusters that satisfy preset constraints are selected from the aforementioned clusters, and an average template is obtained based on the obtained clusters that satisfy the preset constraints. The average template is used to perform template matching with the several clusters respectively, and clusters with matching results lower than a preset threshold are filtered out. The remaining clusters are used as candidate colloidal gold particle targets. The obtained candidate colloidal gold particles are matched with the original image to obtain the estimated position and size of the candidate colloidal gold particles.

2. The method for identifying and locating colloidal gold particles in cryo-electron microscopy images as described in claim 1, characterized in that, The wavelet transform based on the Atrous algorithm is specifically as follows: Step 1: Process the input image With transformation kernel Perform convolution processing to obtain the convolved image. ; Step 2: Transform kernel Perform zero-padding operation between elements; Step 3: Put and Perform the difference operation to obtain the wavelet image at the scale; Step 4: Determine if the preset scale has been reached. If not, then... Assign to And return to step 1; If so, proceed to step 5; Step 5: Output the wavelet image at the last scale and the wavelet image at the first scale, wherein the wavelet image at the first scale is used for threshold calculation of wavelet image binarization.

3. The method for identifying and locating colloidal gold particles in cryo-electron microscopy images as described in claim 1, characterized in that, The preset constraints are specifically: the pixel value of the center point of the cluster is lower than a preset first threshold, and the number of pixels in the cluster is less than twice the area corresponding to the initial radius value.

4. The method for identifying and locating colloidal gold particles in cryo-electron microscopy images as described in claim 1, characterized in that, Gaussian fitting is performed on the original image to obtain the center coordinates of each candidate colloidal gold particle; the diameter of the cluster corresponding to each candidate colloidal gold particle is calculated, and the radius estimate of the candidate colloidal gold particle is obtained based on the diameter.

5. The method for identifying and locating colloidal gold particles in cryo-electron microscopy images as described in claim 1, characterized in that, The step of using the average template to perform template matching with the several clusters respectively involves calculating the normalized cross-correlation value between the average template and the candidate particles.

6. The method for identifying and locating colloidal gold particles in cryo-electron microscopy images as described in claim 1, characterized in that, The clustering operation specifically involves taking the set of pixels corresponding to each target to be identified as a cluster, and performing clustering based on the MeanShift clustering method to obtain the coordinate set of each cluster and the center coordinates of the cluster.

7. The method for identifying and locating colloidal gold particles in cryo-electron microscopy images as described in claim 1, characterized in that, The corresponding preprocessing specifically involves: performing an average filtering process on the electron microscope image to be identified to obtain a smooth electron microscope image.

8. A system for identifying and locating colloidal gold particles in cryo-electron microscopy images, characterized in that, include: The data acquisition unit is used to acquire electron microscope images to be identified and perform corresponding preprocessing. The wavelet transform unit is used to process the preprocessed image using wavelet transform based on the Atrous algorithm to obtain the corresponding wavelet image; The wavelet image processing unit is used to perform binarization and noise point removal on the obtained wavelet image in sequence, and to perform clustering operation on each target to be identified in the binarized image after noise point removal to obtain several clusters corresponding to colloidal gold particles in the binarized image. An average template acquisition unit is used to filter out clusters that satisfy preset constraints from the plurality of clusters, and obtain an average template based on the obtained clusters that satisfy the preset constraints. The colloidal gold particle target acquisition unit is used to perform template matching between the average template and the plurality of clusters respectively, filter out clusters whose matching results are lower than a preset threshold, and use the remaining clusters as candidate colloidal gold particle targets. The localization unit is used to match the obtained candidate colloidal gold particle targets with the original image to obtain the position and size estimate of the candidate colloidal gold particles.

9. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and running thereon, characterized in that, When the processor executes the program, it implements a method for identifying and locating colloidal gold particles in cryo-electron microscopy images as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements a method for identifying and locating colloidal gold particles in cryo-electron microscopy images as described in any one of claims 1-7.

Images