Plant cell morphology-based cell classification method and device, and electronic device

Abstract

The application discloses a plant cell morphology-based cell classification method and device and electronic equipment, and relates to the technical field of biological cell classification or other related technical fields, and the method comprises the following steps: segmenting a plant cell wall staining image obtained, obtaining a plurality of cell contour maps, extracting image features from each cell contour map, performing clustering processing on the image features, obtaining a cell type to which each cell belongs, and mapping the cell type to the cell contour map. The application solves the technical problem that a large amount of labeled data is required and a large amount of human resources is consumed when plant cells are classified in the related art.

Description

Technical Field

[0001] This invention relates to the field of biological cell classification technology or other related fields, specifically to a method, apparatus, and electronic device for classifying plant cell morphology cells. Background Technology

[0002] Cell classification is a crucial task widely applied in the fields of biology and genomics. Its goal is to classify, describe, and label cells in specific cell types or individual samples. The results of cell classification are of great significance for identifying and understanding different cell types, studying cell development and differentiation processes, revealing cell composition and function, and identifying potential abnormal cell types.

[0003] Plant cells exhibit significant differences in structure and morphology, which reflect their different functions and locations within the plant.

[0004] In related technologies, plant cell classification algorithms based on cell morphology mainly employ supervised learning methods, which require a large amount of labeled data and consume significant human and material resources. Furthermore, due to the diversity of cell types, it is necessary to continuously label new cell types, further increasing the workload and time cost.

[0005] There is currently no effective solution to the above problems. Summary of the Invention

[0006] This invention provides a method, apparatus, and electronic device for classifying plant cells based on plant cell morphology, which at least solves the technical problem that plant cell classification requires a large amount of labeled data and consumes a lot of human resources in related technologies.

[0007] According to one aspect of the present invention, a method for classifying cells based on plant cell morphology is provided, comprising: segmenting an acquired plant cell wall staining image to obtain multiple cell outline maps; extracting image features from each cell outline map, wherein the image features include at least: the position of the outline center point, the outline area, the outline perimeter, and the coordinates of the intersection point on the major axis of the outline; performing clustering processing on the image features to obtain the cell type to which each cell belongs, and mapping the cell type onto the cell outline map.

[0008] Optionally, the step of segmenting the acquired plant cell wall staining image to obtain multiple cell outline maps includes: acquiring the plant cell wall staining image through microscopic imaging; inputting the acquired plant cell wall staining image into a cell segmentation model, locating cell regions by the cell segmentation model, and segmenting the cell outline map from the cell regions, wherein the cell outline map includes at least: cell morphology and cell boundary.

[0009] Optionally, the step of extracting image features from each of the cell contour maps includes: deploying a new layer on a virtual layer in each of the cell contour maps; drawing cell contours on the new layer, wherein the center of the new layer coincides with the center point of the cell contour; extracting contour features of the cell contour, wherein the contour features include at least: center point position, contour area, and contour perimeter; extracting intersection coordinates at specified angle intervals using the contour major axis of each cell contour as the x-axis and the contour center point position as the origin, to obtain the intersection coordinates on the contour major axis; and summarizing the extracted contour center point position, contour area, contour perimeter, and intersection coordinates on the contour major axis into the image features.

[0010] Optionally, after extracting image features from each of the cell contour maps, the method further includes: normalizing the contour area, the contour perimeter, and the coordinates of the intersection points on the major axis of the contour.

[0011] Optionally, the step of clustering the image features to obtain the cell type to which each cell belongs includes: determining the nodes of the graph structure based on the center point of the contour, and determining the feature representation of each node in the graph structure based on the area of ​​the contour, the perimeter of the contour, and the coordinates of the intersection points on the major axis of the contour; constructing a graph structure based on the nodes and the feature representation of each node; and performing graph clustering based on the graph structure using a graph convolutional network to obtain the cell type to which each cell belongs.

[0012] Optionally, after constructing the graph structure based on the nodes and the feature representation of each node, the method further includes: constructing an undirected weighted graph: G = (V, E), where V represents the node and E represents the edge connecting different contour center points; constructing an adjacency matrix based on the undirected weighted graph, where the adjacency matrix includes the distance between each node; and constructing a feature matrix: X = [x_1, x_2, ..., x_n], where x_i represents the feature representation of the i-th node.

[0013] Optionally, the step of using a graph convolutional network to perform graph clustering based on the graph structure to obtain the cell type of each cell includes: multiplying the feature matrix with the weight matrix in the graph convolutional network and passing the output result to the activation function; updating the cell features: H^(l+1)=f(A,H^l,W^l), where H^l represents the feature vector of the l-th layer node in the graph convolutional network, W represents the weight matrix, A represents the adjacency matrix, and f is the activation function; and obtaining the cell classification result based on the updated cell features: Y=softmax(H^L*W^L), where L represents the feature vector output by the last fully connected layer in the graph convolutional network.

[0014] According to another aspect of the present invention, a plant cell morphology-based cell classification device is also provided, comprising: an image segmentation unit for segmenting an acquired plant cell wall staining image to obtain multiple cell contour maps; a feature extraction unit for extracting image features from each of the cell contour maps, wherein the image features include at least: the position of the contour center point, the contour area, the contour perimeter, and the coordinates of the intersection point on the contour major axis; and a cell classification unit for performing clustering processing on the image features to obtain the cell type to which each cell belongs, and mapping the cell type onto the cell contour map.

[0015] Optionally, the image segmentation unit includes: an image acquisition module for acquiring a stained image of plant cell walls through microscopic imaging; and an image input module for inputting the acquired stained image of plant cell walls into a cell segmentation model, wherein the cell segmentation model locates cell regions and segments the cell outline map from the cell regions, wherein the cell outline map includes at least: cell morphology and cell boundary.

[0016] Optionally, the feature extraction unit includes: a layer deployment module, configured to deploy a new layer on a virtual layer in each cell contour image, and draw cell contours on the new layer, wherein the center of the new layer coincides with the center point of the cell contour; a first extraction module, configured to extract contour features of the cell contour, wherein the contour features include at least: center point position, contour area, and contour perimeter; a second extraction module, configured to extract intersection coordinates at specified angle intervals using the contour major axis of each cell contour as the x-axis and the contour center point position as the origin, to obtain the intersection coordinates on the contour major axis; and a summarization module, configured to summarize the extracted contour center point position, contour area, contour perimeter, and intersection coordinates on the contour major axis into the image features.

[0017] Optionally, the plant cell morphology-based cell classification device further includes a normalization unit, used to normalize the area of ​​the outline, the perimeter of the outline, and the coordinates of the intersection points on the major axis of the outline after extracting image features from each cell outline image.

[0018] Optionally, the cell classification unit includes: a first determining module, used to determine the nodes of the graph structure based on the center point of the contour, and to determine the feature representation of each node in the graph structure based on the area of ​​the contour, the perimeter of the contour, and the coordinates of the intersection points on the major axis of the contour; a network graph construction module, used to construct a graph structure based on the nodes and the feature representation of each node; and a clustering module, used to perform graph clustering processing based on the graph structure using a graph convolutional network to obtain the cell type to which each cell belongs.

[0019] Optionally, the plant cell morphology-based cell classification device further includes: a weighted graph construction unit, used to construct an undirected weighted graph: G = (V, E) after constructing a graph structure based on the nodes and the feature representation of each node, where V represents the node and E represents the edge connecting different contour center points; a first matrix construction unit, used to construct an adjacency matrix based on the undirected weighted graph, where the adjacency matrix includes the distance between each node; and a second matrix construction unit, used to construct a feature matrix: X = [x_1, x_2, ..., x_n], where x_i represents the feature representation of the i-th node.

[0020] Optionally, the clustering module includes: a product submodule, used to multiply the feature matrix with the weight matrix in the graph convolutional network and pass the output result to the activation function; a feature update submodule, used to update cell features: H^(l+1)=f(A,H^l,W^l), where H^l represents the feature vector of the l-th layer node in the graph convolutional network, W represents the weight matrix, A represents the adjacency matrix, and f is the activation function; and a classification result determination submodule, used to obtain the cell classification result based on the updated cell features: Y=softmax(H^L*W^L), where L represents the feature vector output by the last fully connected layer in the graph convolutional network.

[0021] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform any of the above-described plant cell morphology-based cell classification methods.

[0022] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the plant cell morphology-based cell classification method described above.

[0023] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps of the plant cell morphology-based cell classification method described above.

[0024] In the present invention, the obtained plant cell wall staining image can be segmented to obtain multiple cell outline maps. Then, image features are extracted from each cell outline map, and the image features are clustered to obtain the cell type to which each cell belongs. The cell type is then mapped onto the cell outline map. The image features include at least: the position of the outline center point, the outline area, the outline perimeter, and the coordinates of the intersection point on the outline major axis.

[0025] Based on the above-mentioned disclosure, it is possible to classify cells by utilizing differences in plant cell morphology without the need for data annotation. By analyzing and comparing cell characteristics, morphological differences between different cell types can be discovered. By clustering image features, the cell type to which each cell belongs can be obtained, thereby solving the technical problem of requiring a large amount of labeled data and consuming a lot of human resources when classifying plant cells in related technologies. Attached Figure Description

[0026] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0027] Figure 1 This is a flowchart of an optional plant cell morphology-based cell classification method according to an embodiment of the present invention;

[0028] Figure 2 This is a schematic diagram obtained by segmenting the cell wall image of a Chinese cabbage root using a cell segmentation model, according to one embodiment of the present invention.

[0029] Figure 3 This is a flowchart of an optional cell classification method based on plant cell morphology characteristics according to an embodiment of the present invention;

[0030] Figure 4 This is a schematic diagram of an optional plant cell morphology-based cell classification device according to an embodiment of the present invention;

[0031] Figure 5 This is a hardware structure block diagram of an electronic device (or mobile device) that performs a plant cell morphology-based cell classification method according to an embodiment of the present invention. Detailed Implementation

[0032] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0033] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0034] To facilitate understanding of the present invention by those skilled in the art, some terms or nouns involved in the various embodiments of the present invention are explained below:

[0035] Unsupervised learning is a machine learning method that can automatically learn features and patterns from unlabeled data.

[0036] This invention employs an unsupervised learning algorithm to classify plant cells. Through training a deep learning model, it can learn latent representations of plant cell morphology without prior cell type labeling. The learned representations and features can then be used to classify new cell images. This unsupervised learning method fully utilizes large-scale unlabeled cell image data and reduces the need for labeled data and human resources. Furthermore, due to the strong generalization ability of unsupervised learning methods, the learned model can be applied to different cell types, thereby reducing the workload and time cost of labeling different cell types. This enables the completion of cell classification tasks based on plant cell morphology, addressing the issues of labeling data requirements and cell type variations, thus improving classification efficiency and reducing costs.

[0037] Plant cells exhibit significant differences in structure and morphology. For example, characteristics such as cell wall thickness, cell shape, cell size, and the distribution and morphology of organelles can be used to distinguish different cell types. Within plant tissues, different cell types typically possess specific morphological features. For instance, epidermal cells usually form dense epidermal tissue with thick cell walls that provide protection. Cells in vascular elements (such as xylem and phloem) have specialized morphologies; for example, vascular cells have hollow tubular structures used for the transport of water and nutrients. Furthermore, cell morphology is closely related to its function. For example, mesophyll cells in leaves are typically bead-shaped and contain numerous chloroplasts for photosynthesis. Apical cells in roots are usually long and have hair-like structures, facilitating the absorption of water and nutrients from the soil.

[0038] Plant cell morphology plays a crucial role in cell classification. Compared to other transcriptome sequencing technologies, plant cell morphology provides richer information. The importance of plant cell morphology in cell classification is reflected in the following aspects: 1. Reflection of cell type and function: The morphological characteristics of plant cells typically reflect their type and function. For example, different types of cells have different shapes, sizes, and organelle distributions. By observing the morphological characteristics of plant cells, the cell type and function can be preliminarily determined. 2. Cell identification and classification: Due to the variability in plant cell morphology, different types of cells usually have unique morphological characteristics. Cell morphology can be used to identify and classify plant cells. By observing characteristics such as cell shape, size, and cell wall thickness, cells can be divided into different types, and a cell classification system can be established to better understand the diversity and distribution of plant cells. 3. Reflection of biological development and environmental adaptation: The morphological characteristics of plant cells are regulated by biological development and environmental adaptation. During plant growth, cells undergo processes such as division, expansion, and differentiation, and their morphology changes accordingly. Observing changes in cell morphology can help understand the process and mechanism of cell development, as well as the plant's adaptability to different environmental factors. 4. Description of Spatial Distribution and Tissue Structure: The morphological characteristics of plant cells help describe the spatial distribution and tissue structure of cells within tissues. By observing cell morphology, arrangement, and intercellular interactions, we can understand the location and relationships of cells within tissues, thereby revealing the structure and function of tissues.

[0039] By observing and analyzing the morphological characteristics of plant cells, we can obtain information about cell types, functions, development, and environmental adaptations, thereby gaining a deeper understanding of the diversity and tissue structure of plant cells.

[0040] The following embodiments of the present invention can be applied to various systems / applications / devices based on plant cell morphology classification. Because plant cells vary greatly in morphology, structure, and function, the present invention employs a cell classification method based on plant cell morphology to group similar cells into one category, thereby providing a better understanding of plant cell diversity.

[0041] This invention utilizes an unsupervised cell clustering method to rapidly and accurately extract cell features from large-scale cell image data and perform cluster analysis. This algorithm is not only highly efficient and automated, but also capable of uncovering hidden patterns and structures within cell populations.

[0042] This invention proposes a cell classification scheme based on plant cell morphology, which can classify cells using differences in plant cell morphology without requiring data annotation. By analyzing and comparing cell morphological characteristics, we can discover morphological differences between different cell types and classify cells accordingly. The core idea of ​​this scheme is to use cell morphological characteristics as the basis for classification, rather than relying on labeled data.

[0043] The present invention will now be described in detail with reference to various embodiments.

[0044] Example 1

[0045] According to an embodiment of the present invention, an embodiment of a cell classification method based on plant cell morphology is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0046] The plant cell morphology-based cell classification method provided in this embodiment is applicable to plant cell data analysis software, grouping similar cells together to better understand plant cell diversity. Furthermore, this method has good generalization properties and can be applied to cell data from different plant species and under different experimental conditions.

[0047] Figure 1 This is a flowchart of an optional plant cell morphology-based cell classification method according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0048] Step S101: Segment the acquired plant cell wall staining image to obtain multiple cell outline maps.

[0049] Cell segmentation based on plant cell wall staining images is an important task, as it can extract accurate cell contour information, providing a foundation for subsequent cell analysis and research. Currently, cell segmentation models are applied to this task, including deep learning models such as DeepCell and CellPose.

[0050] Optionally, step S101 includes: acquiring a plant cell wall staining image through microscopic imaging; inputting the acquired plant cell wall staining image into a cell segmentation model, locating cell regions by the cell segmentation model, and segmenting cell outline maps from the cell regions, wherein the cell outline map includes at least: cell morphology and cell boundaries.

[0051] In this embodiment, when using a cell segmentation model for cell segmentation, appropriate models and parameter settings can be selected according to actual needs, and plant cell wall staining images can be input into the model for processing. The model will analyze and process the image, identify and segment the contour information of each cell. The resulting cell contour map can accurately describe the morphology and boundaries of the cells, providing a reliable foundation for subsequent tasks such as cell feature extraction, cell classification, and functional analysis.

[0052] It should be noted that the cell segmentation model used in this embodiment can utilize pre-trained weights to obtain high-quality cell segmentation results. In particular, the feature extraction part of the model can be used to extract features of the cell contours, forming meaningful feature vectors. By inputting the cell image into the trained or pre-trained cell segmentation model, the contour image of the cell can be obtained.

[0053] For example, taking data from bok choy roots as an example, Figure 2 This is a schematic diagram obtained by segmenting the cell wall image of a Chinese cabbage root using a cell segmentation model, according to an embodiment of the present invention. Here, the CellPose model can be used to segment the cell wall of a stained plant cell wall image.

[0054] Step S102: Extract image features from each cell contour map, wherein the image features include at least: the position of the contour center point, the contour area, the contour perimeter, and the coordinates of the intersection points on the contour major axis.

[0055] Optionally, the step of extracting image features from each cell contour map includes: deploying a new layer on a virtual layer in each cell contour map, drawing cell contours on the new layer, wherein the center of the new layer coincides with the center point of the cell contour; extracting contour features of the cell contour, wherein the contour features include at least: center point position, contour area, and contour perimeter; extracting intersection coordinates at specified angles at intervals with the contour major axis of each cell contour as the x-axis and the contour center point position as the origin, to obtain the intersection coordinates on the contour major axis; and summarizing the extracted contour center point position, contour area, contour perimeter, and intersection coordinates on the contour major axis into image features.

[0056] It should be noted that in this embodiment, when extracting cell features, the outline of each cell can be drawn separately on a new layer, and the center of the layer can be aligned with the center point of the cell outline to ensure that the outline of each cell is on a different layer for subsequent processing. Then, the center point, area, and perimeter of each cell outline are extracted as part of the feature vector. Afterward, with the major axis of each cell outline as the x-axis and the center point as the origin, the coordinates of the intersection points are taken at certain intervals and used as part of the feature vector.

[0057] Optionally, after extracting image features from each cell contour map, the method further includes normalizing the contour area, contour perimeter, and the coordinates of the intersection points on the contour's major axis.

[0058] Each set of information, except for the cell contour center point location, is normalized and mapped to a range between 0 and 1. This ensures that each feature has the same scale, preventing some features from having an excessive impact on the model. Table 1 below illustrates the extracted cell feature information.

[0059] Table 1 Cell Feature Extraction Table

[0060] Cell number Location perimeter area Intersection coordinates 1 (x,y) 10 35 (x1,y1),(x2,y2),(x3,y3)... 2

[0061] Through the above steps, feature vector information such as the center point, area, perimeter, and intersection coordinates on the long axis of each cell contour can be extracted and normalized for use in subsequent cell classification tasks. These features will help the model learn and recognize the morphological features of different cell types. It should be noted that the features extracted from the cell contour in this embodiment can be represented as a feature vector, where each element corresponds to a specific morphological feature. These feature vectors can be used for subsequent cell classification tasks, such as clustering, classification, or pattern recognition.

[0062] It should be noted that, in addition to extracting the feature vector information such as the center point, area, perimeter, and intersection coordinates on the major axis of the cell contour, this embodiment can also extract feature vectors such as the circumcircle, incircle, density ratio, circumcircle rectangle, and incircle matrix of the cell contour.

[0063] In extracting various features of cell contours, deep learning models can also be used for feature extraction. Here, pre-trained weights can be used. These pre-trained weights have been trained on large-scale datasets, thus possessing strong generalization ability and rich feature representation capabilities. By pre-training on large-scale datasets, the model can learn universal feature representations, rather than just task-specific features. This allows the pre-trained deep learning model to better capture the diversity and subtle differences in cell morphology during cell classification.

[0064] By quantifying and analyzing cell morphology, a series of morphological features, such as cell size, shape, and texture, can be extracted. Then, machine learning algorithms or clustering methods can be used to analyze and classify these morphological features.

[0065] Step S103: Cluster the image features to obtain the cell type of each cell, and map the cell type onto the cell outline map.

[0066] It should be noted that the clustering strategies mentioned in this embodiment include, but are not limited to, methods such as K-means, hierarchical clustering, Louvain clustering, Leiden clustering, and Gaussian mixture model clustering.

[0067] Optionally, the step of clustering the image features to obtain the cell type to which each cell belongs includes: determining the nodes of the graph structure based on the center point of the contour, and determining the feature expression of each node in the graph structure based on the contour area, contour perimeter, and the coordinates of the intersection points on the contour major axis; constructing the graph structure based on the nodes and the feature expression of each node; and using a graph convolutional network to perform graph clustering based on the graph structure to obtain the cell type to which each cell belongs.

[0068] This embodiment uses a graph convolutional network for graph clustering. With the advancement of Spatially Resolved Transcriptomics (SRT) technology, it has become possible to analyze feature expression using spatial information in tissues. Combining gene expression, spatial localization, and histological factors in SRT data analysis, this embodiment provides a SpaGCN graph convolutional network method. SpaGCN uses graph convolution to aggregate the feature expression information of each node from neighboring points, constructing a graph structure. Then, a graph convolutional network is used to perform graph clustering based on this graph structure to obtain the cell type of each cell.

[0069] In this embodiment, the center point of the cell contour is considered as spatial information within the tissue, and the feature information of the cell contour is used as the feature representation of each node. Based on the nodes and their feature representations, a graph structure is constructed. A graph convolutional network is then used to perform graph clustering based on the graph structure to obtain the cell type to which each cell belongs. In this embodiment, the center point of the cell contour provides the spatial location information of the cell within the tissue, while the image features extracted from the cell contour map (the contour area, contour perimeter, and coordinates of the intersection points on the long axis of the contour) determine the feature representation of each node in the graph structure.

[0070] Optionally, after constructing the graph structure based on the nodes and the feature representation of each node, the method further includes: constructing an undirected weighted graph: G = (V, E), where V represents the node and E represents the edge connecting different contour center points; constructing an adjacency matrix based on the undirected weighted graph, where the adjacency matrix includes the distance between each node; and constructing a feature matrix: X = [x_1, x_2, ..., x_n], where x_i represents the feature representation of the i-th node.

[0071] It should be noted that the spatial graph convolutional neural network used in this embodiment performs graph clustering based on the graph structure. Specifically, the spatial graph convolutional neural network is defined based on the spatial relationships between nodes. The center point of the cell contour is considered as spatial information within the tissue, and the feature information of the cell contour is used as the feature representation of each point, thereby constructing an undirected weighted graph. This undirected weighted graph representation encodes the relationships and morphological features between cell contours as edge and node attributes. By applying graph convolutional networks or other related algorithms, cell contours can be classified and analyzed.

[0072] Gene expression and histological image data are converted into an undirected weighted graph, G = (V, E). In this graph, each contour center point v represents a contour center point (represented as a node), and E represents the edge connecting different contour center points. Every two vertices are connected by an edge with a specific weight. It should be noted that when calculating the distance between each vertex, the distance between any two vertices reflects the relative similarity between the two corresponding points. This distance is determined by two factors: (1) the physical location of the two contour center points u and v in the tissue slice, and (2) the corresponding histological information of these two contour center points. Understandably, although some contour center points are physically close in the tissue, the histological image may show that they belong to different tissue layers. However, because the physical location of these points is very close, their histological information is unavailable (or unreliable) and cannot be classified by histological information. Therefore, these contour points need to be further classified accurately using the features of contour points in this scheme.

[0073] In the extended three-dimensional space, the coordinates of the contour center point v are defined as (xv, yv, z*v). The distance between any two distinct contour points u and v is calculated as:

[0074]

[0075] Optionally, the step of using a graph convolutional network to perform graph clustering based on the graph structure to obtain the cell type of each cell includes: multiplying the feature matrix with the weight matrix in the graph convolutional network and passing the output result to the activation function; updating the cell features: H^(l+1)=f(A,H^l,W^l), where H^l represents the feature vector of the l-th layer node in the graph convolutional network, W represents the weight matrix, A represents the adjacency matrix, and f is the activation function; and obtaining the cell classification result based on the updated cell features: Y=softmax(H^L*W^L), where L represents the feature vector output by the last fully connected layer in the graph convolutional network.

[0076] In graph convolutional networks, the weight matrix refers to the weight of each edge (u,v) that measures the correlation between points u and v, and is negatively correlated with their distance. The graph structure G is stored in an N×N adjacency matrix A = [w(u,v)], where the edge weights between points u and v are defined as follows:

[0077]

[0078] The hyperparameter *l*, also known as the feature length scale, determines the rate at which the weights decay as a function of distance. A similar function is used in SpatialDE24. Let *I* denote the identity matrix. For a contour center point *v*, the corresponding row sum of AI, denoted by *av*, can be interpreted as the relative contribution of other contour centers to its gene expression. By default, a value of *l* is chosen such that the mean of all contour centers equals a pre-specified value, such as 0.5. For data generated by SRT platforms with smaller tissue capture areas, such as SLIDE-seqV2, STARmap, and MERFISH, this embodiment selects a value of *l* that allows neighboring points / cells to contribute more information in gene expression clusters.

[0079] It should be noted that in the loss function, SpaGCN convolution uses the t-distribution as the kernel function to calculate the distance from each data point to each cluster center. Then, based on these distances, it calculates the soft assignment probability qij (predicted distribution), which is the probability that data point i is assigned to cluster j. An auxiliary target distribution P is defined to supervise qij and increase the weight of high-confidence clusters. Finally, KL divergence is used as the objective function to optimize the clustering process by minimizing the KL divergence loss, so that the distribution q predicted by the model is as close as possible to the auxiliary target distribution P.

[0080] Optionally, the step of mapping cell types to a cell outline map includes: using different colors to represent different cell types, and mapping the colors corresponding to the cell types to the cell outline map, wherein the same color is used for the same cell type.

[0081] Mapping cell classification results onto cell outline maps is an intuitive and effective way to display information, combining classification results with cell morphology to provide more comprehensive information. By displaying classification results on cell outline maps, this embodiment allows for intuitive observation and analysis of the characteristics and distribution of different cell populations.

[0082] In displaying cell outline maps, this embodiment can use different colors or markers to represent different cell types or functions based on the cell classification results. For example, cells with similar classification results can be marked with the same color or symbol to form visual clusters or groups on the cell outline map. This display method allows observers to quickly identify and compare the differences and similarities between different cell types.

[0083] Furthermore, by directly mapping the results of cell classification onto cell outline maps, this embodiment can spatially display the distribution of cell types or functions. Observing the distribution patterns of cells in tissues or organs helps to understand the spatial organization and interactions of cells. This is of great significance for studying plant growth and development, cell differentiation, and tissue function.

[0084] Displaying cell classification results on cell outline maps not only visually presents the classification information but also provides a visual reference for subsequent research and analysis. Researchers can gain an intuitive understanding of cell types, functions, and interrelationships by observing the displayed images. This method of presentation helps deepen the understanding of cell classification results and promotes in-depth research and exploration of plant cells.

[0085] The aforementioned plant cell morphology-based cell classification method first segments the acquired plant cell wall staining image to obtain multiple cell outline maps. Then, image features are extracted from each cell outline map, and clustering is performed on these features to determine the cell type of each cell. The cell type is then mapped onto the cell outline map. The image features include at least: the position of the outline center point, the outline area, the outline perimeter, and the coordinates of the intersection points on the major axis of the outline. In this embodiment, cell classification can be performed based on differences in plant cell morphology without the need for data annotation. By analyzing and comparing cell features, morphological differences between different cell types can be identified. Clustering of image features yields the cell type of each cell, thus solving the technical problem of requiring a large amount of labeled data and consuming significant human resources in related technologies for plant cell classification.

[0086] The following describes in detail another optional implementation method.

[0087] Figure 3 This is a flowchart of an optional cell classification method based on plant cell morphology characteristics according to an embodiment of the present invention, such as... Figure 3 As shown, this cell classification method includes:

[0088] The first step is to take a picture using a microscope.

[0089] Figure 2 This is a schematic diagram obtained by segmenting the cell wall image of a Chinese cabbage root using a cell segmentation model, according to an embodiment of the present invention. Figure 2 As shown, the cell outline map can be obtained by segmenting the plant cell wall staining image using the cell lpose model.

[0090] The second step is cell segmentation.

[0091] Cell segmentation is performed using a cell segmentation model. Plant cell wall staining images are input into the model for processing. The model analyzes and processes the images, identifying and segmenting the contour information of each cell. The resulting cell contour maps accurately describe the cell morphology and boundaries, providing a reliable foundation for subsequent tasks such as cell feature extraction, cell classification, and functional analysis.

[0092] The third step is feature extraction.

[0093] (1) Image feature extraction:

[0094] Draw the outline of each cell separately on a new layer, and align the center of the layer with the center point of the cell outline to ensure that the outline of each cell is on a different layer for subsequent processing.

[0095] The center point, area, and perimeter of each cell outline are extracted as part of the feature vector.

[0096] Using the long axis of each cell outline as the x-axis and the center point as the origin, the coordinates of the intersection points are taken at regular angles and used as part of the feature vector.

[0097] (2) Normalization: Normalize each set of information except for the cell outline center point position, mapping its range to between 0 and 1. This ensures that each feature has the same scale and avoids some features having too much influence on the model.

[0098] Through the above steps, feature vector information such as the center point, area, perimeter, and intersection coordinates on the long axis of each cell contour can be extracted and normalized for use in subsequent cell classification tasks. These features will help the model learn and recognize the morphological characteristics of different cell types.

[0099] The fourth step is cell clustering, including: K-means, hierarchical clustering, Louvain clustering, Leiden clustering, and Gaussian mixture model clustering.

[0100] Based on SRT technology and the concept of graph convolutional networks, this embodiment extends the SpaGCN method and applies it to cell contour classification. By treating the cell contour center point position as spatial information and the cell contour feature information as gene expression, the classification and analysis of cell contours can be achieved.

[0101] Step 5: Results Display.

[0102] Based on the cell classification results, different colors or labels are used to represent different cell types or functions. For example, cells with similar classification results can be labeled with the same color or symbol to create visual clusters or groups on the cell outline map.

[0103] Compared to traditional supervised learning methods, the cell classification scheme based on plant cell morphology provided in this embodiment has the following advantages: First, it does not require a large amount of labeled data, reducing the workload and cost of manual labeling. Second, classification based on differences in cell morphological features has good generalization performance and can adapt to different types of plant cell data. Furthermore, the cell classification scheme provided in this embodiment can also reveal the relationship between cell morphology and function, providing valuable clues for subsequent functional research.

[0104] Based on the above, a cell classification scheme based on plant cell morphology offers a novel approach to solving the cell classification problem with large amounts of plant data. It not only overcomes the dependence of supervised learning methods on labeled data but also enables effective classification and analysis using cell morphology features.

[0105] The following is a detailed description with reference to another embodiment.

[0106] Example 2

[0107] The plant cell morphology-based cell classification device provided in this embodiment includes multiple implementation units, each of which corresponds to a specific implementation step in Embodiment 1 above.

[0108] Figure 4 This is a schematic diagram of an optional plant cell morphology-based cell classification device according to an embodiment of the present invention, such as... Figure 4 As shown, the plant cell morphology-based cell classification device may include: an image segmentation unit 41, a feature extraction unit 42, and a cell classification unit 43.

[0109] The image segmentation unit 41 is used to segment the acquired plant cell wall staining image to obtain multiple cell outline maps.

[0110] The feature extraction unit 42 is used to extract image features from each cell contour map, wherein the image features include at least: the position of the contour center point, the contour area, the contour perimeter, and the coordinates of the intersection point on the contour major axis.

[0111] Cell classification unit 43 is used to perform clustering processing on image features to obtain the cell type to which each cell belongs, and to map the cell type onto the cell outline map.

[0112] The aforementioned plant cell morphology-based cell classification device can segment the acquired plant cell wall staining image using the image segmentation unit 41 to obtain multiple cell outline maps. Then, the feature extraction unit 42 extracts image features from each cell outline map, and the cell classification unit 43 performs clustering processing on the image features to obtain the cell type of each cell. The cell type is then mapped onto the cell outline map. The image features include at least: the position of the outline center point, the outline area, the outline perimeter, and the coordinates of the intersection points on the major axis of the outline. In this embodiment, cell classification can be performed using differences in plant cell morphology without the need for data annotation. By analyzing and comparing cell features, morphological differences between different cell types can be discovered. Clustering of image features yields the cell type of each cell, thus solving the technical problem of requiring a large amount of labeled data and consuming significant human resources in related technologies for plant cell classification.

[0113] Optionally, the image segmentation unit includes: an image acquisition module for acquiring plant cell wall staining images through microscopic imaging; and an image input module for inputting the acquired plant cell wall staining images into a cell segmentation model, whereby the cell segmentation model locates cell regions and segments cell outline maps from the cell regions, wherein the cell outline maps include at least: cell morphology and cell boundaries.

[0114] Optionally, the feature extraction unit includes: a layer deployment module for deploying a new layer on a virtual layer in each cell contour image and drawing cell contours on the new layer, wherein the center of the new layer coincides with the center point of the cell contour; a first extraction module for extracting contour features of the cell contour, wherein the contour features include at least: center point position, contour area, and contour perimeter; a second extraction module for extracting intersection coordinates at specified angles at intervals, with the contour major axis of each cell contour as the x-axis and the contour center point position as the origin, to obtain the intersection coordinates on the contour major axis; and a summarization module for summarizing the extracted contour center point position, contour area, contour perimeter, and intersection coordinates on the contour major axis into image features.

[0115] Optionally, the plant cell morphology-based cell classification device further includes a normalization unit, which normalizes the contour area, contour perimeter, and intersection coordinates on the long axis of the contour after extracting image features from each cell contour map.

[0116] Optionally, the cell classification unit includes: a first determination module, used to determine the nodes of the graph structure based on the center point of the contour, and to determine the feature representation of each node in the graph structure based on the contour area, the contour perimeter, and the coordinates of the intersection points on the long axis of the contour; a network graph construction module, used to construct the graph structure based on the nodes and the feature representation of each node; and a clustering module, used to perform graph clustering processing based on the graph structure using a graph convolutional network to obtain the cell type to which each cell belongs.

[0117] Optionally, the plant cell morphology-based cell classification device further includes: a weighted graph construction unit, used to construct an undirected weighted graph G = (V, E) after constructing a graph structure based on nodes and the feature expression of each node, where V represents a node and E represents an edge connecting different contour center points; a first matrix construction unit, used to construct an adjacency matrix based on the undirected weighted graph, where the adjacency matrix includes the distance between each node; and a second matrix construction unit, used to construct a feature matrix X = [x_1, x_2, ..., x_n], where x_i represents the feature expression of the i-th node.

[0118] Optionally, the clustering module includes: a product submodule, used to multiply the feature matrix with the weight matrix in the graph convolutional network and pass the output to the activation function; a feature update submodule, used to update cell features: H^(l+1)=f(A,H^l,W^l), where H^l represents the feature vector of the l-th layer node in the graph convolutional network, W represents the weight matrix, A represents the adjacency matrix, and f is the activation function; and a classification result determination submodule, used to obtain the cell classification result based on the updated cell features: Y=softmax(H^L*W^L), where L represents the feature vector output by the last fully connected layer in the graph convolutional network.

[0119] The above-mentioned plant cell morphology-based cell classification device may also include a processor and a memory. The image segmentation unit 41, feature extraction unit 42, cell classification unit 43, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize the corresponding functions.

[0120] The processor described above contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured, and cell classification based on plant cell morphology characteristics can be achieved by adjusting kernel parameters.

[0121] The aforementioned memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0122] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute any of the above-described plant cell morphology-based cell classification methods.

[0123] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the plant cell morphology-based cell classification method of any one of the above embodiments.

[0124] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the plant cell morphology-based cell classification method described in various embodiments of this application.

[0125] This application also provides a computer program product, including a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the plant cell morphology-based cell classification method described in various embodiments of this application.

[0126] Figure 5 This is a hardware structure block diagram of an electronic device (or mobile device) that performs a plant cell morphology-based cell classification method according to an embodiment of the present invention. Figure 5 As shown, an electronic device may include one or more ( Figure 5 The processor (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and memory 504 for storing data are illustrated using 502a, 502b, ..., 502n. In addition, it may include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a keyboard, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 5 The structure shown is for illustrative purposes only and does not limit the structure of the electronic device described above. For example, the electronic device may also include components that are more... Figure 5 The more or fewer components shown, or having the same Figure 5 The different configurations shown.

[0127] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0128] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0129] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0130] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0131] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0132] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0133] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for classifying cells based on plant cell morphology, characterized in that, include: The acquired plant cell wall staining images were segmented to obtain multiple cell outline maps; Image features are extracted from each of the cell contour maps, wherein the image features include at least: the position of the contour center point, the contour area, the contour perimeter, and the coordinates of the intersection point on the contour major axis; The image features are clustered to obtain the cell type of each cell, and the cell type is mapped onto a cell outline map.

2. The classification method according to claim 1, characterized in that, The steps for segmenting the acquired plant cell wall staining image to obtain multiple cell outline maps include: Obtain images of stained plant cell walls using a microscope; The obtained plant cell wall staining image is input into the cell segmentation model, which locates the cell region and segments the cell outline map from the cell region. The cell outline map includes at least: cell morphology and cell boundary.

3. The classification method according to claim 1, characterized in that, The step of extracting image features from each of the cell contour maps includes: A new layer is deployed on each of the cell outline maps in a virtual layer, and the cell outline is drawn on the new layer, wherein the center of the new layer coincides with the center point of the cell outline; Extract the contour features of the cell contour, wherein the contour features include at least: center point position, contour area and contour perimeter; Using the long axis of each cell contour as the x-axis and the center point of the contour as the origin, the coordinates of the intersection points are extracted at specified intervals to obtain the coordinates of the intersection points on the long axis of the contour. The extracted center point position of the contour, the area of ​​the contour, the perimeter of the contour, and the coordinates of the intersection points on the major axis of the contour are summarized into the image features.

4. The classification method according to claim 3, characterized in that, After extracting image features from each of the cell contour maps, the process further includes: The area of ​​the contour, the perimeter of the contour, and the coordinates of the intersection points on the major axis of the contour are normalized.

5. The classification method according to claim 1, characterized in that, The step of clustering the image features to obtain the cell type to which each cell belongs includes: The nodes of the graph structure are determined based on the center point of the contour, and the feature representation of each node in the graph structure is determined based on the area of ​​the contour, the perimeter of the contour, and the coordinates of the intersection point on the major axis of the contour. A graph structure is constructed based on the nodes and the feature representation of each node. A graph convolutional network is used to perform graph clustering based on the graph structure to obtain the cell type to which each cell belongs.

6. The classification method according to claim 5, characterized in that, After constructing the graph structure based on the nodes and the feature representation of each node, the process further includes: Construct an undirected weighted graph: G = (V, E), where V represents the node and E represents the edge connecting the center points of different contours; Based on the undirected weighted graph, an adjacency matrix is ​​constructed, wherein the adjacency matrix includes the distance between each node; Construct the feature matrix: X = [x_1, x_2, ..., x_n], where x_i represents the feature representation of the i-th node.

7. The classification method according to claim 5, characterized in that, The step of using a graph convolutional network to perform graph clustering based on the graph structure to obtain the cell type of each cell includes: The feature matrix is ​​multiplied by the weight matrix in the graph convolutional network, and the output is passed to the activation function. Update cell features: H^(l+1)=f(A,H^l,W^l), where H^l represents the feature vector of the l-th layer node in the graph convolutional network, W represents the weight matrix, A represents the adjacency matrix, and f is the activation function; Based on the updated cell features, the cell classification result is obtained: Y = softmax(H^L*W^L), where L represents the feature vector output by the last fully connected layer in the graph convolutional network.

8. A cell classification device based on plant cell morphology, characterized in that, include: The image segmentation unit is used to segment the acquired plant cell wall staining image to obtain multiple cell outline maps; The feature extraction unit is used to extract image features from each of the cell contour maps, wherein the image features include at least: the position of the contour center point, the contour area, the contour perimeter, and the coordinates of the intersection point on the contour major axis; The cell classification unit is used to perform clustering processing on the image features to obtain the cell type to which each cell belongs, and to map the cell type onto the cell outline map.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the plant cell morphology-based cell classification method according to any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the plant cell morphology-based cell classification method according to any one of claims 1 to 7.

Images