Method for segmentation, classification and batch processing of organelles in electron microscope image

Abstract

The invention discloses a method for segmentation, classification and batch processing of organelles in an electron microscope image, and the method comprises the steps: obtaining the three-dimensional image data of cells or biological tissues based on a body electron microscope, and carrying out the processing of the three-dimensional image data, and obtaining the instance segmentation data of the organelles. According to the attributes of the organelles, the instance segmentation data of the organelles are manually marked, and virtual path data composed of a plurality of connected nodes are obtained. And retrieving in the instance segmentation data of the organelle according to the virtual path data, and extracting a corresponding target organelle. According to the method, operations such as selection, association, indexing and batch processing can be performed on the image segmentation individual set by using the virtual path. The data processing complexity is simplified, the organelle extraction efficiency is improved, and regrouping and personalized analysis are flexibly performed on the internal structures of the cells based on the extraction result.

Description

Technical Field

[0001] The present invention relates to the technical field of cell image processing, and in particular to a method for segmenting, classifying and batch processing cell organelles in electron microscope images. Background Art

[0002] In the biomedical field, data analysis of high-throughput electron microscopy is one of the main technical bottlenecks. Existing technologies are mainly based on manual or artificial intelligence technology to draw or annotate cell contours. Artificial intelligence solutions mainly rely on the recognition of the cell outer membrane or cell contour from extracting cells to the corresponding organelles in the cell. However, for large-scale cells with irregular contours (such as neurons, etc.), the above methods have the problems of long manual labeling time, high cost, high computing power requirements, and large errors in the results, which still require a lot of manual verification in the later stage. There is currently no universally applicable and mature neural network model to achieve accurate and efficient extraction of such cells. In addition, based on the results of cell contour definition, it is not possible to flexibly regroup and personalize the internal cell structure in the later stage. Summary of the invention

[0003] The purpose of the present invention is to provide a method for segmenting, classifying and batch processing organelles in electron microscope images, so as to simplify the complexity of data processing, improve the efficiency of organelle extraction, and flexibly regroup and personalize the internal cell structure based on the extraction results.

[0004] In order to achieve the above objectives, the present invention is implemented by the following technical solutions:

[0005] A method for segmenting, classifying and batch processing organelles in electron microscope images, comprising:

[0006] Step S1: Acquire three-dimensional image data of cells or biological tissues based on a volume electron microscope, and process the three-dimensional image data to obtain instance segmentation data of organelles.

[0007] Step S2: manually labeling the target organelles in the target cells in the instance segmentation data of the organelles according to the organelle attributes, and obtaining virtual path data consisting of a plurality of connected nodes.

[0008] Step S3: searching the instance segmentation data of the organelle according to the virtual path data to extract the corresponding target organelle.

[0009] Optionally, the three-dimensional image data is processed by any one of a neural network-based image segmentation method, a threshold-based image segmentation method, a boundary-based image segmentation method and manual annotation to obtain instance segmentation data of the organelle.

[0010] Optionally, the step S1 includes: step S1.1, preparing a training sample set, wherein the training sample set includes three-dimensional image data of multiple cells or biological tissues.

[0011] Step S1.2: Input the training sample set into the residual 3D U-Net-based neural network for training to obtain a semantic segmentation model of organelles.

[0012] Step S1.3: input the three-dimensional image data containing the organelles to be extracted into the trained semantic segmentation model of the organelles to obtain the semantic segmentation data of the organelles.

[0013] Step S1.4, converting the semantic segmentation data of the organelle into instance segmentation data of the organelle by using a 3D watershed algorithm.

[0014] Optionally, in the organelle instance segmentation data, if the number of organelle instances is between 0 and 2 8 If the number of organelle instances is between 2 8 ~2 16 If the number of instances is between 2 16 ~2 32 The pixel segmentation number of each organelle is a 32-bit unsigned integer data, and different organelles have different pixel segmentation numbers, and the pixel segmentation number at any pixel spatial position of the same organelle is the same.

[0015] Optionally, the step S1.1 includes: intercepting a plurality of sub-blocks from image data collected by a volume electron microscope, and manually marking organelles on the sub-blocks as the training sample set.

[0016] Optionally, the step S2 includes: step S2.1, displaying the three-dimensional image data of the cell and the instance segmentation data of the organelle by layer overlay in image annotation software.

[0017] Step S2.2, manually annotating nodes of the target organelles in the target cells in the instance segmentation data of the organelles through the image annotation software to obtain a plurality of the nodes and the spatial coordinates of the nodes, connecting the plurality of the nodes in sequence according to the spatial coordinates of the nodes to obtain the virtual path data and store them.

[0018] Optionally, the step S3 includes: step S3.1, searching in the instance segmentation data of the organelle according to the spatial coordinates of the corresponding node, obtaining and saving the same pixel segmentation number corresponding to the spatial coordinates of the node, and the pixel set of the entire organelle associated with this pixel segmentation number.

[0019] The pixel set includes the pixel spatial positions of the entire organelle region and the pixel values ​​corresponding to the pixel spatial positions one by one.

[0020] Step S3.2: Reconstruct the organelle according to the plurality of identical pixel segmentation numbers and the pixel set to extract the target organelle.

[0021] Optionally, the step S3.2 includes: step S3.2.1, saving several identical pixel segmentation numbers extracted, and assigning all remaining pixel segmentation numbers different from this pixel segmentation number to zero, so as to achieve the extraction of the target organelle, and then use it for three-dimensional reconstruction of batch organelles.

[0022] Alternatively, the step S3.2 includes: step S3.2.2, saving the extracted plurality of identical pixel segmentation numbers and the pixel set of the entire organelle corresponding to the pixel segmentation number.

[0023] Step S3.2.3, establish a blank three-dimensional matrix, and fill each pixel value in the saved pixel set of the plurality of organelles into the blank three-dimensional matrix one by one according to the pixel spatial position, so as to extract the target organelles, and then use it for three-dimensional reconstruction of batch organelles.

[0024] Optionally, the organelles include: mitochondria of neuronal cells in brain tissue, neural synapses in brain tissue, mitochondria of cells in testicular tissue, cell nuclei in testicular tissue, mitochondria of hair cells in cochlear tissue and ribbon synapses, or a combination thereof.

[0025] In yet another aspect, the present invention further provides a readable storage medium, wherein the readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described above is implemented.

[0026] The present invention has at least the following technical effects:

[0027] The present invention can utilize the association between virtual path data and image segmentation individuals to achieve the latter's grouping / classification, reconstruction and quantification operations. That is, the image segmentation individuals rather than pixels are taken as the main body, and the virtual path is used as an efficient method for annotation. Compared with the existing manual pixel annotation, the present invention has higher processing efficiency; compared with the method of artificial intelligence-assisted automatic segmentation of cell contours (cell membranes), the present invention has lower computing power costs, better target selectivity, and flexible post-processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A schematic flow chart of a method for segmenting, classifying and batch processing organelles in electron microscope images according to the present invention;

[0029] Figure 2 A schematic diagram of a simulation of a method for segmenting, classifying and batch processing organelles in electron microscope images provided by one embodiment of the present invention;

[0030] Figure 3 This is a schematic diagram of the results of applying the present invention to extract mitochondria from neurons in brain tissue;

[0031] Figure 4 This is a schematic diagram of the results of applying the present invention to extracting neural synapses in brain tissue;

[0032] Figure 5 This is a schematic diagram of the results of applying the present invention to extract mitochondria from cells in testicular tissue;

[0033] Figure 6 This is a schematic diagram of the results of applying the present invention to extracting cell nuclei in testicular tissue;

[0034] Figure 7 It is a schematic diagram of the results of applying the present invention to extract mitochondria and ribbon synapses of hair cells in cochlear tissue. DETAILED DESCRIPTION

[0035] The following is a further detailed description of a method for segmenting, classifying and batch processing of organelles in electron microscope images proposed by the present invention in conjunction with the accompanying drawings and specific embodiments. According to the following description, the advantages and features of the present invention will be clearer. It should be noted that the accompanying drawings are in a very simplified form and all use non-precise proportions, which are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention. In order to make the purposes, features and advantages of the present invention more obvious and easy to understand, please refer to the accompanying drawings. It should be noted that the structures, proportions, sizes, etc. illustrated in the accompanying drawings of this specification are only used to match the contents disclosed in the specification, so that people familiar with this technology can understand and read, and are not used to limit the limiting conditions for the implementation of the present invention, so they have no technical substantive significance. Any modification of the structure, change in the proportional relationship or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope of the technical content disclosed by the present invention.

[0036] As described in the background technology, with the development of artificial intelligence technology, trained artificial neural networks can realize accurate recognition and pixel segmentation of all target structures in the data (such as organelles of a certain form), greatly reducing the need for human annotation. However, in order to further analyze the internal structure of cells and conduct morphological comparisons between cells, it is necessary to correspond the structural individuals obtained by automatic segmentation to their respective cells (or cell subregions), and on this basis, perform batch morphological characterization of cell structures. This is not achievable with current neural network models.

[0037] In view of this, the present invention provides a method for the segmentation, classification and batch processing of organelles in electron microscope images. Based on the image segmentation results of artificial intelligence, virtual paths are used to connect the effective segmented individuals one by one. The virtual paths can be used to select, associate, index and perform batch processing on the image segmentation individual set (for example, a certain type of organelle in the cell).

[0038] like Figure 1 and Figure 2 As shown, this embodiment provides a method for segmentation, classification and batch processing of organelles in electron microscope images, including: step S1, obtaining three-dimensional image data of cells or biological tissues based on a volume electron microscope, and processing the three-dimensional image data to obtain instance segmentation data of organelles.

[0039] Step S2: labeling the target organelles in the target cells in the instance segmentation data of the organelles according to the organelle attributes, and obtaining virtual path data consisting of a plurality of connected nodes.

[0040] Step S3: searching the instance segmentation data of the organelle according to the virtual path data to extract the corresponding target organelle.

[0041] This embodiment can use virtual path data (virtual path) to select, associate, index and perform batch processing on a set of image segmentation individuals (e.g., a certain type of organelle in a cell). This simplifies the complexity of data processing, improves the efficiency and accuracy of organelle extraction, and flexibly regroups and performs personalized analysis on internal cell structures based on the extraction results.

[0042] In this embodiment, the three-dimensional image data is processed in any of the following ways to obtain the instance segmentation data of the organelle: an image segmentation method based on a neural network, an image segmentation method based on a threshold, an image segmentation method based on a boundary, and other methods that can achieve segmentation, such as manual annotation. The present invention is not limited thereto.

[0043] In this embodiment or some other embodiments, step S1 includes:

[0044] Step S1.1: prepare a training sample set, wherein the training sample set includes three-dimensional image data of multiple cells or biological tissues.

[0045] Step S1.2: Input the training sample set into the residual 3D U-Net-based neural network for training to obtain a semantic segmentation model of organelles.

[0046] Step S1.3: input the three-dimensional image data containing the organelles to be extracted into the trained semantic segmentation model of the organelles to obtain the semantic segmentation data of the organelles.

[0047] In this embodiment, the semantic segmentation data of the organelles includes: the probability of the organelle mask and the organelle outline (values ​​are in the range of 0 to 1, including 0 and 1). If converted into an 8-bit image for storage, the pixel grayscale value is between 0 and 255, including 0 and 255.

[0048] Step S1.4: converting the semantic segmentation data of the organelle into instance segmentation data of the organelle by using a 3D watershed algorithm.

[0049] The difference between the instance segmentation data of the organelle and the semantic segmentation data of the organelle is that all organelles of the same type in the semantic segmentation data of the organelle have the same label (pixel segmentation number). For example, when the semantic segmentation data of the organelle is the semantic segmentation data of mitochondria, all mitochondria are assigned the same label. On this basis, the mitochondria of different cells will be retrieved by using the label for retrieval. It is necessary to perform a lot of screening work to find the mitochondria in the same cell, and it is impossible to effectively regroup them.

[0050] All organelles of the same type in the instance segmentation data of the organelles have different labels (pixel segmentation numbers). For example, when the instance segmentation data of the organelles is the instance segmentation data of mitochondria, all mitochondria are assigned labels with different values. Since a mitochondria occupies several pixels in the image, in order to facilitate the retrieval of the corresponding entire mitochondria, the pixels in each independent mitochondrial region are assigned a unique and consistent label, which is convenient for subsequent classification processing and other operations.

[0051] In the organelle instance segmentation data, if the number of organelle instances is between 0 and 2 8 If the number of organelle instances is between 2 8 ~2 16 If the number of instances is between 2 16 ~2 32 The pixel segmentation number of each organelle is a 32-bit unsigned integer data, and different organelles have different pixel segmentation numbers, and the pixel segmentation number at any pixel spatial position of the same organelle is the same.

[0052] In this embodiment, the step S1.1 includes: intercepting a plurality of sub-blocks from the image data collected by the volume electron microscope, and manually marking the organelles on the sub-blocks as the training sample set.

[0053] In this embodiment, the volume electron microscopy is, for example, a serial section scanning electron microscope, a focused ion beam scanning electron microscope, a continuous section scanning electron microscope, and a continuous section transmission electron microscope, but the present invention is not limited thereto.

[0054] The number of sub-blocks captured in the collected image data is, for example, 1 to 2 (for example, a size of 1000×1000×100 pixels), and the organelle structure is manually annotated, but the present invention is not limited thereto.

[0055] In this embodiment, the step S2 includes: step S2.1, displaying the three-dimensional image data (original electron microscope data) of the cell or biological tissue and the instance segmentation data of the organelle in the image annotation software by using layer overlay (for example, two layers can be overlaid or multiple layers can be overlaid).

[0056] The image annotation software is, for example, open source webKnossos software, which is applicable to various types of organelles and cell structures, but the present invention is not limited thereto.

[0057] Specifically, the original electron microscopy data and the organelle instance segmentation data are uploaded to the image annotation software, and two layers are used to display the electron microscopy data and the organelle instance segmentation data respectively, and the latter can be closed or set to transparency.

[0058] Step S2.2, manually annotating nodes of the target organelles in the target cells in the instance segmentation data of the organelles through the image annotation software to obtain a plurality of nodes (Node) and the spatial coordinates of the nodes, and connecting the plurality of nodes in sequence according to the spatial coordinates of the nodes to obtain the virtual path data and store them.

[0059] That is, a node is set up at any position in the instance segmentation data of each target organelle in turn to form a virtual path. That is, the virtual path is associated with the individual organelles it passes through, and is distinguished from the individual organelles that do not contain the spatial coordinates of the node (the organelle population that is not selected in the data).

[0060] In this embodiment, step S3 includes:

[0061] Step S3.1, search in the instance segmentation data of the organelle according to the spatial coordinates of the corresponding node, obtain and save the same pixel segmentation number corresponding to the spatial coordinates of the node, and the pixel set of the entire organelle associated with this pixel segmentation number.

[0062] The pixel set includes the pixel spatial positions of the entire organelle region and the pixel values ​​corresponding to the pixel spatial positions one by one.

[0063] Step S3.2: Reconstruct the organelle according to the plurality of identical pixel segmentation numbers and the pixel set to extract the target organelle.

[0064] Wherein, the step S3.2 comprises:

[0065] Step S3.2.1, save the extracted pixel segmentation numbers that are identical, and assign all other pixel segmentation numbers that are different from the pixel segmentation numbers to zero, so as to extract the target organelles, and then use them for three-dimensional reconstruction of batch organelles.

[0066] Alternatively, the step S3.2 comprises:

[0067] Step S3.2.2,

[0068] The extracted plurality of identical pixel segmentation numbers and the pixel set of the entire organelle corresponding to the pixel segmentation number are saved.

[0069] Step S3.2.3, establish a blank three-dimensional matrix, and fill each pixel value in the saved pixel set of the plurality of organelles into the blank three-dimensional matrix one by one according to the pixel spatial position, so as to extract the target organelles, and then use it for three-dimensional reconstruction of batch organelles.

[0070] That is, through the spatial coordinates of each node in the virtual path, the organelle segmentation data at the location can be quickly extracted, and the batch extraction of specific organelle groups (such as mitochondria belonging to the same cell) in the instance segmentation data of the organelle can be realized. Multiple virtual paths can be generated at the same time to define different organelle groups or subgroups.

[0071] All organelle instance segmentations containing nodes can be automatically associated using overlapping spatial coordinates, that is, the pixel value at the node corresponding to the assignment (label) of the organelle instance segmentation is retrieved, and the entire organelle can be retrieved by retrieving the pixel value with the assignment (label).

[0072] Combination Figure 3 to Figure 7 As shown, the organelles include but are not limited to: mitochondria in neuronal cells of brain tissue, neural synapses in brain tissue, mitochondria in cells in testicular tissue, cell nuclei in cells in testicular tissue, mitochondria in hair cells in cochlear tissue and ribbon synapses, or a combination thereof.

[0073] Please continue to refer to Figure 3 As shown, Figure 3 Figure (a) shows a schematic diagram of the process of extracting mitochondria of neurons in brain tissue using the method described in this embodiment, wherein according to the above step S1, the three-dimensional image data of neurons in brain tissue (the gray image part in Figure (a)) is processed into instance segmentation data of mitochondria of neurons in brain tissue (the color image part in Figure (a)). Figure 3 As shown in step S1 in Figure (a), aligned EM volume is formed and mitochondrial segmentation is achieved.

[0074] According to the above step S2, the node n is marked in the instance segmentation data of the mitochondria of the neurons in the brain tissue, and a virtual path (path) p is generated. According to the above step S3, the corresponding mitochondria are extracted according to the node n. That is, the instance is assigned to the cell (Instance assignment to cell), and the mitochondrial reconstruction (Mitochondrial reconstruction) is realized according to the virtual path.

[0075] Figure 3Figure (b) shows the original electron microscopy data of neurons in brain tissue (three-dimensional image data, shown in the gray three-dimensional stereogram in the figure, where the three-dimensional coordinate axes are x=260μm, y=230μm, z=160μm) and a schematic diagram of the three-dimensional structure of three neurons (Neuron-1~Neuron-3).

[0076] Figure 3 Figure (c) shows a schematic diagram of mitochondria (D, S, A) extracted from different compartments of neurons in brain tissue, where D represents dendritic mitochondria, S represents somatic mitochondria, and A represents axonal mitochondria. It can be seen that this embodiment can independently extract various types of mitochondria to facilitate further targeted analysis.

[0077] Please continue to refer to Figure 4 As shown, Figure 4 The middle figure (a) shows a 3D EM data block (3D EM stack) of the original EM data (3D image data) of the synapses in the brain tissue. Figure 4 The middle (b) figure shows the automatically reconstructed image data of the dense band of synapses in brain tissue (3D segmentation of synapses). Figure 4 In the middle (c) figure, synapses are assigned to corresponding dendrites by tracing the virtual neuron path (green lines). Figure 4 In the middle (d) figure, all synapses related to the dendrites (D1, D2) in the image are labeled. It can be seen that this embodiment can extract and classify the neural synapses in the brain tissue, which is convenient for analyzing the synapse as a cell organelle.

[0078] Please continue to refer to Figure 5 As shown, Figure 5 (a) shows the original electron microscopy data in testicular tissue. Figure 5 (b) shows the three-dimensional image data of testicular tissue (shown in the gray three-dimensional stereogram in the figure, where the three-dimensional coordinate axes are x=49.8μm, y=66μm, z=125μm) and the reconstruction schematic diagram of mitochondria in 12 testicular cells; Figure 5 (c) shows the mitochondrial network extracted from 12 testis cells.

[0079] Please continue to refer to Figure 6 As shown, Figure 6Figure (a) shows the schematic diagram of the identification of cell nuclei in the original electron microscopy data of testicular tissue. Figure 6 Figure (b) shows Figure 6 (a) Magnified view of the red box in the figure, examples of spermatogonium, spermatocyte, spermatoblast and Sertoli cell; Figure 6 Figure (c) shows the result of labeling and classifying the four types of cells by virtual paths. It can be seen that this embodiment can label and classify the above four types of cells, which is convenient for analyzing them.

[0080] Please continue to refer to Figure 7 As shown, Figure 7 Figure (a) shows the schematic diagram of automatic identification of mitochondria and ribbon synapses in the original electron microscopy data of cochlear tissue. Figure 7 The figure in the middle (b) shows the reconstruction of mitochondria and ribbon synapses in inner hair cells in cochlear tissue, and the results of associating ribbon synapses and their nearby mitochondria through virtual paths.

[0081] In other aspects, the present invention further provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the method described above is implemented.

[0082] In addition, the present invention further provides a readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the method described above is implemented.

[0083] In summary, this embodiment can utilize the association between virtual paths and image segmentation individuals to achieve the latter's grouping, reconstruction, and quantification operations. That is, the image segmentation individuals rather than pixels are taken as the main body, and the virtual path is used as an efficient direction for annotation. Compared with the existing manual pixel annotation, this embodiment has higher efficiency and accuracy; compared with artificial intelligence-assisted cell contour (cell membrane) automatic segmentation, this embodiment has lower computing power cost, better target selectivity, and flexible post-processing.

[0084] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0085] It should be noted that the devices and methods disclosed in the embodiments of this article can also be implemented in other ways. The device implementation described above is only schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of this article. In this regard, each box in the flowchart or block diagram can represent a part of a module, program or code, and the module, program segment or part of the code contains one or more executable instructions for implementing the specified logical function, and the module, program segment or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented by a dedicated hardware-based system for performing a specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0086] In addition, the functional modules in the various embodiments of this document may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0087] Although the content of the present invention has been described in detail through the above preferred embodiments, it should be appreciated that the above description should not be considered as a limitation of the present invention. After reading the above content, it will be apparent to those skilled in the art that various modifications and substitutions of the present invention will occur. Therefore, the protection scope of the present invention should be limited by the appended claims.

Claims

1. A method for segmenting, classifying and batch processing organelles in electron microscope images, characterized in that: include: Step S1, acquiring three-dimensional image data of cells or biological tissues based on a volume electron microscope, and processing the three-dimensional image data to obtain instance segmentation data of organelles; Step S2, manually labeling the target organelles in the target cells in the instance segmentation data of the organelles according to the organelle attributes, and obtaining virtual path data consisting of a plurality of connected nodes; Step S3: searching the instance segmentation data of the organelle according to the virtual path data to extract the corresponding target organelle.

2. The method for segmenting, classifying and batch processing organelles in electron microscope images according to claim 1, characterized in that: The three-dimensional image data is processed by using any one of a neural network-based image segmentation method, a threshold-based image segmentation method, a boundary-based image segmentation method and manual annotation to obtain instance segmentation data of the organelle.

3. The method for segmenting, classifying and batch processing organelles in electron microscope images according to claim 1, characterized in that: The step S1 comprises: Step S1.1, preparing a training sample set, wherein the training sample set includes three-dimensional image data of multiple cells or biological tissues; Step S1.2, inputting the training sample set into a neural network based on residual 3D U-Net for training to obtain a semantic segmentation model of organelles; Step S1.3, inputting the three-dimensional image data containing the organelles to be extracted into the trained semantic segmentation model of the organelles to obtain semantic segmentation data of the organelles; Step S1.4, converting the semantic segmentation data of the organelle into instance segmentation data of the organelle by using a 3D watershed algorithm.

4. The method for segmenting, classifying and batch processing organelles in electron microscope images according to claim 3, characterized in that: In the organelle instance segmentation data, if the number of organelle instances is between 0 and 2 8 If the number of organelle instances is between 2 8 ~2 16 If the number of instances is between 2 16 ~2 32 The pixel segmentation number of each organelle is a 32-bit unsigned integer data, and different organelles have different pixel segmentation numbers, and the pixel segmentation number at any pixel spatial position of the same organelle is the same.

5. The method for segmenting, classifying and batch processing organelles in electron microscope images according to claim 4, characterized in that: The step S1.1 includes: intercepting a plurality of sub-blocks from the image data collected by the volume electron microscope, and manually marking the organelles on the sub-blocks as the training sample set.

6. The method for segmenting, classifying and batch processing organelles in electron microscope images according to claim 5, characterized in that: The step S2 comprises: Step S2.1, displaying the three-dimensional image data of the cell and the instance segmentation data of the organelle by layer overlay in image annotation software; Step S2.2, manually annotating nodes of the target organelles in the target cells in the instance segmentation data of the organelles through the image annotation software to obtain a plurality of the nodes and the spatial coordinates of the nodes, connecting the plurality of the nodes in sequence according to the spatial coordinates of the nodes to obtain the virtual path data and store them.

7. The method for segmenting, classifying and batch processing organelles in electron microscope images according to claim 6, characterized in that: The step S3 comprises: Step S3.1, searching the instance segmentation data of the organelle according to the spatial coordinates of the corresponding node, obtaining and saving the same pixel segmentation number corresponding to the spatial coordinates of the node, and the pixel set of the entire organelle associated with the pixel segmentation number; The pixel set includes the pixel spatial positions of the entire organelle region and the pixel values ​​corresponding to the pixel spatial positions one by one; Step S3.2: Reconstruct the organelle according to the plurality of identical pixel segmentation numbers and the pixel set to extract the target organelle.

8. The method for segmenting, classifying and batch processing organelles in electron microscope images according to claim 7, characterized in that: The step S3.2 comprises: Step S3.2.1, saving a number of identical pixel segmentation numbers extracted, and assigning all other pixel segmentation numbers different from the pixel segmentation number to zero, so as to extract the target organelle, and then use it for three-dimensional reconstruction of batch organelles; Alternatively, the step S3.2 comprises: Step S3.2.2, saving the extracted plurality of identical pixel segmentation numbers and the pixel set of the entire organelle corresponding to the pixel segmentation number; Step S3.2.3, establish a blank three-dimensional matrix, and fill each pixel value in the saved pixel set of the plurality of organelles into the blank three-dimensional matrix one by one according to the pixel spatial position, so as to extract the target organelles, and then use it for three-dimensional reconstruction of batch organelles.

9. The method for segmenting, classifying and batch processing organelles in electron microscope images according to claim 1, characterized in that: The organelles include: mitochondria of neuronal cells in brain tissue, neural synapses in brain tissue, mitochondria of cells in testicular tissue, cell nuclei in testicular tissue, mitochondria of hair cells in cochlear tissue and ribbon synapses, or a combination thereof.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

Images